Information processing system, method for processing information, control program, and recording medium

Abstract

To estimate the physical property of an odor material detected by an odor sensor and display visualized estimated physical property.SOLUTION: An information processing system (100) includes: an estimation unit (212) for estimating at least one type of index corresponding to a detection signal from a detection signal of an odor sensor by using an estimation model (221) that has learned the correspondence between the detection signal (114) of the odor sensor (10) and at least one type of index relating to intermolecular interaction; and an output control unit (214) for causing an output device (3) to output an estimation result.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing system, an information processing method, a control program, and a recording medium. [Background technology]

[0002] A technique for mapping and displaying the results of detection by an odor sensor onto a two-dimensional graph using principal component analysis is known in the prior art. For example, Patent Document 1 describes a graph in which the odors of pure water, black coffee, wine, and soy sauce were detected by a sensor and the detection results were subjected to principal component analysis. This graph has two coordinate axes, with the horizontal axis representing the first principal component PC1 and the vertical axis representing the second principal component PC2. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-114254 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology described in Patent Document 1, the principal components PC1 and PC2 are not coordinate axes corresponding to values ​​with chemical meaning, so while it is possible to classify odors, it is not possible to evaluate the physical properties of the odors.

[0005] One aspect of the present invention has been made in consideration of the above-mentioned problems, and aims to provide an information processing system, etc. that outputs an indicator related to the physical properties of a detected odor from a detection signal obtained from an odor sensor. [Means for solving the problem]

[0006] In order to solve the above problem, an information processing system according to one embodiment of the present invention comprises an estimation unit that uses an estimation model that has learned a correspondence between a detection signal obtained from an odor sensor when a gas containing an odorous substance is supplied to the odor sensor and at least one type of indicator related to the intermolecular interaction of the odorous substance to estimate at least one type of indicator corresponding to the target detection signal from the target detection signal obtained from the odor sensor to which a target gas has been supplied, and an output control unit that outputs the estimation result of the estimation unit from an output device.

[0007] In addition, in order to solve the above-mentioned problems, an information processing method according to one embodiment of the present invention is an information processing method executed by an information processing system, and includes an estimation step of estimating at least one type of indicator corresponding to a target detection signal obtained from an odor sensor supplied with a target gas, using an estimation model that has learned a correspondence between the detection signal obtained from the odor sensor when a gas containing an odorous substance is supplied to the odor sensor and at least one type of indicator related to the intermolecular interaction of the odorous substance, and an output control step of outputting the estimation result to an output device. [Effects of the Invention]

[0008] According to one aspect of the present invention, it is possible to provide an information processing system or the like that outputs an index relating to the physical properties of a detected odor from a detection signal acquired from an odor sensor. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram illustrating an example of a configuration of a main part of an odor detection system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram showing an example of the configuration of the odor sensor element shown in FIG. [Figure 3] 1A and 1B are diagrams illustrating the detection of odorous substances by an odor sensor element. [Figure 4] FIG. 1 is an example of a three-dimensional graph according to an embodiment of the present invention, showing the HSP values ​​of odorous substances. [Figure 5]FIG. 1 is a diagram illustrating learning data used in machine learning. [Figure 6] 2 is a flowchart showing an example of the flow of an odor estimation process executed by a control unit of the odor estimation device shown in FIG. [Figure 7] FIG. 10 is a graph showing a comparison between analytical values ​​of HSP values ​​of gases estimated by the odor estimation device and literature values. [Figure 8] FIG. 10 is a graph showing a comparison between analytical values ​​of Hildebrand solubility parameters of gases estimated by the odor estimation device and literature values. [Figure 9] This is a diagram showing a comparison of analytical values ​​of the HSP values ​​and molecular weights of each gas estimated by the odor estimation device with literature values. [Figure 10] FIG. 1 is a diagram illustrating a conventional method for expressing odor substances. DETAILED DESCRIPTION OF THE INVENTION

[0010] [Embodiment] <Configuration of odor detection system 100> DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will now be described in detail. Fig. 1 is a diagram showing an example of the configuration of the main parts of an odor detection system 100 (information processing system) according to an embodiment of the present invention.

