Gas discrimination method and gas discrimination system

Abstract

This gas identification method, which employs a sensor that outputs a signal corresponding to an adsorption concentration of a gas, includes: an acquisition step (step S12) of acquiring the signal output from the sensor which has been exposed to a sample gas during only a second period from within a measuring period comprising a first period, the second period, which follows the first period, and a third period, which follows the second period; an extraction step (step S13) of extracting at least one feature quantity corresponding to drift of the acquired signal; and an identification step (step S14) of employing a trained logical model for identifying sample gases, to identify the sample gas on the basis of the at least one extracted feature quantity, and outputting the identification result.

Description

[Technical Field]

[0001] The present disclosure relates to gas discrimination methods and systems. [Background technology]

[0002] For example, gas identification is performed based on a signal acquired from a sensor exposed to the gas. Patent Document 1 discloses a method for identifying an analyte using data of a pulsed signal that detects an analyte, in which the intensity, wavelength, intensity ratio, kurtosis, etc. of the pulsed signal are used as feature quantities. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2018 / 207524 Summary of the Invention [Problem to be solved by the invention]

[0004] When a sensor is used to identify a sample gas containing chemical substances such as volatile organic compounds, it is desired to reduce erroneous determinations.

[0005] Therefore, the present disclosure provides a gas identification method and the like that can improve identification accuracy. [Means for solving the problem]

[0006] A gas identification method according to one aspect of the present disclosure is a gas identification method using a sensor that outputs a signal corresponding to an adsorption concentration of a gas, the method including: a first step of acquiring a signal output from the sensor exposed to a sample gas only during a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; a second step of extracting one or more feature quantities corresponding to drift in the acquired signal; and a third step of identifying the sample gas based on the one or more extracted feature quantities using a trained logical model for identifying the sample gas, and outputting an identification result.

[0007] A gas identification system according to one aspect of the present disclosure includes a sensor that outputs a signal corresponding to an adsorption concentration of a gas; an exposure unit that exposes the sensor to a sample gas only during a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; an acquisition circuit that acquires the signal output from the sensor during the measurement period; an extraction circuit that extracts one or more feature quantities corresponding to drift in the acquired signal; a memory that stores a trained logical model that identifies the sample gas; and an identification circuit that uses the trained logical model to identify the sample gas based on the one or more extracted feature quantities and outputs an identification result. [Effects of the Invention]

[0008] According to a gas identification method and the like according to an aspect of the present disclosure, identification accuracy can be improved. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a block diagram showing a schematic configuration of a gas identification system according to an embodiment. [Figure 2] FIG. 2 is a schematic diagram showing an example of the configuration of the exposed portion according to the embodiment. [Figure 3] FIG. 3 is a block diagram showing a schematic configuration of a gas identification system according to a modified example of the embodiment. [Figure 4]FIG. 4 is a flowchart illustrating the operation of the gas identification system according to the embodiment. [Figure 5] FIG. 5 is a diagram showing an example of a control signal to the exposure unit and a signal output from the sensor according to the embodiment. [Figure 6] FIG. 6 is a diagram showing an example of a signal output from the sensor according to the embodiment during a plurality of consecutive measurement periods. [Figure 7] FIG. 7 is a diagram for explaining the value of a signal acquired by an extraction circuit according to an embodiment. [Figure 8] FIG. 8 is another diagram for explaining the value of the signal acquired by the extraction circuit according to the embodiment. [Figure 9] FIG. 9 is a diagram for explaining signal values ​​acquired by the extraction circuit according to the embodiment during a plurality of consecutive measurement periods. [Figure 10] FIG. 10 is another diagram for explaining the signal values ​​acquired by the extraction circuit according to the embodiment during a plurality of consecutive measurement periods. [Figure 11] FIG. 11 is a diagram for explaining the rate and amount of change in the value of the signal output from the sensor according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] (How one aspect of the present disclosure was achieved) When a sensor that outputs a signal corresponding to the adsorption concentration of a gas is used for gas identification, the chemical substance contained in the sample gas is identified as an identification target substance based on a feature quantity using the signal output from the sensor when the sample gas containing chemical substances such as volatile organic compounds is exposed. For example, the adsorption concentration to the sensor varies depending on the type of chemical substance, and therefore the signal output from the sensor changes. Therefore, for example, the amount and rate of change in the signal output from the sensor during the period in which the sensor is exposed to the sample gas are used as feature quantities to identify the chemical substances contained in the sample gas. However, depending on the type of chemical substance, the signal changes due to the adsorption of the chemical substance to the sensor during that period may be similar between multiple chemical substances, which may result in erroneous identification. Therefore, gas identification methods are required to improve the identification accuracy.

[0011] On the other hand, in gas identification, a period in which the sensor is not exposed to the sample gas is often provided before and after the period in which the sensor is exposed to the sample gas. The inventors have found that, after the period in which the sensor is exposed to the sample gas, the signal value that has changed due to the exposure to the sample gas is unlikely to return to its original value, and drift may occur in the signal output from the sensor. The inventors have also found that such drift may occur even when the signal changes associated with the adsorption of chemical substances to the sensor are similar. Based on this finding, the present disclosure provides a gas identification method and the like that can improve identification accuracy.

[0012] (Summary of the Disclosure) An outline of one aspect of the present disclosure is as follows.

[0013] A gas identification method according to one aspect of the present disclosure is a gas identification method using a sensor that outputs a signal corresponding to an adsorption concentration of a gas, the method including: a first step of acquiring a signal output from the sensor exposed to a sample gas only during a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; a second step of extracting one or more feature quantities corresponding to drift in the acquired signal; and a third step of identifying the sample gas based on the one or more extracted feature quantities using a trained logical model for identifying the sample gas, and outputting an identification result.

[0014] As a result, in the second step, one or more feature quantities corresponding to the signal drift, which differ from the feature quantities corresponding to the signal output during the second period in which the sensor is exposed to the sample gas, i.e., the output dependent on the concentration of gas adsorbed to the sensor, are extracted. Therefore, in the third step, identification based on the signal drift is performed, and sample gases can be identified with high identification accuracy even when identifying sample gases with similar outputs from the sensor dependent on the concentration of gas adsorbed.

[0015] Also, for example, the method may further include a fourth step of exposing the sensor to the sample gas only during the second period of the measurement period, and in the first step, the signal output from the sensor exposed in the fourth step may be acquired.

[0016] This allows one or more feature quantities to be extracted using the signal output from the sensor exposed in the fourth step.

[0017] Furthermore, for example, in the fourth step, the sensor may be exposed to a reference gas during the first period and the third period.

[0018] As a result, the sensor is exposed to the reference gas that serves as a measurement standard during the first and third periods, so that a reference signal can be obtained from the sensor even if the surrounding environment changes.

[0019] Furthermore, for example, in the first step, the signal output from the sensor may be acquired via a network.

[0020] This allows the output from the sensor to be obtained as measured at another location.

[0021] Also, for example, in the second step, a first value, which is the value of the signal when the value of the signal that fluctuated due to the sensor being exposed to the sample gas during the second period is returning to a reference value during the third period, may be acquired, and at least one of the one or more feature quantities may be extracted using the acquired first value.

[0022] This makes it possible to use the first value, which is the value of the signal when it is returning to a reference value directly related to the drift, for extracting the feature amount.

[0023] Also, for example, in the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods, and in the second step, the first value is acquired during each of two or more of the plurality of measurement periods, and the difference between the acquired first values ​​may be extracted as at least one of the one or more feature amounts.

[0024] As a result, acquiring signals over multiple consecutive measurement periods tends to increase the drift of the acquired signals, which tends to increase the difference between the first values ​​over the multiple measurement periods, and therefore tends to increase the difference in the extracted feature amounts depending on the type of sample gas, further improving the identification accuracy.

[0025] Also, for example, in the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods, and in the second step, the first value is acquired during each of two or more of the plurality of measurement periods, an approximation formula is derived using the acquired first value, and a coefficient of the derived approximation formula may be extracted as at least one of the one or more feature quantities.

[0026] Therefore, acquiring signals over multiple consecutive measurement periods tends to increase drift in the acquired signals. Also, by extracting the coefficients of an approximation formula using the first values ​​over each of the multiple measurement periods as feature quantities, feature quantities in which the variation in the first values ​​is smoothed can be extracted.

[0027] Also, for example, in the second step, a second value, which is the value of the last signal in the first period, may be acquired, and the difference between the acquired first value and the acquired second value may be extracted as at least one of the one or more feature amounts.

