JP3992231B2 - NO2 detector - Google Patents

Description

[0001]
[Field of the Invention]
The present invention relates to NO2 detection, and more particularly, to outside air introduction control of a vehicle, detection of NO2 concentration during warming-up of the vehicle, detection of NO2 concentration in a parking lot, and the like.
[0002]
[Prior art]
[Patent Document 1]
Japanese Patent No. 3024217 discloses that a metal oxide semiconductor is used to detect the NO2 concentration in the outside air and control the introduction of the outside air into the automobile. Further, the detection range when introducing outside air into the automobile is a low concentration of about 0.3 to 3 ppm of NO2, and further, it is required to detect NO2 promptly from the point of application characteristics.
[0003]
[Problems of the Invention]
An object of the present invention is to enable quick and accurate measurement of low concentration NO2.
[0004]
[Structure of the invention]
The NO2 detection device of the present invention is obtained by means for periodically changing the heater power of the metal oxide semiconductor gas sensor, and means for frequency-converting the sensor signal in synchronization with the change of the heater power. And a regression analysis means for converting the frequency conversion value into the NO2 concentration by multiple regression analysis.
Here, the change period of the heater power is preferably 2 seconds or less, particularly preferably 1 second or less. The sensor preferably has a bead-shaped metal oxide semiconductor in which a coil-shaped electrode serving as a heater and a central electrode arranged at the center thereof are embedded.
[0005]
Preferably, a means for sampling and storing a reference value from the obtained frequency conversion value and a means for obtaining a difference value between the obtained frequency conversion value and the reference value are provided. The difference value is subjected to multiple regression analysis by the regression analysis means.
[0006]
Preferably, the regression analysis means obtains a logarithmic value of the NO2 concentration by the multiple regression analysis. For example, the obtained logarithmic value is exponentially converted into a NO2 concentration and output by D / A conversion or the like.
[0007]
Preferably, the frequency conversion is performed by short-time Fourier transform by multiplying the sensor signal by a window function for one period of temperature change centering on a region where the sensor temperature is low.
[0008]
[Operation and effect of the invention]
The inventors, for example, change the temperature of a metal oxide semiconductor gas sensor at a frequency of about 1 Hz, convert the response waveform by short-time Fourier transform, etc., and detect the low concentration of NO2 accurately by performing multiple regression analysis of the converted value. I found out that I can do it. Even if the metal oxide semiconductor gas sensor was driven at a frequency of about 1 Hz, the temperature change was too fast, so that the NO2 detection sensitivity was expected to decrease, but the NO2 concentration could be obtained accurately. Various values such as a neural network were considered for the conversion value processing, but the NO2 concentration could be obtained by multiple regression analysis. For this reason, the circuit cost could be suppressed to the extent that it can be installed in an outside air introduction control system of an automobile.
[0009]
The output waveform (response waveform) of the metal oxide semiconductor gas sensor at the time of temperature change was changed by the change of the sensor with time, ambient temperature, humidity, background gas, and the like. However, if the difference value from the reference waveform conversion value is used as the response waveform (reference waveform) in the air instead of the response waveform conversion value itself, the sensor signal at the start (when the engine is started), etc. It was tough and accurate.
[0010]
In the sensor signal processing, when the difference value was subjected to multiple regression analysis to obtain the logarithmic value of the NO2 concentration, highly accurate detection was possible. In other words, higher accuracy was obtained in the multiple regression analysis of the difference value and the logarithm of the concentration than the difference value and the NO2 concentration itself or the logarithm of the difference value and the NO2 concentration. The detection accuracy obtained in this way is such that the detected value falls within the range of about 1/3 to 3 times the true value in the range of about 0.1 to 3 ppm, for example, data for 6 months. It was. This is a very good value considering that the detection concentration is low and the measurement period is long. That is, detection of low concentration of NO2 using a metal oxide semiconductor gas sensor has been qualitative in the past, but it has become possible to detect low concentration of NO2 to a certain extent.
[0011]
In outside air introduction control of automobiles, it is important to shorten the response time. In normal FFT (Fast Fourier Transform), a response time of about 10 cycles is required. On the other hand, for example, when sensor signals for one cycle are processed by short-time Fourier transform, NO2 can be detected in about one second, for example.