[0011] 1, the odor detection system 100 includes an odor detection device 1, an odor estimation device 2 (information processing device, information processing system), and an output device 3. The odor estimation device 2 is connected to the odor detection device 1 and the output device 3 so that they can communicate with each other.

[0012] The odor detection system 100 is not limited to this, and for example, the odor estimation device 2 may be configured to include the output device 3, or may be configured to include the odor detection device 1. Furthermore, the odor detection system 100 may be configured to include both the output device 3 and the odor detection device 1.

[0013] In the following, an example will be described in which the odor estimation device 2 is a single device, but the odor detection system 100 is not limited to this configuration. For example, in the odor detection system 100, some of the functions of the control unit 21 (described later) of the odor estimation device 2 may be provided in a device other than the odor estimation device 2 that is communicatively connected to the odor estimation device 2. In this case, the same processing as that of the odor estimation device 2 is performed by multiple devices instead of the odor estimation device 2.

[0014] The odor detection device 1 includes an odor sensor 10, a sensor chamber 12, and a communication unit 13 for communicating with other devices.

[0015] The sensor chamber 12 has a storage space in which the odor sensor 10 is placed, and a gas containing an odorant to be detected is supplied into the storage space. Here, the odorant may contain one type of molecule (odor molecule) that adsorbs to a specific olfactory receptor. Alternatively, the odorant may be a mixture containing multiple types of odor molecules that adsorb to a specific olfactory receptor.

[0016] The odor sensor 10 is a sensor that detects odorous substances and includes multiple odor sensor elements 11 with different characteristics. The number of odor sensor elements 11 included in the odor sensor 10 is not particularly limited, and may be one or 64. In one example, the number of odor sensor elements 11 included in the odor sensor 10 may be four or more, or 32 or less.

[0017] The multiple odor sensor elements 11 each have different characteristics and are therefore capable of adsorbing different types of odor substances. Each odor sensor element 11 detects a change in resistance (ΔR) due to the adsorption of an odor substance and outputs a detection signal for each odor sensor element 11. The odor detection device 1 transmits detection data 114, which includes the detection signals output by each odor sensor element 11, to the odor estimation device 2 via the communication unit 13. Details of the odor detection device 1 will be described later.

[0018] The odor estimation device 2 includes a control unit 21, a storage unit 22, and a communication unit 23 for communicating with other devices.

[0019] The storage unit 22 stores various data used by the control unit 21 and an estimation model 221. The estimation model 221 is a trained model generated by machine learning. Details of the estimation model 221 will be described later. The storage unit 22 may be a storage unit within the odor estimation device 2, or may be an external storage communicatively connected to the outside of the odor estimation device 2.

[0020] The control unit 21 comprehensively controls each unit of the odor estimation device 2. The control unit 21 includes an acquisition unit 211, an odor estimation unit 212 (estimation unit), a graph data generation unit 213, an output control unit 214, and a communication control unit 215.

[0021] The acquisition unit 211 acquires the detection data 114 transmitted from the odor detection device 1 via the communication unit 23. The detection data 114 is data including detection signals indicating changes in resistance value (ΔR) detected by each of the multiple odor sensor elements 11, as described above.

[0022] The odor estimation unit 212 estimates an index of an odor substance from the detection data 114 using an estimation model 221 generated by machine learning. The index of an odor substance may be, for example, the Hansen Solubility Parameter (HSP) value of the odor substance, but is not limited to this. The following describes an example in which the index of an odor substance estimated by the odor estimation unit 212 is an HSP value. Specifically, the estimation model 221 is read from the memory unit 22, and the detection data 114 is input to the estimation model 221 to estimate the HSP value of the odor substance.

[0023] The graph data generation unit 213 generates three-dimensional graph data from the estimated HSP values. The output control unit 214 displays the three-dimensional graph data on the display unit 31 of the output device 3 via the communication unit 23. The communication control unit 215 transmits the three-dimensional graph data to an external device other than the output device 3 (e.g., a data management device that stores and manages the estimation results) via the communication unit 23. Note that the output control unit 214 may transmit the estimated HSP values ​​(i.e., the coordinate values ​​of the points corresponding to the odor substances plotted on the three-dimensional graph) instead of the generated three-dimensional graph data.