[0028] This makes it possible to extract a feature amount corresponding to the drift using the second value as a reference value.

[0029] Also, for example, in the second step, the value of the signal at the end of the third period may be acquired as the first value.

[0030] This allows the final signal value of the third period, at which the value of the signal that is returning to the reference value during the third period is likely to be stable, to be used as the first value.

[0031] Also, for example, in the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods, and in the second step, a third value, which is the value of the signal when it fluctuates due to the sensor being exposed to the sample gas during the second period, is acquired during each of two or more of the plurality of measurement periods, and the difference between the acquired third values ​​may be extracted as at least one of the one or more feature quantities.

[0032] As a result, acquiring signals over multiple consecutive measurement periods tends to increase the drift of the acquired signals, which tends to increase the difference between the third values ​​over the multiple measurement periods, which tends to increase the difference in the extracted feature amounts depending on the type of sample gas, thereby further improving the identification accuracy.

[0033] Also, for example, in the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods, and in the second step, a third value, which is the value of the signal when it fluctuates due to the sensor being exposed to the sample gas during the second period, is acquired during each of two or more of the measurement periods, an approximation equation is derived using the acquired third value, and a coefficient of the derived approximation equation may be extracted as at least one of the one or more feature quantities.

[0034] Therefore, acquiring signals over multiple consecutive measurement periods tends to increase the drift of the acquired signals.In addition, by extracting the coefficients of the approximation formula using the third value over each of the multiple measurement periods as feature quantities, it is possible to extract feature quantities in which the variation in the third value is smoothed out.

[0035] Also, for example, in the first step, the signal output from the sensor may be acquired during a plurality of consecutive measurement periods, and in the second step, at least one of the one or more feature quantities may be extracted based on the signal acquired during a second or subsequent measurement period among the plurality of measurement periods.

[0036] As a result, feature values ​​are extracted using signals output from the second and subsequent measurement periods, which tend to have larger drifts, among signals acquired over multiple consecutive measurement periods. This means that the extracted feature values ​​tend to vary more depending on the type of sample gas, further improving identification accuracy.

[0037] a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; an exposure unit that exposes the sensor to a sample gas only during the second period; an acquisition circuit that acquires the signal output from the sensor during the measurement period; an extraction circuit that extracts one or more feature quantities corresponding to drift in the acquired signal; a memory that stores a trained logical model that identifies the sample gas; and an identification circuit that uses the trained logical model to identify the sample gas based on the one or more extracted feature quantities and outputs an identification result.

[0038] As a result, the extraction circuit extracts one or more feature quantities corresponding to the signal drift, which differ from the feature quantities corresponding to the signal output during the second period in which the sensor is exposed to the sample gas, i.e., the output dependent on the concentration of gas adsorbed to the sensor. Therefore, the identification circuit performs identification based on the signal drift, and can identify the sample gas with high identification accuracy even when identifying sample gases with similar output from the sensor dependent on the concentration of gas adsorbed.

[0039] Hereinafter, the embodiments will be described in detail with reference to the drawings as appropriate. Note that the embodiments described below are all comprehensive or specific examples. The numerical values, shapes, materials, components, component placement and connection forms, steps, and step order shown in the following embodiments are merely examples and are not intended to limit the present disclosure. Furthermore, among the components in the following embodiments, components not recited in independent claims will be described as optional components.

[0040] Furthermore, in this specification, terms indicating the relationship between elements, such as parallelism, terms indicating the shape of elements, and numerical ranges are not expressions that only express a strict meaning, but are expressions that also include a substantially equivalent range, for example, a difference of about a few percent.

[0041] In addition, the drawings are not necessarily strict illustrations, and the same reference numerals are used to designate substantially the same components in the drawings, and redundant explanations are omitted or simplified.

[0042] Furthermore, in this specification, unless otherwise specified, ordinal numbers such as "first" and "second" do not refer to the number or order of steps, components, etc., but are used for the purpose of avoiding confusion and distinguishing between steps and similar components, etc.

[0043] (Embodiment) [composition] First, the configuration of a gas identifying system according to an embodiment will be described.

[0044] FIG. 1 is a block diagram showing a schematic configuration of a gas identifying system 100 in accordance with the present embodiment.

[0045] As shown in FIG. 1 , gas identification system 100 according to this embodiment includes sensor 10, exposure unit 20, control circuit 31, acquisition circuit 32, extraction circuit 33, identification circuit 34, and memory 40. Gas identification system 100 identifies a sample gas based on the output of sensor 10 exposed to the sample gas. The sample gas may contain, for example, a chemical substance to be identified. Examples of sample gas include gas collected from food, exhaled breath collected from a human body, air surrounding a human body, or air collected from a room in a building.

[0046] Gas identification system 100 identifies, for example, chemical substances contained in a sample gas. Specifically, gas identification system 100 identifies which of a plurality of target substances is contained in the sample gas as a chemical substance. Gas identification system 100 may also identify whether or not the target substance is contained in the sample gas.

[0047] The substance to be identified is, for example, a volatile organic compound, but may also be an inorganic gas such as ammonia or carbon monoxide. Gas identification system 100 is used, for example, to identify odors. In this case, the volatile organic compound is, for example, a molecule that becomes an odor component.

[0048] The sensor 10 is a sensor that outputs a signal corresponding to the concentration of adsorbed gas. The sensor 10 may be, for example, an electrochemical type, a semiconductor type, a field effect transistor type, a surface acoustic wave type, a quartz oscillator type, or a resistance change type sensor.

[0049] The sensor 10 includes, for example, a sensing unit and a pair of electrodes electrically connected to the sensing unit. The electrical resistance of the sensing unit changes depending on, for example, the concentration of adsorbed gas. A signal corresponding to the electrical resistance of the sensing unit of the sensor 10 is acquired by the acquisition circuit 32 as, for example, a voltage signal or a current signal via the pair of electrodes.

[0050] The sensing unit is composed of, for example, a resin material that is an adsorbent for adsorbing gas and conductive particles dispersed in the resin material. Examples of the resin material include polyalkylene glycol resin, polyester resin, and silicone resin. The resin material is, for example, a material commercially available as a stationary phase for gas chromatography columns, with side chains containing various substituents such as phenyl and methyl groups. From the viewpoint of durability and gas adsorption, the resin material may be, for example, a silicone resin commercially available as a stationary phase for columns, with side chains containing various substituents such as phenyl and methyl groups. Furthermore, the sensing unit is not limited to a composition of a resin material and conductive particles, and may be any material whose electrical resistance value changes upon gas adsorption. The sensing unit may be composed of an inorganic material such as a metal oxide, for example, porous ceramics.

[0051] Gas identification system 100 includes, for example, a plurality of sensors 10. The sensing portions of at least two of the plurality of sensors 10 (specifically, the resin materials constituting the sensing portions) are made of, for example, different materials. Alternatively, the sensing portions of all of the plurality of sensors 10 may be made of different materials. Different materials exhibit different adsorption behaviors toward the same chemical substance. Therefore, the plurality of sensors 10 output different signals in response to the same chemical substance. This allows different features to be extracted from the outputs of the plurality of sensors 10, thereby improving the identification accuracy of gas identification system 100.

[0052] The exposure unit 20 is an exposure mechanism that exposes the sensor 10 to a gas under the control of the control circuit 31. Specifically, the exposure unit 20 exposes the sensor 10 to the sample gas only during the second period of a measurement period consisting of a first period, a second period following the first period, and a third period following the second period. The exposure unit 20 may also expose the sensor 10 to a reference gas during the first and third periods. The reference gas is a gas that serves as a reference for measurement and is, for example, a gas that does not contain the target substance. The reference gas is, for example, a gas that is less likely to be adsorbed by the sensing unit of the sensor 10 than the target substance. Specific examples of reference gases include inert gases such as air and nitrogen, and gas obtained by removing chemical substances from the sample gas using a filter or the like. By outputting a signal from the sensor 10 exposed to the reference gas during the first and third periods, a reference signal corresponding to the ambient environment for each measurement can be obtained even when the ambient environment changes, and using such a signal can improve the identification accuracy described below.

[0053] Here, a specific configuration of the exposure section 20 will be described. Fig. 2 is a schematic diagram showing an example of the configuration of the exposure section 20 according to the present embodiment. As shown in Fig. 2, the exposure section 20 has, for example, a storage section 21, a three-way electromagnetic valve 22, an intake pump 23, and a plurality of pipes 25a, 25b, 25c, 25d, and 25e.