[0012]
【Example】
1 to 8 show an NO2 detector according to an embodiment. In these figures, 2 is a power source and 4 is a metal oxide semiconductor gas sensor, for example, in which a coiled heater combined electrode and a center electrode are embedded in a bead of a SnO2 metal oxide semiconductor. SnO2 beads are almost spherical and have a diameter of about 400 μm. Further, a Pt-W wire having a diameter of about 15 μm was used for the heater combined electrode and the center electrode. The gas sensor 4 is used as a gas sensor by welding four terminals of a heater combined electrode and a center electrode to a stem or the like and covering with a cap provided with a metal mesh on the top. The structure of the gas sensor itself is arbitrary, and the material thereof is also arbitrary. Tr1 and Tr2 are switching transistors, which may be other switches, and RL is a load resistance. RH represents the heater resistance of the gas sensor 4, and RS represents the sensor resistance (resistance value of the metal oxide semiconductor).
[0013]
Reference numeral 6 denotes a microcomputer, and reference numeral 8 denotes a sensor drive, which controls the heater power of the gas sensor 4 by turning on / off the transistors Tr1 and Tr2, and also controls the timing of applying a detection voltage. That is, the heater power is changed in a sine wave shape with a cycle of 1 second (1 Hz), for example, and the sensor temperature is changed with a cycle of 1 second. The power waveform may be a square wave or a ramp wave. The maximum sensor temperature is 500 ° C., and the minimum value is about 150 ° C. The sensor drive 8 turns on the transistor Tr2 to apply the detection voltage to the series piece of the load resistance RL and the sensor resistance RS, and to add the output (sensor signal) to the AD converter 10, For example, the sensor signal for one period to the past five periods is Fourier transformed. Note that wavelet transform, discrete cosine transform, or the like may be used instead of the short-time Fourier transform. A reference value learning unit 14 stores a Fourier transform value as a reference value after a predetermined time, for example, 1 minute, after turning on the power of the NO2 detecting device.
[0014]
The NO2 detection device of the embodiment is used for outside air introduction control of a car, detection of NO2 and CO concentrations in a parking lot, or detection of NO2 and CO concentrations in a garage during a warm-up operation of a car. For example, the waveform of the sensor signal one minute after the start of the operation corresponds to the Fourier transform value of the sensor signal in the air, and NO2 is detected using the difference between the two. The calculation method of the reference Fourier transform value is arbitrary. For example, it is assumed that the signal in clean air is sampled on the condition that the waveform of the sensor signal is stable while the vehicle is running. The reference value may be used.
[0015]
A difference unit 16 calculates a difference value between the reference value of the Fourier transform value stored in the reference value learning unit 14 and the Fourier transform value newly obtained by the short-time Fourier transform unit 12. The Fourier transform value is a signal composed of a combination of a DC component and an AC sine or cosine component, and can be regarded as a kind of vector. In this case, it can be said that the difference unit 16 calculates the difference between the reference transformation value vector and the obtained Fourier transformation value vector.
[0016]
The multiple regression analysis unit 18 calculates the logarithmic value of the NO2 concentration by multiple regression analysis using the obtained difference vector. Then, the exponent conversion unit 20 exponentially converts the obtained logarithmic value to convert it into NO2 concentration, and the output unit 22 performs D / A conversion, for example, and outputs it as an analog output. By using the output of the output unit 22, on / off of outside air introduction of the automobile, air conditioning such as a parking lot, and air pollution due to long-time warm-up operation in the garage are detected. Further, since the output unit 22 can obtain an output corresponding to the analog value of the NO2 concentration, the concentration can be easily displayed.
[0017]
FIG. 2 schematically shows a waveform of a sensor signal accompanying a temperature change. The sensor resistance RS shows a maximum value in the vicinity of the lowest temperature of 150 ° C., and shows a minimum value in the vicinity of the highest temperature of 500 ° C. Of these, the NO2 sensitivity is highest near the lowest temperature. For example, as shown in the lower part of FIG. 2, a window function for Fourier-transforming the signal waveform for one period is used, and the value of the window function is multiplied by the sensor resistance. And Fourier transform. For example, a Hamming window may be used as the window function. In this way, a short-time Fourier transform can be performed with the signal waveform for the past one period, and NO2 can be detected in a short time. The window function is preferably a symmetric function on both sides of the low temperature region where the NO2 sensitivity is maximized.