[0024] The output device 3 includes a display unit 31 and a communication unit 32 for communicating with other devices. The display unit 31 is, for example, a liquid crystal display, an organic EL display, or electronic paper, and displays various types of information.

[0025] The output device 3 displays on the display unit 31 the estimated HSP value of the odor substance and / or a three-dimensional graph showing the estimated HSP value of the odor substance (see Figure 4) from the three-dimensional graph data received via the communication unit 32.

[0026] Unlike PC1, PC2, and other numerical values ​​obtained by principal component analysis, HSP values ​​are numerical values ​​that indicate the physical properties of odor substances. Therefore, the odor estimation device 2 can generate a three-dimensional graph that visualizes the physical properties of the estimated odor substances. This makes it easy to understand the physical properties of the detected odor substances, making it easy to consider, for example, methods for removing the odor substances. Details of the three-dimensional graph showing HSP values ​​will be described later.

[0027] <Odor sensor element 11> Next, the odor sensor element 11 will be described with reference to the drawings. Each odor sensor element 11 outputs a value indicating the change in electrical conductivity of the odor sensor element 11 over time before and after an odor substance is adsorbed to the odor sensor element 11. This allows the odor detection device 1 to detect a variety of odor substances.

[0028] Fig. 2 is a diagram showing an example of the configuration of the odor sensor element 11. As shown in Fig. 2, the odor sensor element 11 is, for example, a chemiresistor-type odor sensor element, and includes a receptive layer 111 (detection unit), a pair of electrodes 112, and a substrate 113. The receptive layer 111 and the pair of electrodes 112 are disposed on the substrate 113.

[0029] The receptor layer 111 includes various resins 111a (adsorbent molecules) and conductive materials 111b uniformly contained in the resins 111a, and selectively adsorbs the odor substances to be detected. The receptor layer 111 may further include a surfactant.

[0030] Examples of the composition of resin 111a include polyvinylidene fluoride, polyolefin, silicone resin, (meth)acrylic resin, polyester resin, polycarbonate, polyacetal resin, furan resin, ketone resin, polyvinyl chloride resin, polyurethane resin, polyamide resin, polyimide resin, polyallylamine resin, polyvinyl acetal resin, and polyether resin. Modified versions or copolymers of these may also be used, with silicone resin, polyester resin, and polyvinyl acetal resin being preferred. Resin 111a is not particularly limited as long as it is a resin that adsorbs odorous substances. Because different substances are easily adsorbed depending on the composition of resin 111a, odorous substances can be selectively adsorbed onto receiving layer 111 by changing the composition of resin 111a.

[0031] The conductive material 111b is realized by, for example, carbon black, metal particles, conductive polymers, etc. Examples of metal particles include silver, copper, nickel, aluminum, and alloys thereof. Examples of conductive polymers include polyaniline, polythiophene, polypyrrole, polyacetylene, polyphenylene vinylene, and polynaphthalene. When the conductive material 111b is a conductive polymer, the odorant can also be adsorbed onto the conductive material 111b, and therefore the conductive material 111b can also be an adsorbed molecule.

[0032] The electrode 112 includes a pair of electrodes 112a and 112b spaced apart from each other.

[0033] Receptor layer 111 is disposed, for example, between electrode 112a and electrode 112b so as to be in contact with both electrode 112a and electrode 112b. When a voltage is applied between electrode 112a and electrode 112b, a current flows from electrode 112a to electrode 112b via receptor layer 111.

[0034] When an odorant is adsorbed to the receptive layer 111, the receptive layer 111 swells due to the adsorbed odorant. As a result, it becomes more difficult for current to flow through the receptive layer 111 compared to before adsorption, and the resistance value (R) increases.

[0035] At this time, the detected resistance value (R) differs depending on the odorant being adsorbed. For example, the resistance value (R) differs when odorant A is adsorbed from when odorant B, which is different from odorant A, is adsorbed. By detecting this change in resistance value (R), odor detection device 1 can detect and identify the adsorbed odorant.

[0036] The composition of the receptor layers 111 included in the multiple odor sensor elements 11 may be the same or different. When the receptor layers 111 have the same composition, the multiple odor sensor elements 11 can detect the same odor substance.