[0054] One end of pipe 25a is provided with inlet 26a for introducing sample gas. Inlet 26a is provided, for example, in a space filled with sample gas. One end of pipe 25b is provided with inlet 26b for introducing reference gas. Inlet 26b is provided, for example, in a space filled with reference gas. One end of pipe 25e is provided with exhaust port 26e for discharging the introduced sample gas and reference gas.

[0055] The accommodation unit 21 is a box-shaped container that accommodates the sensors 10. For example, a plurality of sensors 10 are arranged in an array inside the accommodation unit 21. One end of each of pipes 25c and 25d is connected to the accommodation unit 21. When the intake pump 23, which will be described later, is operated, gas flows from one end of pipe 25c to one end of pipe 25d. The plurality of sensors 10 are arranged in the flow path through which the gas flows.

[0056] The sample gas introduced from intake port 26a is introduced into storage unit 21 via pipe 25a, three-way solenoid valve 22, and pipe 25c. The reference gas introduced from intake port 26b is introduced into storage unit 21 via pipe 25b, three-way solenoid valve 22, and pipe 25c. The sample gas and reference gas introduced into storage unit 21 are exhausted from exhaust port 26e via pipe 25d, intake pump 23, and pipe 25e.

[0057] The three-way solenoid valve 22 is a solenoid valve for switching the gas introduced into the storage section 21. The three-way solenoid valve 22 has an input port P1 connected to the other end of the pipe 25a, an input port P2 connected to the other end of the pipe 25b, and an output port P3 connected to the other end of the pipe 25c. The opening and closing of each port of the three-way solenoid valve 22 is controlled by the control circuit 31. The three-way solenoid valve 22 is switched, under the control of the control circuit 31, between a first state in which the input port P1 and the output port P3 are electrically connected, and a second state in which the input port P2 and the output port P3 are electrically connected. In the first state, the input port P1 and the output port P3 are open, and the input port P2 is closed. In the second state, the input port P2 and the output port P3 are open, and the input port P1 is closed.

[0058] The intake pump 23 is a pump for introducing the sample gas and the reference gas into the storage section 21 and discharging the introduced sample gas and reference gas from the exhaust port 26e. The operation of the intake pump 23 is controlled by the control circuit 31. The intake port of the intake pump 23 is connected to the other end of the pipe 25d. The exhaust port of the intake pump 23 is connected to the other end of the pipe 25e.

[0059] With this configuration, when the intake pump 23 is operating and the three-way solenoid valve 22 is in the first state, the sample gas is introduced into the storage unit 21. As a result, the exposure unit 20 exposes the multiple sensors 10 to the sample gas. Also, when the intake pump 23 is operating and the three-way solenoid valve 22 is in the second state, the reference gas is introduced into the storage unit 21. As a result, the exposure unit 20 exposes the multiple sensors 10 to the reference gas.

[0060] The configuration of the exposure unit 20 is not limited to the configuration shown in FIG. 2 , and is not particularly limited as long as it can expose the sensor 10 to the sample gas. The exposure unit 20 may be configured, for example, so that the sample gas and the reference gas are introduced into the storage unit 21 through separate pipes without passing through the three-way solenoid valve 22. The exposure unit 20 may also be configured without the intake pump 23, so that a carrier gas is constantly flowing into the storage unit 21 and the sample gas is mixed into the carrier gas. Alternatively, the reference gas may not be introduced, and the storage unit 21 may be evacuated by the intake pump 23 after the sensor 10 is exposed to the sample gas. The exposure unit 20 may further include various removal filters for removing moisture or particulates from the sample gas and the reference gas, electromagnetic adjustment valves for adjusting the flow rate of each pipe, check valves for preventing backflow in each pipe, etc.

[0061] 1 again, as described above, the control circuit 31 controls the operation of the exposure unit 20, specifically, the operation of the three-way solenoid valve 22 and the suction pump 23. The control circuit 31 may also output information indicating the timing of the operation of the exposure unit 20 to the acquisition circuit 32.

[0062] The acquisition circuit 32 acquires a signal output from the sensor 10 during the measurement period. The acquisition circuit 32 acquires, for example, a voltage signal or a current signal as a signal output corresponding to the electrical resistance value of the sensing unit of the sensor 10.

[0063] The extraction circuit 33 extracts one or more feature amounts corresponding to the drift of the signal acquired by the acquisition circuit 32. The extraction circuit 33 may extract feature amounts other than the feature amount corresponding to the drift from the signal acquired by the acquisition circuit 32. When there are multiple sensors 10, the extraction circuit 33 extracts one or more feature amounts from the signals output by each of the multiple sensors 10.

[0064] The identification circuit 34 uses the trained logical model to identify the sample gas based on one or more feature quantities extracted by the extraction circuit 33. The identification circuit 34, for example, identifies which of a plurality of identification target substances is contained in the sample gas. The identification circuit 34 may also identify whether or not the sample gas contains the identification target substance. The identification circuit 34 receives one or more feature quantities as input and outputs an identification result. The identification circuit 34 outputs, for example, information for displaying the identification result on a display (not shown) or the like provided in the gas identification system. The identification circuit 34 may output information indicating the identification result to the memory 40 and store the information in the memory 40. The identification circuit 34 may also output information indicating the identification result to an external device.

[0065] The control circuit 31, the acquisition circuit 32, the extraction circuit 33, and the identification circuit 34 are realized by a microcomputer or a processor that has a built-in program for performing the above-mentioned processing. The control circuit 31, the acquisition circuit 32, the extraction circuit 33, and the identification circuit 34 may each be realized by a dedicated logic circuit that performs the above-mentioned processing.

[0066] The memory 40 is a storage device that stores a trained logic model used in the classification circuit 34. The memory 40 is realized by, for example, a semiconductor memory.

[0067] The trained logical model is a logical model that identifies a sample gas. Specifically, the trained logical model is, for example, a logical model that identifies which of a plurality of identification target substances is contained in the sample gas. The trained logical model, for example, receives as input one or more feature quantities extracted by the extraction circuit 33, and outputs which of a plurality of identification target substances is contained in the sample gas. The trained logical model may also output whether or not the sample gas contains the identification target substance.

[0068] The trained logical model is constructed by performing machine learning using, for example, a known substance to be identified and one or more feature quantities extracted by the extraction circuit 33 using the known substance to be identified as training data. The method used to construct the logical model in machine learning is not particularly limited. For example, a neural network is used to construct the logical model in machine learning. That is, the trained logical model includes, for example, a neural network. For constructing the logical model in machine learning, a random forest, a support vector machine, a self-organizing map, or the like may also be used.

[0069] Gas identification system 100 is implemented, for example, as a single gas identification device including the above components, but may also be implemented using multiple devices. When gas identification system 100 is implemented using multiple devices, the components of gas identification system 100 may be distributed among the devices in any manner. An example of a gas identification system implemented using multiple devices will now be described with reference to FIG. 3. FIG. 3 is a block diagram showing the schematic configuration of gas identification system 100a according to a variation of the embodiment.

[0070] As shown in FIG. 3, gas identifying system 100 a includes detection device 200 and identification device 300 .

[0071] Detection device 200 includes sensor 10, exposure unit 20, control circuit 31, detection unit 50, and communication unit 51. Sensor 10, exposure unit 20, and control circuit 31 have the same configuration as, for example, gas identifying system 100 described above.

[0072] The detection unit 50 acquires a signal output from the sensor 10 during the measurement period. For example, a voltage signal or a current signal is acquired as a signal corresponding to the electrical resistance value of the sensing unit of the sensor 10. The detection unit 50 also acquires information indicating the timing to control the exposure unit 20 from the control circuit 31. The detection unit 50 transmits the acquired signal and information to the identification device 300 using the communication unit 51. The detection unit 50 is realized by a microcomputer or processor that has a program that performs the above processing built in. The detection unit 50 may also be realized by a dedicated logic circuit that performs the above processing.

[0073] The communication unit 51 is a communication module (communication circuit) that enables the detection device 200 to communicate with the identification device 300 via a wide area communication network 90 such as the Internet, which is an example of a network. The communication unit 51 may perform wired communication or wireless communication. There are no particular limitations on the communication standard used for communication by the communication unit 51.

[0074] Identification device 300 includes acquisition circuit 32a, extraction circuit 33, identification circuit 34, memory 40, and communication unit 60. Extraction circuit 33, identification circuit 34, and memory 40 have the same configuration as, for example, gas identification system 100 described above.