[0018]
FIG. 3 shows a response waveform of the gas sensor 4 at about 3 ppm NO 2 in the air. With NO2 of about 3 ppm, the waveform characteristics are not so different from those in the air, and the waveform has no characteristics.
[0019]
FIG. 4 shows the NO2 detection algorithm. The temperature of the metal oxide semiconductor gas sensor is changed at an appropriate cycle. The frequency of the temperature change (reciprocal of the cycle) is, for example, 0.5 Hz or more, preferably 0.8 Hz to 4 Hz, and particularly preferably 1 Hz to 2 Hz. In the embodiment, the change width of the sensor temperature is 150 to 500 ° C. However, if the frequency of the temperature change is increased, the temperature change does not follow even if the heater power is changed, and the width of the temperature change decreases.
[0020]
Next, for example, using the waveform of the sensor signal for one cycle, short-time Fourier transform is performed using a window function that is symmetric about a low temperature region with high NO2 sensitivity. The obtained Fourier transform value is a kind of vector, and when emphasizing that it consists of a plurality of components, it may be called a Fourier transform spectrum.
[0021]
A reference Fourier transform spectrum is obtained in order to detect low concentrations of NO2 by canceling out the sensor change with time, the influence of ambient temperature and humidity, and other noise factors. In the embodiment, a Fourier transform spectrum at the time of engine start (such as one minute after the start of operation) is stored as a reference value (standard value). When NO2 is detected, a difference value between the obtained Fourier transform value and the reference value is obtained, and the difference value is converted into a logarithm of NO2 concentration by multiple regression analysis. In this way, the NO2 concentration is obtained, control such as outside air introduction control is performed, and the NO2 concentration is displayed.
[0022]
When the temperature of the gas sensor is changed, the influence of the surrounding wind is reduced. FIG. 5 shows the waveform of the sensor signal in the air when there is no wind and the waveform in the air when the wind speed is 7.6 m / sec. The two waveforms match and no wind effect is seen.
[0023]
FIG. 6 shows Fourier transform values when the waveform of FIG. 5 is subjected to a short-time Fourier transform. The information No. on the horizontal axis indicates each component of the Fourier transform value, information No. 1 is a direct current component, information No. 2 and 3 are 1 Hz components having the same frequency as the temperature change, and No. 4 and 5 are Components of 2 Hz and Nos. 6 and 7 are components of 3 Hz. An even component indicates a sine component and an odd component indicates a cosine component. Thus, FIG. 6 shows the change of the Fourier transform spectrum in the presence of wind with respect to the Fourier transform spectrum in the air when there is no wind. From FIG. 6, the influence of wind cannot be detected even if Fourier transform is performed. As described above, when NO2 is detected using the temperature change of the sensor, the influence of the wind is reduced. The reason why the influence of the wind is reduced by the temperature change is unknown.
[0024]
Next, a conversion method between the difference value from the reference value of the Fourier transform spectrum and the NO2 concentration was examined. Therefore, for example, for each of five sensors, the NO2 concentration was changed in the range of 0.3 ppm to 3 ppm, and measurement was performed 15 times each. Then, we examined what kind of method would give the best result by performing multiple regression analysis. If the difference value and the NO2 concentration are directly associated with each other, the coefficient of determination in the multiple regression analysis is about 0.9, and if the NO2 concentration is made to correspond to the logarithm of the difference value, the coefficient of determination is about 0.75, and the difference value and the NO2 concentration When the logarithm was made to correspond, the coefficient of determination was 0.999 or more. When the difference value and the concentration are both logarithmic, the coefficient of determination is about 0.95. Thus, accurate analysis could be performed by performing multiple regression analysis of the difference value to the logarithm of the NO2 concentration.