[0037] Furthermore, if each receiving layer 111 has a different composition, the amount of adsorption will differ even for the same odor substance depending on the composition of the receiving layer 111, and each of the multiple odor sensor elements 11 can output a different rate of change in resistance value (R) for the same odor substance.

[0038] By combining multiple odor sensor elements 11 having receptor layers 111 of different compositions, the odor sensor 10 can detect odor substances containing various components that have been difficult to distinguish in the past. This also improves the accuracy of odor substance identification by the odor sensor 10. The combination of odor sensor elements 11 can be changed as needed depending on the type of odor substance to be detected.

[0039] The following is an example of the configuration of the receiving layer 111 when eight odor sensor elements 11 are provided. 1) relates to the resin 111a, 2) relates to the conductive material 111b, 3) relates to the surfactant, and 4) is the weight ratio of the resin 111a described in 1), the conductive material 111b described in 2), and the surfactant described in 3).

[0040] (1)1) Polyvinyl butyral (Mobital (registered trademark) B75H, manufactured by Kuraray Co., Ltd.) 2) SUPER C-65 (MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon (registered trademark) DA-325, manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (2) 1) Polymethyl methacrylate (Fujifilm Wako Pure Chemical Industries, Ltd.) 2) SUPER C-65 (Manufactured by MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (3) 1) Polyvinyl chloride (Fujifilm Wako Pure Chemical Industries, Ltd.) 2) SUPER C-65 (Manufactured by MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (4)1) Silicone resin (DOWSIL (registered trademark) RSN-0255 Flake Resin, manufactured by DOW Corporation) 2) SUPER C-65 (MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (5) 1) Cellulose acetate (Fujifilm Wako Pure Chemical Industries, Ltd.) 2) SUPER C-65 (MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (6) 1) Polyvinylpyrrolidone (Polyvinylpyrrolidone K90, Fujifilm Wako Pure Chemical Industries, Ltd.) 2) SUPER C-65 (MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (7) 1) Polyvinylidene fluoride (W#850, manufactured by Kureha Corporation) 2) SUPER C-65 (MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 (8) 1) Polyester (SunEstar 4610, manufactured by Sanyo Chemical Industries, Ltd.) 2) SUPER C-65 (MTI Corporation, USA) 3) Polyether phosphate ester amine (Disparlon DA-325 manufactured by Kusumoto Chemicals Co., Ltd.) 4) 72:20:8 The odor sensor element 11 is not limited to a chemiresistor-type odor sensor element, and may include one or more types of odor sensor elements used in known odor sensors. For example, it may be a QCM odor sensor using a quartz crystal oscillator, or a MEMS piezo-type sensor equipped with a piezoelectric thin film such as lead zirconate titanate that identifies odors based on changes in resonance frequency. It may also be an odor sensor using a membrane-type surface stress sensor (MSS), or a metal oxide semiconductor sensor that identifies odors based on changes in the resistance value of the semiconductor.

[0041] <Detection of odor substances> Figure 3 is a diagram illustrating the detection of odor substances by the odor sensor element 11. The upper diagram in Figure 3 is a diagram illustrating the detection of odor substances in the sensor chamber 12 in chronological order, and the lower diagram is a graph showing the change in resistance value (R) corresponding to the upper diagram.

[0042] 3, first, in the first step (time: T0 to T1), nitrogen gas N2 is supplied as a purge gas into sensor chamber 12 to expel odor substances remaining in sensor chamber 12. At this time, the resistance value (R) of odor sensor element 11 remains almost constant with little change. Instead of supplying nitrogen gas N2 as a purge gas, argon gas Ar, air, or the like may be supplied as a purge gas.

[0043] Next, in the second step (time: T1 to T2), a gas containing the odor substance to be detected (target gas) is supplied into the sensor chamber 12. The odor substance is adsorbed to the receptive layer 111 of the odor sensor element 11, causing the receptive layer 111 to swell. As the receptive layer 111 swells, the resistance value (R) increases.

[0044] Once detection is complete, in the third step (time: T2 to T3), nitrogen gas N2 is supplied into sensor chamber 12 as a purge gas to discharge the odorant from sensor chamber 12. At this time, as the odorant is discharged, the resistance value (R) decreases and returns to the resistance value (R) before the gas containing the odorant was supplied. Instead of supplying nitrogen gas N2 as a purge gas, argon gas Ar, air, or the like may be supplied as a purge gas.