[0075] The acquisition circuit 32a acquires the signal output from the sensor 10 during the measurement period, which is acquired by the detection unit 50, via the wide area communication network 90. ​​The acquisition circuit 32a communicates with the detection device 200 via the wide area communication network 90 using the communication unit 60.

[0076] The communication unit 60 is a communication module (communication circuit) that enables the identification device 300 to communicate with the detection device 200 via the wide area communication network 90. ​​The communication unit 60 may perform wired communication or wireless communication. There are no particular limitations on the communication standard used for communication by the communication unit 60.

[0077] [Operation] Next, the operation of the gas identification system according to this embodiment will be described. The following mainly describes the operation of gas identification system 100, but unless otherwise specified, gas identification system 100a also operates in a similar manner.

[0078] FIG. 4 is a flowchart illustrating the operation of gas identification system 100 according to this embodiment. In other words, FIG. 4 is a flowchart of a gas identification method performed by gas identification system 100. The gas identification method according to this embodiment includes an exposure step, an acquisition step, an extraction step, and an identification step. In this specification, the exposure step is an example of a fourth step, the acquisition step is an example of a first step, the extraction step is an example of a second step, and the identification step is an example of a third step.

[0079] (1) Exposure step and acquisition step As shown in FIG. 4, first, in the exposure step, the exposure unit 20 exposes the sensor 10 to the sample gas only during the second period of the measurement period (step S11). The exposure unit 20 also exposes the sensor 10 to the reference gas during the first and third periods. For example, the control circuit 31 operates the intake pump 23 and controls the opening and closing of each port of the three-way solenoid valve 22, thereby exposing the sensor 10 to the reference gas during the first and third periods and exposing the sensor 10 to the sample gas during the second period. In step S11, for example, the sensor 10 is exposed to the sample gas only during the second period of the measurement periods in a plurality of consecutive measurement periods.

[0080] Then, in the acquisition step, the acquisition circuit 32 acquires a signal output from the sensor 10 exposed in step S11 (step S12). That is, the acquisition circuit 32 acquires a signal output from the sensor 10 exposed to the sample gas only during the second period of the measurement period.

[0081] In gas identification system 100a, detection unit 50 acquires the signal output from sensor 10 exposed in step S11. Acquisition circuit 32a acquires the signal output from sensor 10 exposed in step S11 from detection unit 50 via wide area communication network 90. ​​This allows acquisition circuit 32a to acquire the signal output from sensor 10 even when sensor 10 is located far away from identification device 300.

[0082] FIG. 5 is a diagram showing an example of a control signal to the exposure unit 20 and a signal output from the sensor 10. FIG. 5 shows the intensities of the control signal to the three-way solenoid valve 22 and the signal output from the sensor 10 during a measurement period Tm consisting of a first period T1, a second period T2, and a third period T3. FIG. 5(a) is a graph showing an example of the change over time in the control signal output from the control circuit 31. In FIG. 5(a), when the control signal is at a high level, the three-way solenoid valve 22 is controlled to be in the first state, and when the control signal is at a low level, the three-way solenoid valve 22 is controlled to be in the second state. FIG. 5(b) is a graph showing an example of the change over time in the intensity (e.g., voltage) of the signal output from the sensor 10.

[0083] In step S11, for example, as shown in FIG. 5A, during the first period T1 and the third period T3, the three-way solenoid valve 22 is in the second state, and the exposure unit 20 exposes the sensor 10 to the reference gas. During the second period T2, the three-way solenoid valve 22 is in the first state, and the exposure unit 20 exposes the sensor 10 to the sample gas. As a result, the signal acquired in step S12 changes, for example, as shown in FIG. 5B. First, during the first period T1 in which the sensor 10 is exposed to the reference gas, the signal value hardly changes. Next, during the second period T2 in which the sensor 10 is exposed to the sample gas, the sensing unit of the sensor 10 adsorbs the sample gas (mainly chemical substances contained in the sample gas), causing the signal value to fluctuate (e.g., increase). Then, during the third period T3 in which the sensor 10 is again exposed to the reference gas, the sample gas (mainly chemical substances contained in the sample gas) is released from the sensing unit of the sensor 10, causing the signal value that fluctuated during the second period to return to the reference value. The reference value is, for example, the value before the signal value begins to fluctuate due to exposure of the sensor 10 to the sample gas.

[0084] The lengths of the first, second, and third periods are not particularly limited and are set, for example, according to the type of sensor 10 and the type of material to be identified. The length of the first period T1 is, for example, 1 second or more and 10 seconds or less. The length of the second period T2 is, for example, 5 seconds or more and 30 seconds or less. The length of the third period is, for example, 10 seconds or more and 100 seconds or less.

[0085] In step S11, for example, the exposure unit 20 exposes the sensor 10 to the sample gas only during a second period of each of a plurality of consecutive measurement periods Tm. Then, in step S12, the acquisition circuit 32 acquires signals output from the sensor 10 during a plurality of consecutive measurement periods Tm.

[0086] FIG. 6 is a diagram illustrating an example of signals output from the sensor 10 during multiple consecutive measurement periods. As shown in FIG. 6, the acquisition circuit 32 acquires signals output from the sensor 10 during, for example, seven consecutive measurement periods Tm-1 to Tm-7. During each of the measurement periods Tm-1 to Tm-7, the operation described in FIG. 5 is performed, and the same operation is repeated. Thus, during the second and subsequent measurement periods Tm-2 to Tm-7, the next measurement period Tm begins before the signal value that fluctuated during the second period T2 has completely returned to the reference value, resulting in a large drift, which will be described in detail later. The number of consecutive measurement periods Tm is not particularly limited and can be set, for example, depending on the type of sensor 10 and the type of material to be identified.

[0087] (2) Extraction step 4 again, next, in the extraction step, the extraction circuit 33 extracts one or more feature amounts corresponding to the drift of the signal acquired in step S12 (step S13). The extraction circuit 33 extracts one or more feature amounts using the value of the acquired signal.

[0088] As shown in FIG. 6, the signal output from sensor 10 exhibits drift, with the baseline deviating from the reference value. This signal drift is thought to be caused, for example, by the difficulty of gas escaping from the sensing section of sensor 10, and its magnitude varies depending on the combination of the sample gas (specifically, the chemical substances contained in the sample gas) and the material of the sensing section of sensor 10. For example, if the sample gas is difficult to escape from the sensing section of sensor 10, the output value from sensor 10 does not return to the reference value and the drift becomes large. On the other hand, if the sample gas is easy to escape from the sensing section of sensor 10, the drift becomes small or does not occur. Extraction circuit 33 extracts one or more feature quantities corresponding to the signal drift for use in identifying the sample gas.

[0089] For example, the extraction circuit 33 acquires at least one of a first value, a second value, and a third value as a signal value from the acquired signal, and extracts one or more feature amounts using the acquired value.

[0090] First, the first value, the second value, and the third value acquired by the extraction circuit 33 will be described.

[0091] 7 and 8 are diagrams for explaining the values ​​of the signals acquired by the extraction circuit 33. Temporal changes in the intensity (e.g., voltage) of the signals output from the sensor 10 and acquired by the acquisition circuit 32 are shown in FIGS.

[0092] The first value is the signal value when the signal value that fluctuated due to exposure of the sensor 10 to the sample gas during the second period T2 attempts to return to the reference value during the third period T3. The first value is, for example, the signal value V1 at a predetermined time during the third period T3, as shown in Fig. 7. Alternatively, the first value may be the average value of the signal values ​​during a predetermined interval S1 during the third period T3, as shown in Fig. 8. The length of the interval S1 is, for example, 0.1 seconds or more and 5 seconds or less.

[0093] The first value may be the final signal value in the third period T3. In this case, the first value may be the signal value V1 at the final point in the third period T3, or may be the average value of the signal values ​​in the section S1 that includes the final point in the third period T3. This allows the extraction circuit 33 to use the signal value that is stable during the third period T3 as the first value. Furthermore, the first value may be the signal value V1 at a point in time after the start of the third period T3, or may be the average value of the signal values ​​in the section S1 that starts after the predetermined time has elapsed from the start of the third period T3. The predetermined time is, for example, a time that is at least half the length of the third period T3.

[0094] In this way, the first value is the signal value during the third period T3 when the sensor 10 is not exposed to the sample gas, after the second period T2 when the sensor 10 is exposed to the sample gas, and is therefore less affected by exposure to the sample gas and is suitable as a value indicating baseline fluctuations in the signal.

