JPH1183780A - Selection method of gas sensor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable selection of a gas sensor with a uniform waveform by selecting the gas sensor based on the analogy between a temperature waveform of a gas sensor signal and a reference waveform when the gas sensor is changed in temperature within a gas to be detected. SOLUTION: A pair of heaters h1 and h2 are arranged at both ends of a metal oxide semiconductor 2 of a gas sensor S. The temperature of the metal oxide semiconductor 2 is changed cyclically by varying the duty ratio of transistors T1 and T2 when they are ON. Then, a temperature waveform of a gas sensor signal is measured when the gas sensor S is changed in temperature within a gas to be detected to select the gas sensor S based on the analogy between the temperature waveform and a reference waveform. Preferably, the temperature waveform employs gas sensor signals at least at four points in the process of changes in the temperature. Preferably, the analogy is determined after the temperature waveform of the gas sensor signals and the reference waveform are both normalized. In the comparison of the temperature waveforms, a Fourier transform may be carried out.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to sorting of metal oxide semiconductor gas sensors.

[0002]

2. Description of the Related Art As a metal oxide semiconductor gas sensor using a temperature change, there is an SnO2 type CO sensor TGS203 or the like (TGS203 is a trade name of FIGARO GIKEN). This gas sensor operates at a cycle of 150 seconds, assigning the first 60 seconds to the high temperature area and the next 90 seconds to the low temperature area, the final temperature in the high temperature area is 300 ° C, and the final temperature in the low temperature area is 80 ° C.
CO is detected from the resistance value of the metal oxide semiconductor at the end of the low temperature range. The resistance value of the sensor is almost inversely proportional to the CO concentration, and the relative sensitivity between hydrogen and CO is 1:10.
0 ppm and 100 ppm of CO are equivalent. Further, the initial distribution of the resistance value is 1 to 10 KΩ in 100 ppm of CO.

[0003] The inventor has studied to improve the accuracy of the CO detection device using the TGS203 and improve the detection accuracy by a factor of two or more. The methodology used was to consider the TGS 203 as a complex system and obtain a highly reliable CO detection signal using the temperature waveform of the gas sensor signal. In this process, the inventor succeeded in eliminating the drift with time of the TGS 203 and compensating for the temperature and humidity dependence.

[0004] However, the above technique requires a new sorting method for the metal oxide semiconductor gas sensor, that is, a new inspection method at the time of shipment or when incorporated in a CO alarm. The selection criteria of the conventional TGS203 are that the relative sensitivity of CO to hydrogen in the last signal in the low temperature range is 10 times or more, and the resistance value in 100 ppm of CO is 1 to 10 KΩ. However, such selection criteria are inadequate. Since the temperature of the TGS 203 changes, a plurality of signals can be obtained by using the waveform of the sensor signal when the temperature changes.
The detection accuracy is improved. However, the conventional selection criterion focuses on only the last signal in the low temperature range. Further, the fact that the range of the sensor resistance in 100 ppm of CO is in the range of 1 to 10 KΩ has little relation to the characteristics of the gas sensor and merely lowers the production yield.

Here, related prior art is shown. It has been proposed by Yoshikawa et al. To change the temperature of a gas sensor and regard the behavior of its resistance value as a temperature waveform and to perform a Fourier transform on this to detect gas (Analytical Chemistry VoL68, No. 13, 2067-2072
1996). In addition, there are many studies on combining a high-temperature signal and a low-temperature signal of a gas sensor (for example, US Pat. Nos. 4,896,143 and 4,399,684).

[0006]

An object of the present invention is to provide a new sorting method for a metal oxide semiconductor gas sensor.

[0007]

According to the present invention, in a method for selecting a gas sensor including a heater and a metal oxide semiconductor having a resistance value changed by a gas, the temperature of a gas sensor signal when the temperature of the gas sensor is changed in a gas to be detected. It is characterized in that a waveform is measured and a gas sensor is selected based on the similarity between the temperature waveform and the standard waveform.

[0008] Preferably, at least four gas sensor signals in a temperature change process are used as the temperature waveform.
Preferably, the similarity is obtained after normalizing both the temperature waveform and the standard waveform of the gas sensor signal. Preferably, a generating function between the temperature waveform and the standard waveform is obtained, and a similarity between the temperature waveform of the gas sensor and the standard waveform is obtained based on the similarity between the generating functions. Preferably, a value of a correlation function between the temperature waveform and the standard waveform is obtained, and a similarity between the temperature waveform of the gas sensor and the standard waveform is obtained from the value of the correlation function.

[0009]

According to the present invention, the temperature of the metal oxide semiconductor gas sensor is changed in the gas to be detected, and the temperature of the gas sensor is sampled at preferably 4 points or more, more preferably 10 points or more to obtain a temperature waveform. . The obtained temperature waveform is compared with a standard waveform corresponding to a standard gas sensor whose characteristics have been confirmed in advance. Alternatively, an average waveform in the group of gas sensors to be selected is set as a standard waveform and compared with the standard waveform. Then, the gas sensor is selected based on the similarity between the measured temperature waveform and the standard waveform. In this way, a gas sensor having a uniform waveform can be selected, which is convenient when detecting a gas using a temperature waveform.

[0010] The number of sampling points required to determine the temperature waveform is related to the information amount of the gas sensor. For example, the TGS 203 used in the embodiment includes four types of different signals of the late stage of the low-temperature region, the early stage of the high-temperature region, the late stage of the high-temperature region, and the early stage of the low-temperature region. Can be. Therefore, sampling at four points is T
This is a condition for expressing the temperature waveform of the GS 203. For gas sensors with more complex temperature waveforms, the number of required sampling points increases.

In comparing the temperature waveforms, the resistance value of the gas sensor itself is not so important. Therefore, in order to mask and compare the difference in resistance value, it is preferable to normalize the resistance value of both the temperature waveform and the standard waveform.

In the comparison of the temperature waveforms, the temperature waveform or the standard waveform may be converted into a generating function such as Fourier transform or Laplace transform for comparison. In the generating function, the data is compressed as compared with the original temperature waveform. However, when a generating function is used, it is necessary to sample the temperature waveform at many points, for example, at 10 or more points, and preferably at 20 or more points for conversion into the generating function.