[0045] The odor detection device 1 outputs, as detection data 114, a value related to at least either the signal intensity or the change over time of the detection signal of the resistance value (R) for each of a plurality of odor sensor elements 11. The odor detection device 1 in the present embodiment outputs, as detection data 114, the resistance value (R) and the detection time for each of the plurality of odor sensor elements 11. Here, the detection time may be the time when each detection signal is output.

[0046] Further, the detection data 114 may be the signal intensity ratio of the detection signals output by the plurality of odor sensor elements 11, or may be the difference in the signal intensity of the detection signals output by other odor sensor elements 11 based on the signal intensity of the detection signal output by a specific odor sensor element 11.

[0047] The odor detection device 1 transmits the detection data 114 to the odor estimation device 2. The odor estimation device 2 estimates the HSP value of the odor substance from the received detection data 114 using the estimation model 221 and generates a three-dimensional graph. The odor estimation device 1 may output the generated three-dimensional graph and / or the estimated HSP value.

[0048] <Three-dimensional graph of HSP value> FIG. 10 is a diagram for explaining a conventional representation method of odor substances, and is an example of a scatter diagram in which the detected odor substances are plotted with the horizontal axis being the first principal component PC1 and the vertical axis being the second principal component PC2. As shown in FIG. 10, conventionally, the odor substances detected by an odor sensor or the like are displayed as a two-dimensional graph by principal component analysis. The numerical values of the detected odor substances are plotted to show which group among alcohol-based, aromatic-based, and amine-based groups they are close to.

[0049] However, the coordinate axes indicated by the principal components PC1 and PC2 are not coordinate axes corresponding to values having chemical meanings. Therefore, although the classification of odor substances can be done depending on which group's plot the plot of the detected odor substance is close to, the physical properties of the detected odor substance cannot be evaluated based on the coordinate position of the plot.

[0050] FIG. 4 is an example of a three-dimensional graph according to an embodiment of the present invention, showing the HSP values ​​of odor substances.

[0051] 4, the odor estimation device 2 estimates HSP values ​​from the received detection data 114 using an estimation model 221 and generates a three-dimensional graph. The estimation model 221 is a trained model generated by machine learning. Details of the prediction of HSP values ​​using the estimation model 221 will be described later.

[0052] Unlike PC1, PC2, and other numerical values ​​obtained by principal component analysis, the HSP values ​​obtained by the above method are numerical values ​​that indicate the physical properties of odor substances, so the estimated physical properties of odor substances can be visualized and displayed in a three-dimensional graph. This makes it easy to understand the physical properties of the detected odor substances, making it easy to consider methods for removing odor substances, for example.

[0053] The HSP value is a type of solubility parameter and is an index of the intermolecular interactions of a substance. The HSP value is an index that takes into account the polarity of the substance, and is calculated by multiplying the Hildebrand solubility parameter (SP value) by the dispersion force δ d (Dispersion), polar interaction δ p (Polarity), hydrogen bond δ h (Hydrogen Bond)

[0054] Dispersion force 421 (first index) is an index related to non-polar intermolecular interactions, and is the force that acts between non-polar molecules with no charge separation. Polar interaction 422 (second index) is an index related to intermolecular interactions between dipoles, and is the force that acts between polarized molecules. Hydrogen bond 423 (third index) is an index related to intermolecular interactions that are different from non-polar interactions and dipole-dipole interactions, and is a bond based on electrostatic attraction (Coulomb force). It is known that substances with similar HSP values ​​tend to mix easily.

[0055] In this way, the three-dimensional graph according to the embodiment of the present invention is generated using three coordinate axes representing the three components of the HSP value, which is an index showing the physical properties of an odorant: dispersion force 421, polar interaction 422, and hydrogen bond 423. This allows odorants to be evaluated and classified based on their chemically meaningful physical properties.

[0056] The value estimated by the odor estimation unit 212 is not limited to the HSP value, and may be an index related to the intermolecular interaction of odor substances. For example, the value estimated by the odor estimation unit 212 may be an SP value, a molecular weight, a vapor pressure, or the like.