[0095] The second value is the final signal value in the first period T1. For example, the second value is the signal value V2 at the end of the first period T1, as shown in Fig. 7. Alternatively, the second value may be the average value of the signal values ​​in a section S2 that includes the end of the first period T1, as shown in Fig. 8. The length of section S2 is, for example, 0.1 seconds or more and 5 seconds or less.

[0096] In this way, the second value is the signal value during the first period T1 when the sensor 10 is not exposed to the sample gas, before the second period T2 when the sensor 10 is exposed to the sample gas, and is therefore suitable as a value indicating the reference value of the signal.

[0097] The third value is the signal value when it fluctuates due to exposure of the sensor 10 to the sample gas during the second period T2. The third value is, for example, the signal value V3 at a predetermined time point during the second period T2, as shown in Fig. 7. Alternatively, the third value may be the average value of the signal values ​​during a predetermined interval S3 during the second period T2, as shown in Fig. 8. The length of the interval S3 is, for example, 0.1 seconds or more and 5 seconds or less.

[0098] The third value is, for example, the signal value at the timing when the signal value is maximum in the second period T2. In this case, the third value may be the signal value V3 at the time when the signal value is maximum in the second period T2, or may be the average value of the signal values ​​in the section S3 that includes the time when the signal value is maximum in the second period T2. Furthermore, the third value may be the signal value V3 at a time when a predetermined time has elapsed since the start of the second period T2, or may be the average value of the signal values ​​in the section S3 that starts after a predetermined time has elapsed since the start of the second period T2. The predetermined time is, for example, a time that is at least half the length of the second period T2.

[0099] The third value is a value that fluctuates due to adsorption of the sample gas to the sensor 10, and if the baseline of the signal shifts, the third value also shifts. Therefore, it is possible to extract a feature corresponding to the drift using the third value.

[0100] The extraction circuit 33 acquires at least one of a first value, a second value, and a third value from the signal output from the sensor 10, for example, in each of two or more measurement periods among a plurality of consecutive measurement periods Tm1 to Tm7. By acquiring signals in a plurality of consecutive measurement periods Tm1 to Tm7 in this way, drift in the acquired signals tends to increase, and the identification accuracy, which will be described later, can be improved.

[0101] 9 and 10 are diagrams for explaining the values ​​of the signal acquired by the extraction circuit 33 during a plurality of consecutive measurement periods Tm-1 to Tm-7. Each of Figs. 9 and 10 shows the change over time in the intensity (e.g., voltage) of the signal output from the sensor 10 and acquired by the acquisition circuit 32.

[0102] 9 and 10, when acquiring the first value, the extraction circuit 33 acquires at least one of the first values ​​for each of the first measurement period Tm-1 to the seventh measurement period Tm-7. The same process is carried out when acquiring the second and third values. When acquiring the first value, the extraction circuit 33 acquires the first value, the second value, or the third value in multiple measurement periods Tm, for example, at the same timing in each measurement period Tm.

[0103] Specifically, the extraction circuit 33 acquires, as the first value, for example, at least one of the signal values ​​V1-1 to V1-7 corresponding to the signal value V1 shown in Fig. 7. Alternatively, the extraction circuit 33 acquires, as the first value, for example, the average value of the signal values ​​in at least one of the sections S1-1 to S1-7 corresponding to the section S1 shown in Fig. 8.

[0104] Furthermore, the extraction circuit 33 acquires, as the third value, for example, at least one of the signal values ​​V3-1 to V3-7 corresponding to the signal value V3 shown in Fig. 7. Alternatively, the extraction circuit 33 acquires, as the third value, for example, the average value of the signal values ​​in at least one of the sections S3-1 to S3-7 corresponding to the section S3 shown in Fig. 8. Although not shown, the process for acquiring the second value is similar to that for the first and third values.

[0105] Next, a specific example of a method for extracting a feature amount by the extraction circuit 33 will be described.

[0106] First, a first example of a method for extracting a feature quantity by the extraction circuit 33 will be described. In the first example, the extraction circuit 33 acquires a first value for each of two or more measurement periods Tm among the consecutive measurement periods Tm-1 to Tm-7 shown in FIGS. 9 and 10, and extracts the difference between the acquired first values ​​as a feature quantity. For example, the extraction circuit 33 extracts the difference between the first values ​​for each of two spaced apart measurement periods Tm as a feature quantity. The two spaced apart measurement periods Tm are, for example, measurement periods Tm-2 and Tm-7, which are the second and last measurement periods Tm among the consecutive measurement periods Tm-1 to Tm-7. This tends to increase the difference between the first values, which tends to increase the accuracy of the identification described below.

[0107] Furthermore, the extraction circuit 33 may extract the difference between the first values ​​in two consecutive measurement periods Tm (for example, measurement periods Tm-2 and Tm-3) as the feature amount.

[0108] Furthermore, the extraction circuit 33 may extract multiple feature amounts by changing the combination of two measurement periods Tm for calculating the difference between the first values ​​in multiple ways. For example, the extraction circuit 33 may extract, as multiple feature amounts, the differences between the first values ​​in all combinations of two consecutive measurement periods Tm. When changing the combination of two measurement periods Tm in multiple ways, the extraction circuit 33 may extract each of the differences between the multiple extracted first values ​​as a feature amount, or may extract the average value of the differences between the multiple extracted first values ​​as a feature amount.

[0109] Furthermore, in the first example, the extraction circuit 33 may acquire a third value instead of the first value. That is, the extraction circuit 33 may acquire a third value for each of two or more measurement periods Tm among the consecutive measurement periods Tm-1 to Tm-7 shown in FIGS. 9 and 10, and extract the difference between the acquired third values ​​as a feature. In this case, the first value in the above description can be replaced with the third value. Furthermore, the extraction circuit 33 may extract both the difference between the first values ​​and the difference between the third values ​​as feature amounts.

[0110] In this way, in the first example, the extraction circuit 33 can extract a feature amount corresponding to drift indicating that the baseline of the signal has fluctuated by taking the difference between values ​​at the same timing during each of the multiple measurement periods Tm. Furthermore, because the extraction circuit 33 takes the difference between the first values ​​or the third values ​​acquired over the multiple measurement periods Tm, the extracted feature amount tends to be large.

[0111] Next, a second example of a feature extraction method by the extraction circuit 33 will be described. In the second example, the extraction circuit 33 acquires a first value and a second value during a measurement period Tm and extracts the difference between the acquired first and second values ​​as a feature. For example, the extraction circuit 33 acquires the first value and the second value during at least one measurement period Tm among a plurality of consecutive measurement periods Tm-1 to Tm-7 shown in FIGS. 9 and 10 and extracts the difference between the first value and the second value acquired during the same measurement period Tm as a feature. When extracting the differences between the first value and the second value during a plurality of measurement periods, the extraction circuit 33 may extract each of the differences between the extracted first value and the second value as a feature, or may extract the average value of the differences between the extracted first value and the second value as a feature.

[0112] In this way, in the second example, the extraction circuit 33 can extract a feature amount corresponding to the drift of the signal by calculating the difference between the first value and the second value that serves as a reference value.

[0113] Next, a third example of a feature extraction method by the extraction circuit 33 will be described. In the third example, the extraction circuit 33 acquires a first value in each of two or more measurement periods Tm among the plurality of consecutive measurement periods Tm-1 to Tm-7 shown in Figures 9 and 10. The extraction circuit 33 acquires, for example, a first value in each of all measurement periods Tm from the second measurement period Tm-2 onwards among the plurality of consecutive measurement periods Tm-1 to Tm-7.

[0114] Furthermore, among the measurement periods Tm after the second measurement period Tm-2, there may be a measurement period Tm from which the extraction circuit 33 does not acquire the first value. That is, the extraction circuit 33 may acquire the first value in one or more measurement periods Tm after the second measurement period Tm-2 among the consecutive measurement periods Tm-1 to Tm-7. For example, the extraction circuit 33 may acquire the first value in each of the measurement periods Tm from the third measurement period Tm-3 onwards or from the fourth measurement period Tm-4 onwards among the consecutive measurement periods Tm-1 to Tm-7.

[0115] Alternatively, the extraction circuit 33 may acquire the first value for each of all measurement periods Tm from a plurality of consecutive measurement periods Tm-1 to Tm-7.