In the comparison of the temperature waveforms, a value of a correlation function with a standard waveform may be obtained. The value of the correlation function indicates the degree of similarity between the measured temperature waveform and the standard waveform. For example, a generating function cannot be obtained by sampling four points, but a value of the correlation function can be obtained.

The similarity between the standard waveform and the temperature waveform indicates the overall similarity of characteristics between the standard gas sensor and the gas sensor to be selected, whether it is virtual or real. Therefore, according to the present invention, it is possible to provide a selection method suitable for a gas sensor using a temperature waveform by selecting overall characteristics of gas sensor characteristics.

[0015]

【Example】

[0016]

[Structure of Gas Detector] FIG. 1 to FIG. 36 show an embodiment.
FIG. 1 shows the structure of the developed gas detecting device. S is a metal oxide semiconductor gas sensor, in which a TGS203 is used and a pair of heaters h1 and h2 are arranged at both ends of a SnO2 based metal oxide semiconductor 2. It is. The type, structure, material, and detection target gas of the sensor S are arbitrary. 4 is DC 5
The gas detection device is driven by the output VDD using a DC power supply such as V. In order to drive the pair of heaters h1 and h2 of the gas sensor S together, transistors T1 and T2 are used, and these are turned on / off at the same time. And the transistor T
When both T1 and T2 are turned on, current flows through the heaters h1 and h2, and the temperature of the metal oxide semiconductor 2 is periodically changed by changing the on duty ratio of the transistors T1 and T2. Here, in accordance with the operating conditions of the TGS 203, the high-temperature range is set to 60 seconds and the low-temperature range is set to 90 seconds. The waveform of the heater power is a square wave that changes in two stages, a high-temperature range and a low-temperature range. The final temperature in the low temperature range is 80 ° C. In the embodiment, the time is displayed as
The 0th second immediately before the end of the low temperature range is defined as the 0th second, the 0th to 60th second is defined as the high temperature range, and the 90th to 150th second (the 150th second is equal to the 0th second) is defined as the low temperature range. However, the heater waveform is not limited to a square wave, but may be a sine wave or a sawtooth wave.

A resistance ladder 5 is connected to the metal oxide semiconductor 2, and R1 to Rn are individual resistors. Here, it is assumed that each of the resistors R1 to Rn changes four times, for example, 0.5KΩ, 2KΩ, 8KΩ, 32KΩ, 128KΩ,
Six resistors of 512 KΩ are used. The accuracy of the fixed resistor is ±
An AD conversion error of about 2% is easily obtained, and an AD conversion error based on the switching of the resistance value is about ± 2%. When the transistors T1 and T2 are turned off, the power supply output VDD (hereinafter, referred to as a detection voltage Vc) flows to the resistance ladder 5 via the metal oxide semiconductor 2, and the output voltage to the resistance ladder 5 is A / D converted and processed.

Reference numeral 8 denotes a microcomputer.
Assume a one-chip microcomputer of bits. 1
Reference numeral 0 denotes the bus, reference numeral 12 denotes, for example, an 8-bit AD converter, reference numeral 14 denotes a resistor ladder control unit, which grounds only one of the resistors R1 to Rn, and uses the grounded resistor as a load resistor. As described above, the output voltage to the resistor ladder is AD
AD conversion is performed by the converter 12. It is natural that the output voltage to the resistance ladder 5 may be further divided and A / D-converted.
The same is true even if the voltage on the side is AD-converted. Reference numeral 16 denotes a heater control unit that controls on / off of the transistors T1 and T2 to generate a temperature cycle including a high-temperature region for 60 seconds and a low-temperature region for 90 seconds. Reference numeral 18 denotes an EEPROM control unit.
0 is an EEPROM.

FIG. 3 shows the structure of the EEPROM 20.
For example, here, CO is a detection target, and the detection range is CO5.
The range is about 10 times of 0 to 600 ppm. As the reference signal, three points of 65 ppm, 200 ppm, and 400 ppm of CO are used, and the logarithm LnR0,
The logarithm LnR6 of the sensor resistance at the 6th second and the logarithm LnR69 of the sensor resistance at the 69th second (the initial stage of the low temperature range) are used. Ln represents a natural logarithm, and a subscript such as 0 of R0 represents a timing based on 0 seconds. Similarly, at 200 ppm and 400 ppm of CO, the three reference signals at the 0th, 6th and 69th seconds are stored in the form of the logarithm of the sensor resistance. Cards 51 to 53 are used when the reference signal for each density is considered as one card. In addition, the card 54 is used to record the usage history of the CO detection device. That is, a record of the total use time and the past CO warning is stored as the elapsed time. The total usage time is a cumulative value of the time during which the power supply of the CO detection device is turned on. For example, the time unit is one day, and the cumulative usage time is stored in the card 54. As a record of the alarm, the date is stored each time a buzzer described later sounds.
As the date, a date based on the same standard as the total use time is stored. In this way, the day when the buzzer sounds is known.

Reference numeral 22 denotes an input / output, to which an adjustment switch 23 and a reset switch 24 are connected.
Is turned on, the EEPROM control unit 18
This switch is used only when adjusting the CO detection device. Reset switch 24
Is a switch for stopping the buzzer 38 from sounding.

The microcomputer 8 has a 4-bit arithmetic and logic operation unit 26, a sequence control unit 28 for operating the CO detection device at a period of 150 seconds, and the sequence control unit 28 has a built-in timer. . 3
Reference numeral 0 denotes a RAM, which is used as a volatile memory, and its configuration is shown in FIG. In the RAM 30, LnR0, LnR6, LnR6
Nine measurement data and reference signals at two concentrations are stored. As the reference signal, 65 ppm and 200 ppm on the low concentration side are always used, and when the gas concentration exceeds 200 ppm, the 65 ppm reference signal is replaced with a 400 ppm reference signal. When the gas concentration drops to 200 ppm or less, the reference signal of 400 ppm is replaced with the reference signal of 65 ppm. The gas detection range is 50-600 ppm and 5
The range of 0 to 65 ppm is close to the reference signal 65 ppm. Also 4
The range of 00 to 600 ppm is 1.5 times the range of 400 ppm of the reference signal, and the gas concentration can be accurately obtained using the 400 ppm reference signal. Excluding these ranges, when CO is generated, the gas concentration is determined by interpolation between the two reference signals using the reference signals on both sides of the actual CO concentration.