[0057] <HSP score estimation using machine learning> Next, generation of the estimation model 221 by machine learning will be described with reference to the drawings. The estimation model 221 is a trained model generated by machine learning using the training data 4.

[0058] 5 is a diagram illustrating the training data 4 used in machine learning. As shown in FIG. 5, the training data 4 includes, as explanatory variables 41, detection data (detection signals) 114 acquired from an odor sensor 10 supplied with a gas containing an odorant (a gas containing an odorant). Specifically, the explanatory variables 41 are detection signals of resistance values ​​(R) output by each of the multiple odor sensor elements 11.

[0059] In addition, the HSP value including three components, dispersion force 421, polar interaction 422, and hydrogen bond 423, is included as the objective variable 42.

[0060] The estimation model 221 is generated using a known machine learning algorithm, such as a k-nearest neighbor method, a logistic regression, a support vector machine, a random forest, and a neural network.

[0061] Through machine learning using the learning data 4, an estimated model 221 is generated that learns the correspondence between the detection data 114 of the resistance value (R) obtained from the odor sensor 10 and the HSP value, which includes three components of the odor substance's dispersion force 421, polar interaction 422, and hydrogen bond 423.

[0062] The odor estimation device 2 uses the estimation model 221 to estimate that the HSP value corresponding to the detection data 114 (target detection signal) acquired from the odor sensor 10 to which an odor substance (target gas) has been supplied is the HSP value of the odor substance.

[0063] <Flow of odor estimation process by control unit 21> An example of the flow of processing performed by the control unit 21 of the odor estimation device 2 as described above will be described with reference to Fig. 6. Fig. 6 is a flowchart showing an example of the flow of odor estimation processing performed by the control unit 21.

[0064] First, the acquisition unit 211 acquires the detection data 114 transmitted from the odor detection device 1 via the communication unit 23, and sends the acquired detection data 114 to the odor estimation unit 212 (S1). Note that the acquisition unit 211 may also send the detection data 114 to the memory unit 22 and store the detection data 114 in the memory unit 22.

[0065] The odor estimation unit 212 receives the detection data 114 sent from the acquisition unit 211. The odor estimation unit 212 estimates the HSP values ​​of the detected odor substance from the received detection data 114 using the estimation model 221 (S2: estimation step). The odor estimation unit 212 sends the estimated HSP values ​​to the graph data generation unit 213. The odor estimation unit 212 may also send the estimated HSP values ​​to the memory unit 22, which may store the HSP values ​​associated with the detection data 114.

[0066] The graph data generation unit 213 receives the estimated HSP values ​​sent from the odor estimation unit 212. The graph data generation unit 213 plots the received HSP values ​​on a three-dimensional graph using three components of the HSP values, namely, dispersion force 421, polar interaction 422, and hydrogen bond 423, as the three coordinate axes, to generate data of the three-dimensional graph (S3). The graph data generation unit 213 sends the generated three-dimensional graph data to the output control unit 214. The generated three-dimensional graph data may be stored in the memory unit 22 in association with the detection data 114.

[0067] The output control unit 214 transmits the three-dimensional graph data sent from the graph data generation unit 213 to the output device 3 so that it can be displayed on the display unit 31 (S4: output control step). This allows the odor detection system 100 to display coordinate values ​​generated from the estimated HSP values ​​of the odor substance and / or a three-dimensional graph that visualizes the physical properties of the odor substance on the display unit 31. The output control unit 214 may also transmit the estimated HSP values ​​and / or the generated three-dimensional graph data.

[0068] [Variation 1] Each process described in the above embodiment can be modified as appropriate. For example, in the above embodiment, the graph data generation unit 213 plots the estimated HSP value on a three-dimensional graph with three components of the HSP value, i.e., dispersion force 421, polar interaction 422, and hydrogen bond 423, as three coordinate axes to generate three-dimensional graph data.

[0069] The graph to be generated may be any graph that visualizes and displays the physical properties of the odorant; for example, a four-dimensional graph may be created in which the inner product 424 (fourth index) of dispersion forces 421, polar interactions 422, and hydrogen bonds 423 is used as the fourth coordinate axis.

[0070] In this case, the objective variable 42 of the training data 4 includes the dispersion force 421 , the polar interaction 422 , the hydrogen bond 423 , and the inner product 424 .