[0116] The extraction circuit 33 then derives an approximation formula using the acquired first value and extracts the coefficients of the derived approximation formula as feature quantities. The approximation formula is, for example, an approximation formula in which the first value in each measurement period Tm is a function of time. The approximation formula is, for example, a linear formula (i.e., linear approximation) or a quadratic formula. The approximation formula may be a polynomial other than a quadratic formula, or may be an exponential, logarithmic, or power formula. A known method can be used to derive the approximation formula. In the case of a linear approximation formula, the approximation formula can be derived using, for example, the least squares method.

[0117] In addition, in the third example, the extraction circuit 33 may acquire a third value instead of the first value. That is, the extraction circuit 33 may acquire a third value for each of two or more measurement periods Tm among the consecutive measurement periods Tm-1 to Tm-7 shown in FIGS. 9 and 10, derive an approximate expression using the acquired third value, and extract the coefficient of the derived approximate expression as a feature. In this case, the first value in the above description can be replaced with the third value. Furthermore, the extraction circuit 33 may extract both the coefficient of the approximate expression using the first value and the coefficient of the approximate expression using the third value as feature quantities.

[0118] In this way, in the third example, an approximation formula corresponding to the baseline of the signal is derived, and therefore the extraction circuit 33 can extract a feature quantity corresponding to the drift of the signal. Furthermore, by extracting the coefficient of the approximation formula using the first value or the third value for each of the multiple measurement periods Tm as a feature quantity, it is possible to extract a feature quantity in which the variation in the first value or the third value is smoothed.

[0119] In the first and third examples, the number of measurement periods Tm for acquiring the first value or the second value may be three or more, or may be four or more.

[0120] In addition, the extraction circuit 33 may extract at least one feature using any one of the methods from the first to third examples described above, or may extract multiple feature amounts using two or more of the methods from the first to third examples described above.

[0121] Furthermore, in the first to third examples described above, the extraction circuit 33 may extract one or more feature quantities based on signals acquired during the second or subsequent measurement periods Tm (i.e., measurement periods Tm-2 and later) among the multiple measurement periods Tm-1 to Tm-7 shown in FIGS. 9 and 10 . This allows feature quantities to be extracted using signals output during the second or subsequent measurement periods Tm, which tend to exhibit large drift. Therefore, the extracted feature quantities tend to vary significantly depending on the type of sample gas, further improving the identification accuracy (described later). Furthermore, when the extraction circuit 33 extracts one or more feature quantities based on signals acquired during the second or subsequent measurement periods Tm, there may be a measurement period Tm during which the extraction circuit 33 does not acquire the first value. In other words, the extraction circuit 33 may extract one or more feature quantities based on signals acquired during any measurement period Tm from the second or subsequent measurement periods Tm. For example, the extraction circuit 33 may extract one or more feature amounts based on a signal acquired in the third or subsequent measurement period Tm among the multiple measurement periods Tm-1 to Tm-7, or may extract one or more feature amounts based on a signal acquired in the fourth or subsequent measurement period Tm among the multiple measurement periods Tm-1 to Tm-7.

[0122] The extraction circuit 33 may extract a feature quantity other than the feature quantity corresponding to the drift of the acquired signal. For example, the extraction circuit 33 may acquire, as the feature quantity, each signal value at a predetermined interval in at least one of the second period T2 and the third period T3. The predetermined interval is, for example, not less than 0.1 seconds and not more than 10 seconds.

[0123] Furthermore, to further improve the identification accuracy, the extraction circuit 33 may acquire, as a feature, at least one of the rate and amount of change in the signal value during at least one of the second period T2 and the third period T3, in addition to the feature corresponding to the drift. FIG. 11 is a diagram illustrating the rate (i.e., slope) and amount of change in the signal value output from the sensor 10. FIG. 11 shows the time change in the intensity (e.g., voltage) of the signal output from the sensor 10 acquired by the acquisition circuit 32. In the graph of FIG. 11, points a to d respectively indicate the time and signal value after a predetermined time has elapsed since the start of the second period T2, and points e and f respectively indicate the time and signal value after a predetermined time has elapsed since the start of the third period T3. The number of points and the position of each point are set depending on the type of sensor 10, the type of material to be identified, and the like. The position of each point may also be determined based on the shape of the signal waveform, rather than a predetermined time.

[0124] The extraction circuit 33 extracts, as a feature, for example, the slope between two points on a graph of signal intensity versus time. Examples of the extracted slopes include the slope SU1 of the line connecting points a and b in the second period T2, the slope SU2 of the line connecting points c and d in the second period T2 after point b, and the slope SD of the line connecting points e and f in the third period T3. The extraction circuit 33 also extracts, as a feature, the amount of change from a predetermined point in the graph of signal intensity versus time. Examples of the extracted change include the amount of change DU, which is the difference between the signal value at the start of the second period T2 and the signal value at point d, and the amount of change DD, which is the difference between the signal value at the start of the third period T3 and the signal value at point f.

[0125] (3) Identification step 4 again, in the next identification step, the identification circuit 34 uses the trained logical model for identifying the sample gas to identify the sample gas based on the one or more feature amounts extracted in step S13, and outputs the identification result (step S14). For example, the identification circuit 34 uses the trained logical model to input the one or more feature amounts and outputs information indicating which of a plurality of identification target substances is contained in the sample gas. The identification circuit 34 may also output information indicating whether the sample gas contains the identification target substance.

[0126] When the trained logical model includes a neural network, for example, one or more feature quantities are input to an input node of the neural network, and the output node outputs the probability that each of a plurality of identification target substances is contained in the sample gas. In other words, the number of input nodes is the number of one or more feature quantities input, and the number of output nodes is the number of a plurality of identification target substances to be identified. The trained logical model outputs the identification target substance that has the highest probability of being output from the output node among the plurality of identification target substances. Furthermore, the trained logical model may output whether or not the identification target substance is contained in the sample gas based on the probability output from the output node, for example, based on whether or not the probability is equal to or greater than a threshold value.

[0127] As described above, the gas identification method performed by gas identification system 100 is a gas identification method using sensor 10, and includes the following steps: an acquisition step (step S12) of acquiring a signal output from sensor 10; an extraction step (step S13) of extracting one or more feature quantities corresponding to drift in the acquired signal; and an output step (step S14) of identifying the sample gas based on the extracted one or more feature quantities using a trained logical model and outputting the identification result.

[0128] As a result, the extraction step extracts one or more features corresponding to the signal drift that are different from the signal output during the second period in which the sensor 10 is exposed to the sample gas, i.e., the feature dependent on the concentration of gas adsorbed to the sensor 10. Therefore, the identification step performs identification based on the signal drift, and the sample gas can be identified with high identification accuracy even when identifying sample gases with similar outputs from the sensor dependent on the concentration of gas adsorbed.

[0129] For example, in the sensor 10, it is possible to accurately distinguish between a first sample gas in which the drift of the output signal is large, as in the signal shown in Figure 6, and a second sample gas in which the signal waveform in the second period is the same as that of the first sample gas, but which, unlike the signal shown in Figure 6, hardly experiences any signal drift.

[0130] (Example) Next, the present disclosure will be specifically described based on examples, but the present disclosure is not limited to the following examples.

[0131] [Acquisition of discrimination test signal] First, the 16 sensors 10 arranged in the housing section 21 were used to acquire signals output from each of the 16 sensors 10.

[0132] <Sensor> The 16 sensors used each had a different material for the sensing part. For example, the sensors 10 used had a sensing part made of a resin material such as methylphenyl silicone (75% side chain phenyl group) or methylphenyl silicone (35% side chain phenyl group).

[0133] <Gas> The reference gas used was the air in the measurement chamber in which the container 21 was placed. The sample gases used were five types of sample gases A to E, obtained by volatilizing the following five types of chemical substances into the air in the measurement chamber. In other words, sample gases A to E each contain the corresponding chemical substances listed below. Sample gas A: Phenylethyl alcohol Sample gas B: Methylcyclopentenolone Sample gas C: Isovaleric acid Sample gas D: Undecalactone Sample gas E: Skatole

[0134] <Acquisition of voltage signal> The voltage signal acquisition operation involved exposing the 16 sensors 10 to the reference gas for a first period of 5 seconds, then exposing the 16 sensors 10 to the sample gas for a second period of 10 seconds, and then exposing the 16 sensors 10 to the reference gas for a third period of 25 seconds. In one voltage signal acquisition operation, the exposure operation for the measurement periods consisting of the first, second, and third periods was repeated seven consecutive times. Thus, in one acquisition operation, the 16 sensors 10 were exposed to the sample gas and the reference gas for seven consecutive measurement periods, and voltage signals output from each of the 16 sensors 10 were acquired. In other words, one signal voltage acquisition operation acquired a voltage signal set consisting of 16 different voltage signals corresponding to each of the 16 sensors 10. This voltage signal set acquisition operation was performed a total of 158 times for sample gases A to E. Specifically, 30 voltage signal sets were acquired for sample gas A, and 32 voltage signal sets were acquired for each of sample gases B to E.