The RAM 30 also stores the obtained CO concentration, COHb (CO hemoglobin concentration in blood) converted from the CO concentration, and other auxiliary signals (for example, time data for configuring a timer on a daily basis). And so on.

Returning to FIG. 1, reference numeral 32 denotes an alarm control unit which operates the LEDs 39 and 40 via the drive circuit 36, and
The buzzer 38 sounds when the hemoglobin concentration is, for example, 5% or more. When the buzzer 38 sounds, the EEPROM control unit 18 writes the date of the alarm on the card 54. Reference numeral 34 denotes a program memory in which various constants and other data used for temperature correction are recorded. These data are common fixed data even when the sensor S changes. All data for each sensor is recorded in the EEPROM 20. 42 is a thermistor for measuring the ambient temperature, and 44 is a temperature / humidity correction unit.

[0024]

[Sampling and Logarithmic Conversion by Detector] FIG. 4 shows an average temperature waveform of ten sensors. If the sampling point used in the embodiment is indicated by a circle in the waveform of CO 100 ppm, sampling is performed at 150 seconds, 6 seconds, and 69 seconds. The resistance value of the sensor changes about 10 times in the range of 30 ppm to 300 ppm of CO, and the resistance value changes about 10 times between 0 second and 69 seconds. If a variation in sensor resistance, a change in ambient temperature and humidity, and the like are added to the above, the range of the AD conversion is about 0.5 to 500 KΩ in resistance value. So in this range A
The resistances R1 to Rn are set to 0.5 KΩ or more so that D conversion can be performed.
The output voltage VRl is changed to a resistance ladder immediately before each sampling timing, and the load resistance is switched according to the value. A of VRl
The D conversion itself can be performed within one second, and it is sufficient to determine which resistor is used at each sampling point according to the value at that time.

FIG. 5 is an enlarged view of the initial temperature waveform in the high temperature region for another 10 sensors. Atmosphere is 0 ° C, relative humidity 96%, 20 ° C 65%, 50 ° C
40%, the range of ± 2δ (δ is the standard deviation) and the average value are shown. Although the gas concentration is 100 ppm CO, the resistance value changes slightly less than 10 times at each timing due to fluctuations in ambient temperature and humidity. Further, the resistance values at the 0th and 6th seconds are substantially equal. For example, the same load resistance as at the 0th second may be used at the 6th second. However, preferably, for example, a resistance value of 0 second (or 149 second to ensure sampling before shifting to a high temperature region) is determined from a signal of 148 seconds, and a resistance value of 6 seconds from a resistance value of 5 seconds is determined. Determine the load resistance of
Similarly, the load resistance at 69 seconds is determined from the resistance value at 68 seconds.

FIG. 6 shows a sampling algorithm. When the time reaches 148 seconds, the output voltage is AD-converted, and this value is 1/3 to 2 of the detection voltage Vc (same as VDD).
Confirm that it is within the range of / 3. In this range, the ratio of the resistance value between the sensor resistance and the load resistance is in the range of 2: 1 to 1: 2. If the output voltage is correct, the output voltage is kept as it is. If the output voltage is not correct, the load resistance is switched so that the output voltage falls within this range. Next, when it reaches the 0th second, the output voltage is AD-converted, and the output voltage VRl obtained by AD conversion is used to obtain 0 according to the equation (1).
Find the logarithm of the sensor resistance at the second. Similarly, at 5 seconds, it is checked whether the value of the load resistance is correct, and the logarithm of the sensor resistance at 6 seconds is obtained. Further, it is checked whether the value of the load resistance is correct at the 68th second, and the logarithm of the sensor resistance is obtained at the 69th second.

LnR = 2−4VRl / Vc + LnRl (1) When the logarithm of the sensor resistance is approximated to the first-order term as in the equation (1), R / Rl is 1, the error is 0, and R / Rl is 1 /. 2 or 2, the error is 2%, and R / Rl is 1/3 or 3, the error is 11%. In the embodiment, since the purpose is to detect the CO concentration with an error of ± 20% or less, the error of ± 10% is too large. Therefore, the resistance ladder 5 is controlled so that the ratio between the sensor resistance and the load resistance is maintained in the range of 2 to 1/2 at three points of 0 second, 6 seconds, and 69 seconds.

The conversion from VR1 to the logarithm of the sensor resistance according to equation (1) is a linear conversion, and is a very simple conversion. However, six load resistors were required accordingly. In order to reduce the number of load resistors to, for example, four, the range of R / Rl is maintained in the range of 4 to 1/4, more preferably in the range of route 8 to 1 / route 8. For this purpose, conversion up to the third order term is necessary. When the logarithm of the sensor resistance is series-expanded by VRl, there are no second-order terms, and equations (2) and (3) take into account the third-order terms. When equations (2) and (3) are used,
When R / Rl is 1, the conversion error is 0%, when R / Rl is 1/4 or 4, the conversion error is 4%, and when Rl is 1/3 or 3, the conversion error is 2%. Therefore, for example, the values of the resistors R1 to Rn are set to 16
It is changed by a factor of two, more preferably by a factor of eight or nine. For example, when the values of the resistors R1 to Rn are 1KΩ and 8K
Ω, 64 KΩ, and 512 KΩ. In this way, the range of 0.5 to 1 MΩ can be converted to logarithm with an error of 2% or less. LnR = 2x + 2/3 × x 3 + LnRl (2) x = 1-2VRl / Vc (3)

[0029]

[Adjustment of Gas Detection Device] FIG. 7 shows a procedure for adjusting the gas detection device of FIG. At this time, the adjustment switch 23 is turned on to enable the writing of the reference signal to the EEPROM 20. In the following description, it is assumed that the CO detection device is set in the adjustment tank, and after the detection device is set, the power is turned on to operate. Then, for example, 65 ppm of CO is injected. Then, the microcomputer 8 generates LnR0, LnR6, and LnR69 for writing into the RAM 30. This is written on the card 51 of the EEPROM 20. The same procedure is then followed, increasing the CO concentration to 200 ppm.
Further, the CO concentration is increased to 400 ppm. If the CO concentration is increased in a predetermined procedure as described above, the EEPROM
20 can be written with a reference signal. As a result, there is no need to adjust the variable resistance to store the reference signal, and the adjustment operation is simplified.