[0071] 9, a four-dimensional graph may be created in which the molecular weight of the odorant is used as the fourth coordinate axis. In this case, the objective variable 42 includes the molecular weight in addition to the dispersion force 421, polar interaction 422, and hydrogen bond 423.

[0072] [Variation 2] Furthermore, the graph data generation unit 213 may generate graph data having at least one of the dispersion force 421, polar interaction 422, hydrogen bond 423, and inner product 424 as coordinate axes. The graph data generation unit 213 generates data of a one-dimensional graph when there is one coordinate axis, a two-dimensional graph when there are two coordinate axes, and a three-dimensional graph when there are three coordinate axes.

[0073] In this case, the training data 4 also has at least one of dispersion force 421, polar interaction 422, hydrogen bond 423, and inner product 424 as the objective variable 42 corresponding to the coordinate axes of the graph. [Example]

[0074] Example 1 of the present invention will be described below. In Example 1, an odor sensor 10 equipped with the above-described eight odor sensor elements 11 was placed in a sensor chamber 12, and gases containing ethanol, propanol, and toluene, each with a known HSP value, were used as detection targets. Each gas was supplied into the sensor chamber 12 for an exposure time of approximately 30 seconds. Before detecting each gas, nitrogen gas N2 was supplied into the sensor chamber 12 as a purge gas, and any gas remaining in the sensor chamber 12 was evacuated. From the rate of change of the resistance value (R) detected by the odor sensor element 11, an estimation model 221 was used to calculate the three components of the HSP value of each gas, the dispersion term (MPa), 0.5 , polarity term (MPa) 0.5 and hydrogen bond term (MPa) 0.5 was estimated and compared with literature values.

[0075] Figure 7 shows a comparison of analytical values ​​and literature values ​​for the HSP values ​​of each gas estimated by the odor estimation device 2. For ethanol, propanol, and toluene, there was almost no significant difference between the analytical values ​​and literature values, indicating that the estimation was generally accurate. [Example]

[0076] In Example 2, under the same experimental conditions as in Example 1, the Hildebrand solubility parameters of each gas, including isopropanol, 2-methyl-1-propanol, and toluene, were analyzed using the estimation model 221 and compared with literature values.

[0077] FIG. 8 shows the Hildebrand solubility parameters (cal / cm) of each gas estimated by the odor estimation device 2. 3 ) 0.5 This figure shows a comparison of analytical values ​​and literature values ​​for isopropanol, 2-methyl-1-propanol, and toluene. It was found that there was almost no significant difference between the analytical values ​​and literature values, and that the estimation was generally correct. [Example]

[0078] In Example 3, under the same experimental conditions as in Example 1, the HSP value and molecular weight of each gas including ethanol, propanol, and heptane were estimated using the estimation model 221 and compared with literature values.

[0079] Figure 9 shows the HSP values ​​(MPa) of each gas estimated by the odor estimation device 2. 0.5 This figure shows a comparison of analytical values ​​and literature values ​​for molecular weight (g / mol). For all gases, ethanol, propanol, and heptane, there was almost no significant difference between the analytical values ​​and literature values, indicating that the estimation was generally accurate.

[0080] [Software implementation example] The functions of the odor estimation device 2 (hereinafter referred to as the "device") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 21).

[0081] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The control device and storage device execute the program, thereby realizing the functions described in each of the above embodiments.

[0082] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0083] Furthermore, some or all of the functions of the control blocks can be realized by logic circuits. For example, an integrated circuit in which a logic circuit that functions as each of the control blocks is formed is also included in the scope of the present invention. In addition, the functions of the control blocks can also be realized by, for example, a quantum computer.

[0084] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0085] 〔summary〕 The information processing system according to aspect 1 of the present invention comprises an estimation unit that, when a gas containing an odorous substance is supplied to an odor sensor, estimates at least one type of indicator corresponding to the target detection signal from the target detection signal acquired from the odor sensor to which a target gas is supplied, using an estimation model that has learned a correspondence between the detection signal acquired from the odor sensor and at least one type of indicator related to the intermolecular interaction of the odorous substance, and an output control unit that outputs the estimation result of the estimation unit from an output device.