[0135] [Feature extraction] Feature quantities were extracted from each of the 158 voltage signal sets corresponding to sample gases A to E obtained by the above-mentioned [Acquisition of signals for identification test].

[0136] <Reference example> In extracting the features for the reference example, five feature values, i.e., the slopes SU1, SU2, and SD and the variations DU and DD shown in FIG. 11, were extracted from each of the 16 voltage signals constituting the voltage signal set. In other words, a feature value set consisting of 16 values ​​x 5 = 80 feature values ​​was extracted for one voltage signal set. In this way, a feature value set was extracted for each of the 158 voltage signal sets, and the feature value sets for the 158 reference examples were extracted.

[0137] <Example> In the feature extraction in the working example, in addition to the same five feature values ​​as in the reference example, seven feature values ​​corresponding to signal drift were extracted from each of the 16 voltage signals constituting the voltage signal set: the difference between value V1-7 and value V1-2, and the difference between value V3-7 and value V3-2, as shown in FIG. 9 . In other words, a feature set consisting of 16 features x 7 = 112 feature values ​​was extracted for one voltage signal set. In this way, a feature set was extracted for each of the 158 voltage signal sets, and 158 feature sets in the working example were extracted.

[0138] [Identification Test] When the above-mentioned five types of chemical substances were selected as target substances for identification, a test was conducted to identify which of the five target substances was contained in the sample gas.

[0139] <Allocation of feature sets> The 158 sets of feature sets in the Reference Examples and Examples were randomly sorted by computer into 128 sets of training feature sets in the Reference Examples and Examples and 32 sets of prediction feature sets in the Reference Examples and Examples, regardless of the type of sample gas. Note that the sorting may be performed before the extraction of the features.

[0140] <Construction of a trained logical model> For the training feature sets in each of the Reference Example and the Working Example, 128 sets of training feature sets and the identification target substances contained in the sample gas corresponding to the 128 sets of training feature sets were used as training data, and machine learning was performed using a neural network including one hidden layer with five nodes to construct a trained logical model for identifying the identification target substances. The input to the neural network is the features that make up one feature set, and the output of the neural network is the probability that each of the five identification target substances is contained in the sample gas. In the trained logical model, the identification target substance with the highest probability is identified as the identification target substance contained in the sample gas.

[0141] <Sample gas identification> Using the features constituting the 32 prediction feature sets for each of the Reference Example and the Working Example as inputs, the trained logical model constructed above was used to identify the target substances contained in the sample gas. In addition, the allocation of the 158 sets of training feature sets and prediction feature sets was changed, and the trained logical model was reconstructed, and the target substances contained in the sample gas were identified a total of three times.

[0142] The results of the first to third classifications using the prediction feature set in the reference example are shown in Tables 1 to 3. The results of the first to third classifications using the prediction feature set in the working example are shown in Tables 4 to 6.

[0143] In Tables 1 to 6, the alphabets at the top and leftmost correspond to sample gas A to sample gas E, respectively. Each cell also records the number of times that the trained logical model identified the substance to be identified as contained in the sample gas listed at the leftmost part of the row in which the cell is located when the features constituting the prediction feature set extracted from the signals corresponding to sample gas A to sample gas E listed at the top of the column in which the cell is located were input. In other words, the numerical value of a cell in which the sample gas listed at the top and leftmost parts is the same indicates the number of times the identification was correct, and the numerical value of a cell in which the sample gas listed at the top and leftmost parts is different indicates the number of times the identification was incorrect.

[0144] [Table 1]

[0145] [Table 2]

[0146] [Table 3]

[0147] [Table 4]

[0148] [Table 5]

[0149] [Table 6]

[0150] As shown in Tables 1 to 3, in the reference example, there were many incorrect identifications when distinguishing between sample gases D and E. Of the 32 sets of prediction feature sets input, the percentage of incorrect identification results output, i.e., the percentage of misclassifications, was 12.5%, 15.6%, and 18.8% for the first to third runs, respectively, with the average for the first to third runs being 15.6%. There were no incorrect identifications when distinguishing between sample gases A and C.

[0151] In contrast, as shown in Tables 4 to 6, in the Example, the number of incorrect identifications was fewer in the identification of sample gas D and sample gas E than in the Reference Example. The percentage of incorrect identification results, i.e., the percentage of misidentifications, among the 32 sets of prediction feature sets input, was 9.4%, 3.1%, and 3.1% for the first to third runs, respectively, and the average for the first to third runs was 5.2%. Thus, in the Example, the percentage of misidentifications was reduced by more than 10% compared to the Reference Example. There were no misidentifications in the identification of sample gas A to sample gas C.

[0152] From the above results, it can be seen that the results of the example using features corresponding to drift have a lower rate of misjudgment and improved classification accuracy than the results of the reference example using features not corresponding to drift.

[0153] When checking the waveform of the signal obtained from sensor 10 (for example, sensor 10 whose sensing portion is made of methylphenyl silicone), the waveform of the signal during the second period was different when sensor 10 was exposed to sample gas A, when sensor 10 was exposed to sample gas B, and when sensor 10 was exposed to sample gas C. Therefore, it is considered that the accuracy of distinguishing sample gas A from sample gas C was high in both the reference example and the working example.

[0154] On the other hand, the signal waveforms in the second period were almost the same when the sensor 10 was exposed to sample gas D and when the sensor 10 was exposed to sample gas E. Furthermore, the drift of the signal corresponding to sample gas E was larger than the drift of the signal corresponding to sample gas D. Therefore, it is thought that erroneous determinations were particularly likely to occur in the reference example, and that the example using features corresponding to drift had higher identification accuracy.

[0155] (Other embodiments) While the gas identification system and gas identification method according to the present disclosure have been described above based on embodiments and examples, the present disclosure is not limited to these embodiments and examples. Various modifications conceivable by those skilled in the art to the embodiments and examples, as well as other configurations constructed by combining some of the components of the embodiments and examples, are also included within the scope of the present disclosure, provided they do not deviate from the gist of the present disclosure.

[0156] In the above embodiment, the exposure unit 20 exposes the sensor 10 to the reference gas during the first and third periods, but this is not limiting. The exposure unit 20 does not have to expose the sensor 10 to the sample gas during the first and third periods. For example, instead of exposing the sensor 10 to the reference gas, the exposure unit 20 may inhale the sample gas and expose the sensor 10 to a vacuum atmosphere.

[0157] Furthermore, for example, gas identifying system 100a includes detection device 200 and identification device 300, but the present invention is not limited to this. Gas identifying system 100a may be configured with only identification device 300. In this case, for example, step S11 in FIG. 4 may be omitted, and acquisition circuit 32a may acquire, for example, a signal from an already detected sensor via a network.

[0158] Furthermore, for example, in the above-described embodiments, some or all of the components of the gas identification system according to the present disclosure may be configured with dedicated hardware, or may be implemented by executing a software program appropriate for each component. Each component may be implemented by a program execution unit, such as a CPU or processor, reading and executing a software program stored on a recording medium, such as a hard disk drive or semiconductor memory.

[0159] Additionally, components of a gas identification system according to the present disclosure may be configured with one or more electronic circuits, each of which may be a general-purpose circuit or a dedicated circuit.

[0160] The one or more electronic circuits may include, for example, a semiconductor device, an integrated circuit (IC), or a large scale integration (LSI). The IC or LSI may be integrated on a single chip or on multiple chips. Although the IC or LSI is referred to here as an IC or LSI, the name may vary depending on the degree of integration, and may be called a system LSI, a very large scale integration (VLSI), or an ultra large scale integration (ULSI). Also, a field programmable gate array (FPGA), which is programmed after the LSI is manufactured, can be used for the same purpose.

[0161] Furthermore, the general or specific aspects of the present disclosure may be realized as a system, an apparatus, a method, an integrated circuit, or a computer program. Alternatively, they may be realized as a computer-readable non-transitory recording medium such as an optical disk, a HDD, or a semiconductor memory on which the computer program is stored. Alternatively, they may be realized as any combination of a system, an apparatus, a method, an integrated circuit, a computer program, and a recording medium.