Here, the CO detector is set in the adjustment tank, but only the sensor S may be set. Then, the resistance value of the sensor S may be A / D converted by, for example, an AD converter of about 12 bits, recorded in a personal computer or the like, and written in the EEPROM 20.
In this case, the sensor S is not incorporated in the CO detection device, but the sensor S is handled as a set with the EEPROM 20, and these are assembled into a separately assembled CO detection device. The parts other than the sensor S and the EEPROM 20 can be handled in exactly the same manner as a normal electronic circuit, and even a manufacturer having no experience with a gas sensor can assemble a CO detection device.

[0031]

[Drift of Gas Sensor Signal] FIGS. 8 to 12 show drift characteristics of the sensor resistance. This is 45 data of TGS203, including defective (7), good (20), samples left for more than 2 years (8), and samples (1) collected once set in the CO detector.
0). In each figure, the horizontal axis represents the sensor resistance at 0 seconds on a logarithmic scale, and the vertical axes represent the 6th second (FIG. 8), the 12th second (FIG. 9), the 30th second (FIG. 10), and the 60th second (FIG. 11),
The sensor resistance at 120 seconds (FIG. 12) is likewise shown on a logarithmic scale. And 1 on the horizontal axis is CO100pp at 0 second.
In the reference signal in the m (the third day of the start of energization), 1 is 6 on the vertical axis.
This is a reference signal at CO 100 ppm at the second and the like (third day from the start of energization). 8 to 12 are normalized by a reference signal on the third day of energization in CO 100 ppm.

Each point in the figure shows a measurement point associated with 5 weeks of energization. If 45 TGS 203 are used for 5 weeks, the resistance of the sensor is increased or lowered about twice as much. In FIG. 8, the increase in resistance is concentrated on a narrow straight line having a slope of 1 in a two-dimensional phase space of 6 seconds or the like and 0 seconds. This axis is called the drift axis. The reason why the drift axis is not clear at 30 ppm or 300 ppm of CO is due to the dispersion of the concentration dependency of TGS203. That is, the concentration dependency is not uniform, and CO30pp
Since the initial points in m and 300 ppm are not aligned at one point, the drift axis is unclear due to dispersion of the initial points. Also CO
A straight line connecting three points of 30 ppm, 100 ppm, and 300 ppm is called a concentration axis. The initial characteristics of the TGS 203 are on this concentration axis, and the concentration axis moves parallel to the direction of the drift axis with use.

A similar drift axis can be seen in FIG.
The distribution of the phase points around the drift axis widens, indicating that the correlation between the drift of the signal at 0 seconds and the drift of the signal at 12 seconds is weaker than in the case of 0-6 seconds.
In the characteristics at the 30th to 0th seconds in FIG. 10, the distribution around the drift axis is wider, and it is difficult to distinguish between 300 ppm CO and 100 ppm CO. In the characteristics at 60-0 seconds in FIG. 11, among the most drifted ones in CO 300 ppm, there is one that is almost overlapping the reference point at CO 100 ppm. In the characteristics from the 120th second to the 0th second in FIG. 12, since the signal at the 0th second and the signal at the 120th second are very similar, both the drift axis and the concentration axis are common, and all the phase points are one. Are gathered around the straight line.

From these facts, the initial signal in the high temperature range can be used for drift correction, for example, 4 to 20.
The signal at the second, preferably the signal at the 5th to 15th second. The other party to be combined is a signal in the latter half of the low temperature range, for example, a signal at the 90th to 150th seconds, preferably at the 120th to 150th seconds. In any of FIGS. 8 to 11, the concentration axis and the drift axis are oblique, and two axes corresponding to the CO concentration and the drift cannot be obtained in the orthogonal coordinate system. If a concentration axis orthogonal to the drift axis can be obtained, it means that there is an axis that is not affected by the drift, and the coordinates on that axis are determined only by the gas concentration. However, no such axis could be found.

In the embodiment, since the logarithm of the sensor resistance is used, the concentration axis and the drift axis are linear. However, if the sensor resistance itself is used, the concentration axis becomes a curve axis close to a parabola.

[0036]

[Negative hydrogen sensitivity] In each figure, the behavior of a mixed gas of 100 ppm of CO and 300 ppm of hydrogen and the behavior in 1000 ppm of hydrogen are also shown. As is clear from FIG. 8, the sensitivity to hydrogen is slightly negative. For example, in FIG.
When each point of 100 ppm + 300 ppm of hydrogen is translated in parallel along the drift axis and the intersection with the concentration axis is determined, the obtained concentration range is 80 ppm to 60 ppm of CO. On the other hand, CO100
The distribution of each point in ppm for 5 weeks is narrow, and the point is cross-translated along the drift axis to find the intersection with the concentration axis. The distribution range is about 80 to 120 ppm CO. The reason why the sensitivity to hydrogen is negative is that the signal at 6 seconds has a higher hydrogen sensitivity than the signal at 0 seconds. Therefore, in order to correct this, a phase space consisting of signals at the 0th and 69th seconds is used.

FIG. 13 shows the same 5-week energization data in this case. As is apparent from FIG. 13, when hydrogen is generated, the resistance value at the 69th second is significantly reduced, and is located at a place extremely far from the concentration axis. Therefore, a signal representing the hydrogen concentration is defined as the distance from the concentration axis in the downward direction in FIG.

Such a hydrogen detection signal is not accurate, and FIG. 13 does not use an oblique coordinate system. However, since it is used for correcting a small negative hydrogen sensitivity, even a hydrogen detection signal having no quantitative property can be used. Then, in the correction of the hydrogen sensitivity, the hydrogen sensitivity slightly negative in FIG. 8 is returned to 0, that is, a CO detector that is extremely selective only for CO is designed, or the CO characteristic is changed to the original characteristic of the TGS 203 as shown in FIG.
Correcting the relative sensitivity of hydrogen to 10: 1 can be considered. Which of these to choose depends on the CO
This is a matter of the design policy of the detection device.