[0086] In the information processing system of aspect 2 of the present invention, in the above aspect 1, the estimation model may include the detection signal obtained from the odor sensor to which the gas is supplied as an explanatory variable, and may be generated by machine learning using training data that includes at least one of the following as objective variables: (1) a first index related to non-polar intermolecular interactions of the odor substance; (2) a second index related to intermolecular interactions between dipoles; (3) a third index related to intermolecular interactions other than the non-polar interactions and the dipole-dipole interactions; and (4) a fourth index that is the dot product of the first index, the second index, and the third index.

[0087] The information processing system according to aspect 3 of the present invention may, in aspect 1 or 2 above, cause the output device to output at least one of the estimated at least one type of index and a graph using the index as coordinate values.

[0088] An information processing system according to aspect 4 of the present invention may be such that, in any of aspects 1 to 3 above, the odor sensor has a plurality of chemiresistor-type sensor elements, and the detection signal and the target detection signal are obtained from each of the plurality of chemiresistor-type sensor elements.

[0089] An information processing method according to aspect 5 of the present invention is an information processing method executed by an information processing system, and includes an estimation step of estimating at least one type of indicator corresponding to a target detection signal obtained from an odor sensor supplied with a target gas, using an estimation model that has learned a correspondence between the detection signal obtained from the odor sensor when a gas containing an odorous substance is supplied to the odor sensor and at least one type of indicator related to the intermolecular interaction of the odorous substance, and an output control step of outputting the estimation result to an output device.

[0090] The control program according to aspect 6 of the present invention may be a control program for causing a computer to function as the information processing system of aspects 1 to 4, and may also be a control program for causing a computer to function as the estimation unit and the output control unit.

[0091] A recording medium according to the seventh aspect of the present invention may be a computer-readable recording medium on which the control program according to the sixth aspect is recorded. [Explanation of symbols]

[0092] 2. Odor estimation device (information processing device, information processing system) 3 Output Devices 4. Training data 10 Odor Sensor 11 Odor sensor element 41 Explanatory variables 42 Response variable 100 Odor detection system (information processing system) 114 Detection data (detection signal, target detection signal) 212 Odor estimation unit (estimation unit) 214 Output control section 221 Estimation Model 421 Dispersion power (1st index) 422 Polar interaction (second index) 423 Hydrogen Bonds (Third Index) 424 Inner Product (4th Index)

Claims

1. an estimation unit that, when a gas containing an odorant is supplied to the odor sensor, estimates, from a target detection signal acquired from the odor sensor to which a target gas is supplied, at least one type of index corresponding to the target detection signal using an estimation model that has learned a correspondence relationship between the detection signal acquired from the odor sensor and at least one type of index related to the intermolecular interaction of the odorant; an output control unit that causes an output device to output the estimation result of the estimation unit.

2. The estimation model is The odorant is generated by machine learning using training data that includes, as explanatory variables, the detection signal acquired from the odor sensor to which the gas is supplied, and includes, as objective variables, at least one of (1) a first index related to non-polar intermolecular interactions of the odorant, (2) a second index related to dipole-dipole intermolecular interactions, (3) a third index related to intermolecular interactions different from the non-polar interactions and the dipole-dipole interactions, and (4) a fourth index that is the inner product of the first index, the second index, and the third index. The information processing system according to claim 1 .

3. outputting, to the output device, at least one of the estimated at least one type of index and a graph having the index as coordinate values; The information processing system according to claim 1 .

4. The odor sensor comprises: a plurality of chemiresistor-type sensor elements; The detection signal and the object detection signal are obtained from each of the plurality of chemiresistor-type sensor elements; The information processing system according to claim 1 .

5. An information processing method executed by an information processing system, an estimation step of estimating, from a target detection signal acquired from the odor sensor to which a target gas is supplied, at least one type of indicator corresponding to the target detection signal, using an estimation model that has learned a correspondence relationship between the detection signal acquired from the odor sensor and at least one type of indicator related to the intermolecular interaction of the odor substance when a gas containing the odor substance is supplied to the odor sensor; an output control step of causing an output device to output the result of the estimation.

6. A control program for causing a computer to function as the information processing system according to claim 1, the control program causing the computer to function as the estimation unit and the output control unit.

7. A computer-readable recording medium on which the control program according to claim 6 is recorded.

Images