[0162] For example, the present disclosure may be realized as a gas identification method executed by a computer such as a gas identification system, as a program for causing a computer to execute such a gas identification method, or as a computer-readable non-transitory recording medium having such a program recorded thereon. [Industrial Applicability]

[0163] The gas identification system and gas identification method according to the present disclosure are useful for identifying chemical substances and the like in gases. [Explanation of symbols]

[0164] 10 sensors 20 Exposed part 21 Storage unit 22 Three-way solenoid valve 23 Intake pump 25a, 25b, 25c, 25d, 25e piping 26a, 26b air intakes 26e Exhaust port 31 Control circuit 32, 32a acquisition circuit 33 Extraction circuit 34 Identification circuit 40 memory 50 Detector 51, 60 Communications Department 90 Wide Area Communication Network 100, 100a Gas Identification System 200 Detection Device 300 Identification Device P1, P2 input ports P3 output port

Claims

1. A gas identification method using a sensor that outputs a signal corresponding to an adsorption concentration of a gas, comprising: a first step of acquiring a signal output during a measurement period from the sensor exposed to a sample gas only during the second period of a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; a second step of extracting one or more features corresponding to drift of the acquired signal; a third step of identifying the sample gas based on the one or more extracted features using a trained logic model for identifying the sample gas, and outputting an identification result; In the second step, a first value is acquired, which is the value of the signal when the value of the signal that has fluctuated due to exposure of the sensor to the sample gas during the second period is returning to a reference value during the third period, and at least one feature amount among the one or more feature amounts is extracted using the acquired first value. Gas identification methods.

2. In the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods; In the second step, the first value is acquired in each of two or more of the measurement periods among the plurality of measurement periods, and a difference between the acquired first values ​​is extracted as at least one of the one or more feature amounts. The gas identifying method according to claim 1 .

3. In the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods; In the second step, the first value is acquired in each of two or more of the measurement periods among the plurality of measurement periods, an approximation formula is derived using the acquired first value, and a coefficient of the derived approximation formula is extracted as at least one feature amount among the one or more feature amounts. The gas identifying method according to claim 1 or 2.

4. In the second step, a second value that is a last value of the signal in the first period is acquired, and a difference between the acquired first value and the acquired second value is extracted as at least one feature amount of the one or more feature amounts. The gas identifying method according to any one of claims 1 to 3.

5. In the second step, a value of the signal at the end of the third period is acquired as the first value. The gas identifying method according to any one of claims 1 to 4.

6. A gas identification method using a sensor that outputs a signal corresponding to the adsorption concentration of a gas, comprising: a first step of acquiring a signal output during a measurement period from the sensor exposed to a sample gas only during the second period of a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; a second step of extracting one or more features corresponding to drift of the acquired signal; a third step of identifying the sample gas based on the one or more extracted features using a trained logic model for identifying the sample gas, and outputting an identification result; In the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods; In the second step, a third value is acquired during each of two or more of the measurement periods among the plurality of measurement periods, the third value being the value of the signal when the sensor fluctuates due to exposure to the sample gas during the second period, and a difference between the acquired third values ​​is extracted as at least one of the one or more feature amounts. Gas identification methods.

7. A gas identification method using a sensor that outputs a signal corresponding to the adsorption concentration of a gas, comprising: a first step of acquiring a signal output during a measurement period from the sensor exposed to a sample gas only during the second period of a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; a second step of extracting one or more features corresponding to drift of the acquired signal; a third step of identifying the sample gas based on the one or more extracted features using a trained logic model for identifying the sample gas, and outputting an identification result; In the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods; In the second step, a third value is acquired, which is the value of the signal when the signal fluctuates due to exposure of the sensor to the sample gas during the second period, for each of two or more of the plurality of measurement periods, and an approximation equation is derived using the acquired third value, and a coefficient of the derived approximation equation is extracted as at least one of the one or more feature quantities. Gas identification methods.

8. A gas identification method using a sensor that outputs a signal corresponding to the adsorption concentration of a gas, comprising: a first step of acquiring a signal output during a measurement period from the sensor exposed to a sample gas only during the second period of a measurement period consisting of a first period, a second period following the first period, and a third period following the second period; a second step of extracting one or more features corresponding to drift of the acquired signal; a third step of identifying the sample gas based on the one or more extracted features using a trained logic model for identifying the sample gas, and outputting an identification result; In the first step, the signal output from the sensor is acquired during a plurality of consecutive measurement periods; In the second step, at least one feature amount among the one or more feature amounts is extracted based on the signal acquired in a second or subsequent measurement period among the plurality of measurement periods. Gas identification methods.

9. a fourth step of exposing the sensor to the sample gas only during the second period of the measurement period; the signal acquired in the first step is a signal output during the measurement period from the sensor exposed to the sample gas only during the second period of the measurement period in the fourth step; The gas identifying method according to any one of claims 1 to 8.

10. In the fourth step, the sensor is exposed to a reference gas for the first period and the third period. The gas identifying method according to claim 9.

11. In the first step, the signal output from the sensor is acquired via a network. The gas identifying method according to any one of claims 1 to 8.

12. a sensor that outputs a signal corresponding to the adsorption concentration of the gas; an exposure unit that exposes the sensor to the sample gas only during the second period of a measurement period that includes a first period, a second period following the first period, and a third period following the second period; an acquisition circuit for acquiring a signal output from the sensor during the measurement period; an extraction circuit for extracting one or more feature quantities corresponding to drift of the acquired signal; a memory in which a trained logic model for identifying the sample gas is stored; an identification circuit that uses the trained logical model to identify the sample gas based on the one or more extracted feature quantities and outputs an identification result; the extraction circuit acquires a first value, which is the value of the signal when the value of the signal that has fluctuated due to exposure of the sensor to the sample gas during the second period is returning to a reference value during the third period, and extracts at least one feature amount from the one or more feature amounts using the acquired first value. Gas identification system.

13. A sensor that outputs a signal corresponding to the adsorption concentration of a gas; an exposure unit that exposes the sensor to the sample gas only during the second period of a measurement period that includes a first period, a second period following the first period, and a third period following the second period; an acquisition circuit for acquiring a signal output from the sensor during the measurement period; an extraction circuit for extracting one or more feature quantities corresponding to drift of the acquired signal; a memory in which a trained logic model for identifying the sample gas is stored; an identification circuit that uses the trained logical model to identify the sample gas based on the one or more extracted feature quantities and outputs an identification result; the acquisition circuit acquires the signal output from the sensor during a plurality of consecutive measurement periods; the extraction circuit acquires, during each of two or more of the measurement periods among the plurality of measurement periods, a third value that is the value of the signal when the sensor fluctuates due to exposure to the sample gas during the second period, and extracts a difference between the acquired third values ​​as at least one of the one or more feature amounts. Gas identification system.

14. A sensor that outputs a signal corresponding to the adsorption concentration of a gas; an exposure unit that exposes the sensor to the sample gas only during the second period of a measurement period that includes a first period, a second period following the first period, and a third period following the second period; an acquisition circuit for acquiring a signal output from the sensor during the measurement period; an extraction circuit for extracting one or more feature quantities corresponding to drift of the acquired signal; a memory in which a trained logic model for identifying the sample gas is stored; an identification circuit that uses the trained logical model to identify the sample gas based on the one or more extracted feature quantities and outputs an identification result; the acquisition circuit acquires the signal output from the sensor during a plurality of consecutive measurement periods; the extraction circuit acquires, for each of two or more of the measurement periods among the plurality of measurement periods, a third value that is the value of the signal when the signal fluctuates due to exposure of the sensor to the sample gas during the second period, derives an approximation equation using the acquired third value, and extracts a coefficient of the derived approximation equation as at least one feature amount among the one or more feature amounts. Gas identification system.

15. A sensor that outputs a signal corresponding to the adsorption concentration of a gas; an exposure unit that exposes the sensor to the sample gas only during the second period of a measurement period that includes a first period, a second period following the first period, and a third period following the second period; an acquisition circuit for acquiring a signal output from the sensor during the measurement period; an extraction circuit for extracting one or more feature quantities corresponding to drift of the acquired signal; a memory in which a trained logic model for identifying the sample gas is stored; an identification circuit that uses the trained logical model to identify the sample gas based on the one or more extracted feature quantities and outputs an identification result; the acquisition circuit acquires the signal output from the sensor during a plurality of consecutive measurement periods; the extraction circuit extracts at least one feature amount from the one or more feature amounts based on the signal acquired in a second or subsequent measurement period among the plurality of measurement periods; Gas identification system.

Images