[0025]
It is inconvenient to determine multiple regression coefficients for each individual sensor. Therefore, for example, one lot of gas sensors may be regarded as sensors having the same characteristics, and a common multiple regression coefficient may be used for these gas sensors. Further, when the NO2 concentration in the high concentration range of 3 to 30 ppm was obtained using the multiple regression coefficient calibrated at 0.1 to 3 ppm, a low concentration was obtained. This is because the NO2 sensitivity decreases in the high concentration range. In order to solve this problem, it is preferable to increase the change frequency of the sensor temperature to 2 Hz and reduce the change width of the sensor temperature in the detection of NO2 of 3 to 30 ppm. In this way, the sensor temperature changed between a little over 200 ° C. and a little over 450 ° C., and the low temperature portion sensitive to the low concentration decreased, and as a result, the linearity with respect to the high concentration NO 2 increased.
[0026]
An example of multiple regression analysis shows that, for example, when quantifying 1 ppm of NO2, the contribution of the DC component is about 62%, the contribution of the 1 Hz component is about 27%, and the contribution of the 2 Hz and 3 Hz components is 3% each. An example was obtained in which the contribution of 4 Hz and 5 Hz was 2% each. When the regression coefficient is expressed as a relative value, the power spectrum is 5.5 for the 1 Hz component, 2.8 for the 2 Hz component, 9.2 for the 3 Hz component, and 9.2 for the 4 Hz component. It was 8.2 for the 5.2 Hz component and 1.2 for the 6 Hz component. In the embodiment, the DC component and the sine wave and the cosine wave up to the sixth harmonic are each six components, and a total of 13 components are used. However, the DC component and the eleven components up to the fifth harmonic may be used. In this case, a power spectrum from a direct current component to a third harmonic may be used.
[0027]
FIG. 7 shows the results when the NO2 concentration was obtained 20 times in two days using one gas sensor. Extremely high short-term reproducibility is obtained.
[0028]
FIG. 8 shows the detection results of NO2 concentration for about 200 days for five gas sensors. The NO2 concentration used for the measurement is 0.3, 1, 3 ppm, and the distribution width of the detected value obtained for each concentration is shown in FIG. The detection accuracy of NO2 concentration is extremely high.
[Brief description of the drawings]
FIG. 1 is a block diagram of the NO2 detector according to the embodiment. FIG. 2 is a characteristic diagram showing the sensor signal and the window function of the short-time Fourier transform in the embodiment. FIG. Fig. 4 is a characteristic diagram showing a response waveform. Fig. 4 is a flowchart showing an NO2 detection algorithm in the embodiment. Fig. 5 is a characteristic diagram showing the influence of wind on the sensor signal in the embodiment. Fig. 6 is a Fourier diagram of the sensor signal in Fig. 5. Characteristic diagram showing conversion spectrum [Fig. 7] Characteristic diagram showing short-term reproducibility of NO2 detection result (2 days / 20 measurements) in the embodiment [Fig. 8] Long-term reproducibility of NO2 detection value in the embodiment (Fig. 7) Characteristic diagram showing about 200 days) [Explanation of symbols]
2 Power supply 4 Metal oxide semiconductor gas sensor 6 Microcomputer 8 Sensor drive 10 AD converter 12 Short-time Fourier transform unit 14 Reference value learning unit 16 Difference unit 18 Multiple regression analysis unit 20 Exponential conversion unit 22 Output unit RH Heater resistance RS Sensor resistance RL Load resistance Tr1, Tr2 transistor

Claims (5)

Means for periodically changing the heater power of the metal oxide semiconductor gas sensor, means for frequency-converting the sensor signal in synchronization with the change of the heater power, and the obtained frequency conversion value by multiple regression analysis A NO2 detection device comprising a regression analysis means for converting to a NO2 concentration. 2. The NO2 detection device according to claim 1, wherein the change period of the heater power is 2 seconds or less. A difference value obtained by providing means for sampling and storing a reference value from the obtained frequency conversion value and means for obtaining a difference value between the obtained frequency conversion value and the reference value The NO2 detection device according to claim 1 or 2, wherein multiple regression analysis is performed by the regression analysis means. 4. The NO2 detection apparatus according to claim 3, wherein the regression analysis means obtains a logarithmic value of the NO2 concentration by the multiple regression analysis. The frequency conversion is performed by short-time Fourier transform by multiplying the sensor signal by a window function for one period of temperature change centering on a region where the sensor temperature is low. Any NO2 detector.

Images