[0039]

[Drift correction] FIG. 14 shows the principle of drift correction.
The solid line in the figure is the concentration axis, and the broken line is the drift axis. And 6
The reference signal at three points of 5 ppm, 200 ppm and 400 ppm is E
It is recorded in the EPROM 20. By measurement, LnR0
And a point (a, b) on the phase space in two dimensions of LnR6. The coordinates of each reference signal in this phase space are shown in FIG.
Determined as in 4. Then, the object is translated from the point (a, b) along the drift axis, and the coordinates of the intersection with the concentration axis are (e,
f). Once the coordinates (e, f) are obtained, the CO concentration can be obtained from the position on the concentration axis. The movement from the coordinates (a, b) to the coordinates (e, f) is a projection onto the density axis parallel to the drift axis.

The projection method may be arbitrarily determined. For example, data indicating the CO concentration may be written in the phase space of FIG. 14 to form a two-dimensional map, and the CO concentration may be obtained from the position on the map. The fact that the map is coarse and there is no data directly corresponding to each coordinate may be processed by interpolation between each point of the map. Alternatively, each reference point (l, m), (e, f), (q,
From r), three lines parallel to the drift axis are drawn, and a correction value for the CO concentration is written on each line, and the correction value is set to 1 on the concentration axis. Then, a correction value is determined so as to correct the increase in resistance due to the drift. Then, the density axis is moved in parallel so as to pass through the measurement point, the intersection point between the two correction lines on both sides is obtained, and the correction value at each correction line is obtained and interpolated. The logarithm of the sensor resistance at the 0th second is corrected by the correction value thus obtained, and is converted into a CO concentration. In these modified examples, the restriction of the projection can be reflected on the correction value of the correction line and the value of the map, and a fine operation on the projection is easy.

[0041]

[Temperature and Humidity Dependence] FIG. 16 shows another 10 TGS203s at 0 ° C. and a relative humidity of about 96% in 100 ppm CO.
And the ratio of the resistance value between 20 ° C. and 65% relative humidity.
The horizontal axis represents the timing of the temperature change. Most of the temperature-humidity dependence is corrected as a side effect of drift correction at the 0th and 6th seconds.

Referring to FIG. 17, the same ten TGS203s were measured at 50 ° C. and a relative humidity of about 40% in 100 ppm of CO.
And the ratio of the resistance value between 20 ° C. and 65% relative humidity.
Most of the temperature-humidity dependence is corrected as a side effect of drift correction at the 0th and 6th seconds.

[0043]

[Signal Processing] FIGS. 18 to 23 show the calculation of the CO concentration. FIG. 18 shows a main loop. First, three variables a, b, and c are defined from measurement data. Next, the CO concentration is obtained by a temperature correction subroutine (FIG. 19), a drift correction subroutine (FIG. 20), and a hydrogen correction subroutine (FIG. 21). Finally, the blood CO hemoglobin concentration COHb is determined from the CO concentration. The initial value of COHb is set to 0 at the time of reset. This conversion itself is already known, and k2, k3, and k4 are constants. Here, k4 is a value corresponding to about 30 ppm of CO, which is equal to or less than the lower limit of detection, and detection is not performed when the CO concentration is 30 ppm or less.

[0044]

[Temperature Correction Subroutine] In the temperature correction subroutine of FIG. A reference table of correction constants T1, T2, and T3 for a, b, and c from the ambient temperature is prepared in the program memory 34, and is read and added to a, b, and c.

[0045]

[Drift Correction Subroutine] FIG. 20 shows a drift correction subroutine. The inclination of the drift axis is 1, and (e-
a) and (fb) are equal. Therefore, f = e + (b−
a) is established. Therefore, one of the two unknowns e and f can be deleted. Next, it is checked whether np is equal to or larger than ab. If this condition is not satisfied, the measurement point is 200pp
When the drift axis is extended from m, it is below the drift axis, and the detected concentration is 200 ppm or less. Next, the point (e,
f) internally divides a line segment defined by two reference signals of 65 ppm and 200 ppm. From this, e and f are 65 ppm or 2 ppm.
The coordinates n, p, q, and r of the reference signal at 00 ppm are constrained by one relational expression, and the coordinates e can be solved using these.

There is no restriction on the projection of the obtained e, and the projection is similarly performed at a point extremely distant from the density axis and in the vicinity of the density axis. The projection is symmetrical above and below the density axis. On the other hand, as the drift from the concentration axis to the high resistance side becomes remarkable, it is preferable to limit the projection and correct only a part of the drift. In addition, when drifting from the concentration axis to the low resistance side, it is preferable to make correction more conservative than drift to the high resistance side. Further, the inclination of the drift axis at about 30 ppm of CO is slightly larger than the inclination of the drift axis at 100 ppm or more, and it is preferable to change the inclination of the drift axis for each concentration. Also, 30 ppm of CO is harmless and is not included in the detection target, and there is no need to perform drift correction on CO in such a low concentration range. Therefore, as shown in FIG. 22, it is preferable to make the correction asymmetric above and below the density axis, and to partially correct the drift as the distance from the density axis increases.

When a map is used or when a plurality of drift axes are provided for each CO concentration, the above processing can be performed by manipulating the data in the map or manipulating the inclination of the drift axis. However, in the embodiment, after e is obtained, the program memory 3
The above-described processing is performed using the two-dimensional lookup table stored in step S4.
The headings of this look-up table are (ea) and e, where (ea) is proportional to the distance from the concentration axis. The sign of (ea) is inverted above and below the density axis. The value of e indicates the CO concentration, and the processing of the low concentration area or the processing of the high concentration area is determined by the value of e. Therefore, if the value of e is updated from the look-up table in accordance with (ea) and e, the correction is conservative in an area that is asymmetrical above and below the density axis and large in distance from the density axis, and is corrected in a low density area. Can be modest. However, the processing corresponding to FIG. 22 may not be performed.

When the value e is finally found, 65 ppm and 2
The internal division ratio y of the line segment between 00 ppm is determined. When y is 0, the CO concentration is 200 ppm, and when y is 1, the CO concentration is 65 ppm. During this period, there is a change in the CO concentration about three times, and if this is solved as it is, a second-order or higher term is required in the series expansion of exp (y), so consider the midpoint between 65 ppm and 200 ppm. In the vicinity of 200 ppm, the series is developed based on the concentration of 200 ppm, and in the vicinity of 65 ppm, the series is developed based on the concentration of 65 ppm. In this way, exp (y) = 1 +
Even when approximating y, there is almost no approximation error. In this way, the CO concentration before the correction of the hydrogen concentration is determined.

If the phase point obtained is above the drift axis passing through 200 ppm of CO, the CO concentration is 200 pp
m is exceeded. Therefore, in this case, the EEPROM 20 is accessed to read the reference signal of 400 ppm of CO. Hereinafter, the CO concentration is determined in the same manner. The processing in this case is CO6
This is the same as the process when two reference signals of 5 ppm and 200 ppm are used.
A reference signal of 0 ppm may be used.

The temperature / humidity dependence has a gas concentration dependence as shown in FIG. 23, and the temperature / humidity dependence is different between the low concentration region and the high concentration region. However, at the stage of the temperature / humidity correction subroutine, C
O concentration is unknown. Therefore, after temporarily calculating the CO concentration, a two-dimensional look-up table stored in the program memory 34 from the ambient temperature T and the temporarily determined CO concentration is used.
Re-correct the density. This is a method of first-order approximation ignoring the temperature-humidity dependence of the CO concentration, and re-correcting the temperature-humidity dependence of the CO concentration using the obtained temporary CO concentration. In the lookup table, the amount of increase or decrease in the CO concentration is stored using the provisional CO concentration and the ambient temperature as headings.
Obtain the O concentration. The processing corresponding to FIG. 23 can be omitted.

[0051]

[Hydrogen correction subroutine] When the CO concentration is determined, hydrogen correction is performed. FIG. 21 shows the processing, and FIG. 15 shows the principle. In a two-dimensional phase space defined by the logarithm of the resistance value at 0 seconds and the logarithm of the resistance value at 69 seconds, the coordinates of the measurement point are (a,
c). This is 65ppm, 200ppm, 400
The intersection at the time of moving vertically upward in FIG. 15 to the concentration axis of ppm is defined as (a, g). Then, the difference between g and c is h, and h
The hydrogen concentration is determined by the In this case, it is determined whether or not it is necessary to use a signal of 400 ppm as a reference signal from whether or not the value of a exceeds n. If a is equal to or less than n, the EEPROM 20 is accessed to read the reference signal of 400 ppm. . Since the point (a, g) is on the line connecting the 200 ppm reference signal and the 400 ppm reference signal, one equation is generated for the coordinate g, and g
Can be requested. When g is obtained, h is obtained. For example, k1 × h is added to the CO concentration obtained in the main loop of FIG. 12, for example, using k1 as an appropriate positive constant. As a criterion for the addition, for example, the hydrogen concentration dependency of the CO detector is set to 0, or the relative sensitivity of CO to hydrogen is set to an appropriate value such as 10: 1. When a is larger than n, that is, when the coordinate point (a, c) obtained in FIG. 15 is on the right side of the reference signal of 200 ppm, the reference signals of 65 ppm and 200 ppm are used. Then, h is obtained in the same manner as described above, and the hydrogen concentration is corrected.

[0052]

[Information dimension of gas sensor] The number of independent signals included in the gas sensor will be referred to as information dimension. The information dimension is a concept clearly defined in the information science on the basis of entropy, but here, the number of independent signals is simply qualitatively determined and called the information dimension. As described above, the CO concentration and the degree of drift can be known from the combination of the signal at the 0th second and the signal at the 6th second, and the H2 concentration can be determined from the combination of the signal at the 0th second and the signal at the 69th second. I was able to know.

FIG. 24 shows another 40 TGS 203 at 0 ° C. relative humidity 96%, 20 ° C., 65%, 50 ° C. 40
% Dependence on temperature and humidity in CO30ppm, 100ppm, 30ppm
This is for 0 ppm. 20 ° C or 5 at 0 ° C
A phase point exists on a line different from 0 ° C., and a decrease in absolute humidity can be detected. As is well known, the TGS 203 mainly depends on the absolute humidity.

Using the temperature waveform of TGS203, C
Four kinds of signals of O, drift, hydrogen, and absolute humidity (FIG. 24) can be obtained. On the other hand, from the other correlation data of the temperature waveform of the TGS 203 (correlation when the signal whose sampling timing is changed is regarded as the other signal),
It can be seen that the temperature waveforms can be classified into four stages: an early high-temperature region, a late high-temperature region, and an early low-temperature region. The characteristics of the TGS 203 are affected by four factors: CO, hydrogen, absolute humidity, and drift. From these facts, it can be said that the temperature waveform of the TGS 203 includes four signals of CO, hydrogen, absolute humidity, and drift, and the information dimension of the temperature waveform of the TGS 203 is four-dimensional.

[0055]

[Selection of gas sensor]

[0056]

[Dispersion of Temperature Waveform] FIG. 25 shows the characteristics of seven defective products in the drift characteristics of FIG. Similarly, FIG. 26 shows the characteristics of eight samples left for two years or more in the drift characteristics of FIG. A comparison between FIG. 25 and FIG. 8 shows that a sample far from the drift axis appears in FIG. 25, and that the accuracy of drift correction is low for a defective product. Similarly, comparing FIG. 26 with FIG. 8, it can be seen that the sample left undisturbed for two years or more has a small drift and is concentrated on the drift axis and is extremely easy to handle.

FIGS. 27 to 34 show the temperature waveforms of each group in FIG. 8 on the third day from the start of energization. FIGS. 27 and 31 show the waveforms of defective products, and FIGS. 28 and 32 show the samples left for two years or more. 29 and 33 show 10 waveforms randomly extracted from 20 non-defective products, and FIGS. 30 and 34 show waveforms of 10 returned products. 27 to 30, in air (the line with the highest resistance), in CO 30 ppm, in 100 ppm, 300 pp
The line in m indicates the range of the maximum and minimum resistance values at 100 ppm of CO. In addition, a line having a valley of resistance near the 70th second is a line of 100 ppm of CO + 300 ppm of H2. Positive numbers are C
On the average of the resistance values of 100 ppm of O, negative numbers indicate gas concentration dependence. Similarly, FIG. 31 to FIG.
3 shows temperature waveforms of individual sensors in ppm.

When comparing FIGS. 27 and 28, the variance of the temperature waveform is large in the defective product of FIG. 27, the variance of the temperature waveform is small in the sample of FIG. 28, and the variance is medium in the samples of FIGS. . This characteristic is also evident in FIGS. 31 and 32. In the defective product shown in FIG. 31, the dispersion of the temperature waveform is remarkable, and in the sample left as shown in FIG. The samples of FIGS. 33 and 34 have a medium temperature waveform dispersion.

Summarizing the above, the success or failure of the drift correction, that is, whether or not the drift proceeds along a specific drift axis is determined by the degree of dispersion of the temperature waveform. When a sample having a small variance of the temperature waveform is used, the drift proceeds with a narrow width along the drift axis. This means that a sample having a waveform closer to the standard waveform, for example, the waveform shown in FIG. 28 is a better sensor, and FIGS.
This indicates that it is necessary to exclude a sample having a large deviation from the waveform average among the samples in FIG. As is clear from FIG. 27 to FIG. 30, there is no difference between the four groups in the average of the resistance values at 100 ppm CO, and the resistance value at 100 ppm CO is reduced except for the sample with the highest resistance among the defective products. All are in the range of 1 to 10 KΩ. FIG.
The reason why the sample No. 7 was determined to be defective was that the resistance value in 100 ppm of CO exceeded 10 KΩ on the day of the screening, and if the measurement was performed on a different day (in the example, the measurement was performed after being left for several months and then re-measured). 27) The characteristic shown in FIG. 27 is obtained. The inventor measured the temporal change of the temperature waveforms for the four groups used in FIG. 8, and the group of defective products has the largest variance from the average waveform, and the sample left for two years or longer has the most average waveform. Still showing a close, small distribution of variance.

The sensor numbers are shown in FIGS.
The large deviation from the average waveform in FIG.
30 and those close to the average waveform are 24, 25, 26, 29
It is. In FIG. 25, 27 and 30 are included in the drift in the CO of 100 ppm and the deviation from the axis is large.
Data along the axis include 24, 25, 26, and 29. Thus, the sample having a large deviation from the average waveform shows a different behavior from the other samples even in the drift.

[0061]

[Selection Method] FIG. 35 shows a method for selecting a gas sensor using the Fourier transform. The temperature of the gas sensor is changed in the gas to be detected. For example, in this case, the temperature is changed in a high temperature range of 60 seconds and a low temperature range of 90 seconds in 100 ppm of CO. It is sufficient to measure the degree of dispersion of the temperature waveform in one gas concentration, and it is sufficient to use the temperature waveform in one gas concentration for selection. The temperature change condition is preferably the same as the temperature change during actual use of the gas sensor, but is not limited to this. In the course of the temperature change, for example, gas sensor signals at 10 points, here 60 points, are sampled, and Ln is sampled.
Rst (Rs indicates a sensor signal and t indicates a sampling point) is obtained.

The Fourier transform is performed on the temperature waveform of the obtained sensor signal. A Laplace transform or the like may be used instead of the Fourier transform, and any transform may be used as long as the transform finds a generating function of the temperature waveform of the gas sensor signal. The DC component obtained by the Fourier transform is D
1, the sine components are R1 to R15, and the cosine components are D1 to
D15. The fundamental frequency of the Fourier transform is, for example, 6
It is 180 seconds, which is the least common multiple of 0 seconds and 90 seconds. Where R
The components 15 and I15 were obtained because a valley of the initial resistance value in the high temperature region occurs at about 3 seconds, corresponding to a cycle of about 12 seconds. And this is because a heater waveform of a square wave was used, and if a heater waveform such as a sine wave or a sawtooth wave is used,
For example, conversion up to R4 or I4 is sufficient, and sampling points, for example, 10 are sufficient.

Next, the value of the obtained Fourier transform is normalized, and here, it is normalized by the value of the DC component D1. For the normalization, a difference or the like may be used instead of the ratio.
15, I1 to I15 and the like may be used for normalization.

A temperature waveform of a sample that has been separately confirmed to exhibit preferable characteristics, for example, an average value of the temperature waveform in FIG. 32 in the embodiment, is obtained, and similarly Fourier-transformed and normalized to obtain a standard waveform. The normalized standard waveform is compared with the measured standard waveform to determine the similarity S. Note that the normalization is performed before obtaining the similarity, and this is for masking the magnitude of the resistance value. The similarity is R1 to R15, I1 to I15
Can be defined as a distance in a 30-dimensional space determined by the following equation. Since the DC component D1 is used for normalization, it is not considered in the similarity. Various distances known in the topology may be used for the distance. Here, distance weights Ai and Bi are simply prepared for each dimension of R1 to R15 and I1 to I15, and data Rs1 to Rs15 in a standard waveform are prepared. , Is1 to Is15 are added. The similarity S obtained in this way becomes smaller when the similarity of the waveforms is high, and sensors having S equal to or less than a predetermined value are selected as non-defective products.

FIG. 36 shows an embodiment using the value of the correlation function. The correlation function is used instead of the Fourier transform, and the sensor signal at, for example, 4 points, preferably 10 points or more is normalized because the correlation function is used. The comparison may be made later, and the same applies except that the number of sampling points decreases. For example, the logarithm of the sensor signal Rst at 10 points in 100 ppm of CO is obtained. The information dimension of the TGS 203 is four-dimensional, and at least four signals are required to compare the temperature waveforms. For example, if only two signals at the end of the high-temperature region and the end of the low-temperature region are sampled, this is too coarse as a signal representing the temperature waveform. In the three-point signal, one dimension is assigned to the DC component, and two dimensions are assigned to the other AC components, and a signal representative of the temperature waveform is not obtained. Therefore, the minimum number of sampling points is four, preferably ten or more.

The sampled LnRst is normalized by, for example, the signal LnRs0 at the 0th second, and the standard waveform LnS
t is also normalized by LnS0. Then, the value C of the correlation function is
It is obtained as the product sum of the weight factor At and the absolute value of the difference between the measured waveform and the standard waveform. There are various ways to determine At. For example, to reduce the contribution of a component having a large variance, a standard correlation function is obtained by using the product of the square root of the variance of LnRst and the square root of the variance of LnSt. Then, similarly to FIG. 35, those having a small correlation function value are selected as non-defective products.

In the embodiment, the TGS 203 was measured. However, the type of the gas sensor is arbitrary, and it is sufficient that the gas sensor is a metal oxide semiconductor gas sensor. In this embodiment, gas sensors having uniform drift characteristics can be selected, and drift can be accurately corrected.
If the temperature waveforms are uniform, the temperature and humidity are corrected (FIG. 24).
Excellent results are also obtained in the detection of hydrogen and hydrogen (FIG. 13).

[Brief description of the drawings]

FIG. 1 is a block diagram of a gas detection device according to an embodiment.

FIG. 2 is a diagram showing a configuration of a RAM in the gas detection device of the embodiment.

FIG. 3 is a diagram showing a configuration of an EEPROM in the gas detection device of the embodiment.

FIG. 4 is a characteristic diagram showing a waveform of a resistance value of the gas sensor used in the embodiment.

FIG. 5 is a characteristic diagram showing a resistance value waveform of a gas sensor used in an example in an initial high temperature range.

FIG. 6 is a flowchart showing a sampling algorithm in the gas detection device according to the embodiment.

FIG. 7 is a flowchart showing an adjustment algorithm in the gas detection device according to the embodiment.

FIG. 8 is a characteristic diagram showing a drift characteristic on a 0-6 second plane in the example.

FIG. 9 is a characteristic diagram showing a drift characteristic on a 0-12 second plane in the example.

FIG. 10 is a characteristic diagram showing a drift characteristic on a 0-30 second plane in the example.

FIG. 11 is a characteristic diagram showing a drift characteristic in a 0-60 second plane in the example.

FIG. 12 is a characteristic diagram showing a drift characteristic on a 0-120 second plane in the example.

FIG. 13 is a characteristic diagram showing a drift characteristic on a 0-69 second plane in the example.

FIG. 14 is a characteristic diagram showing a mechanism for calculating the CO concentration in the embodiment.

FIG. 15 is a characteristic diagram showing hydrogen correction in the embodiment.

FIG. 16 is a characteristic diagram showing the temperature / humidity dependency between 20 ° C.-65% RH and 0 ° C. of the gas sensor.

FIG. 17: 20 ° C.-65% RH and 50 ° C. of the gas sensor
Characteristic diagram showing temperature and humidity dependence between -40% RH

FIG. 18 is a flowchart illustrating a main program according to the embodiment.

FIG. 19 is a flowchart illustrating temperature and humidity correction in the embodiment.

FIG. 20 is a flowchart illustrating drift correction in the embodiment.

FIG. 21 is a flowchart illustrating correction to coexisting hydrogen in the gas detection device according to the embodiment.

FIG. 22 is a characteristic diagram showing details of drift correction in the embodiment.

FIG. 23 is a characteristic diagram showing details of temperature and humidity correction in the embodiment.

FIG. 24 is a characteristic diagram showing a temperature-humidity characteristic on a 9-60 second plane in the example.

FIG. 25 is a characteristic diagram showing drift in a 0-6 second plane in seven defective products.

FIG. 26 is a characteristic diagram showing a drift in a plane of 0-6 seconds in eight left samples.

FIG. 27: 30 ppm of CO in air with 7 defective products
Medium, 100ppm, 300ppm, and 100ppm of CO
Characteristic diagram showing temperature waveform in H2 300ppm

FIG. 28: CO30 in air with eight samples left untreated
ppm, 100 ppm, 300 ppm, and 100 ppm CO
Characteristic diagram showing temperature waveform in + H2 300ppm

FIG. 29: 10 good products, in air, 30 ppm of CO
Medium, 100ppm, 300ppm, and 100ppm of CO
Characteristic diagram showing temperature waveform in H2 300ppm

FIG. 30: 30 ppm of CO in air with 10 returned products
Medium, 100ppm, 300ppm, and 100ppm of CO
Characteristic diagram showing temperature waveform in H2 300ppm

FIG. 31 is a characteristic diagram showing individual temperature waveforms in 100 ppm of CO in seven defective products.

FIG. 32 is a characteristic diagram showing individual temperature waveforms in 100 ppm of CO in eight left samples.

FIG. 33 is a characteristic diagram showing individual temperature waveforms in 10 good products in 100 ppm of CO.

FIG. 34 is a characteristic diagram showing individual temperature waveforms in 100 ppm of CO in 10 returned products.

FIG. 35 is a flowchart showing selection of a gas sensor using Fourier transform.

FIG. 36 is a flowchart showing selection of a gas sensor using the value of a correlation function.

[Explanation of symbols]

 2 Metal oxide semiconductor 4 DC power supply 5 Resistance ladder 8,48 Microcomputer 10 Bus 12 AD converter 14 Resistance ladder control unit 16 Heater control unit 18 EEPROM control unit 20 EEPROM 22 I / O 23 Adjustment switch 24 Reset switch 26 Arithmetic logic operation unit 28 Sequence control unit 30 RAM 32 Alarm control unit 34 Program memory 36 Drive circuit 38 Buzzer 39, 40 LED 51 to 54 Card 42 Thermistor 44 Temperature / humidity correction unit S Metal oxide semiconductor gas sensor h1, h2 Heater T1, T2 Transistors R1 to Rn resistance

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Kazuko Takamatsu 1-3-5 Senba Nishi, Minoh City Inside Figaro Giken Co., Ltd.

Claims (5)

[Claims] 1. A method for selecting a gas sensor including a heater and a metal oxide semiconductor having a resistance value changed by a gas, comprising: measuring a temperature waveform of a gas sensor signal when the temperature of the gas sensor is changed in a gas to be detected. And selecting a gas sensor based on the similarity between the temperature waveform and the standard waveform. 2. The method according to claim 1, wherein at least four gas sensor signals in a temperature change process are used as the temperature waveform. 3. The method according to claim 1, wherein the similarity is obtained after normalizing both the temperature waveform and the standard waveform of the gas sensor signal. 4. A gas sensor according to claim 1, wherein a generating function between the temperature waveform and the standard waveform is obtained, and a similarity between the temperature waveform and the standard waveform of the gas sensor is obtained based on the similarity between the generating functions. Sorting method. 5. The gas sensor according to claim 1, wherein a value of a correlation function between the temperature waveform and the standard waveform is obtained, and a similarity between the temperature waveform and the standard waveform of the gas sensor is obtained from the value of the correlation function. Sorting method.