Anomaly detection device, anomaly detection device, anomaly detection method, and anomaly detection program

Abstract

In order to enable highly accurate abnormality determination for determination targets, this abnormality determination device (1) comprises: an acquisition unit (11) that acquires measurement data for gas generated from a determination target, from a plurality of gas sensors having different reactivities depending on the composition of the gas; and a determination unit (12) that determines an abnormality related to the determination target by comparing reference data with determination data that is obtained from the measurement data of each of the plurality of gas sensors for the determination target, and obtained in accordance with a relationship between the measurement data.

Description

[Technical Field]

[0001] The present invention relates to an abnormality determination device, etc., for determining abnormalities in a target subject to determination. [Background technology]

[0002] There is a need for technology to determine whether a sample contains foreign substances. For example, Patent Document 1 below describes an odor identification device for the purpose of evaluating the quality of a sample that may contain various odor components. More specifically, the odor identification device described in Patent Document 1 uses the detection output of multiple odor sensors to determine whether an unknown sample, which is the product to be evaluated, is defective. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2003-232759 [Overview of the project] [Problems that the invention aims to solve]

[0004] The technology described in Patent Document 1 had the problem of not being able to perform highly accurate judgments.

[0005] One aspect of the present invention has been made in view of these problems, and one example of its objective is to provide an abnormality detection device, etc., capable of highly accurate abnormality detection of a target for determination. [Means for solving the problem]

[0006] An abnormality determination device according to one aspect of the present invention includes an acquisition means for acquiring measurement data for a gas generated from a target for determination from a plurality of gas sensors that have different reactivity from one another according to the composition of the gas, and a determination means for performing an abnormality determination regarding the target for determination by comparing determination data obtained from each of the measurement data of the plurality of gas sensors for the target for determination, the determination data obtained according to the relationship between the measurement data, with reference data.

[0007] An abnormality determination method according to one aspect of the present invention includes: acquiring measurement data for a gas generated from a target for determination from a plurality of gas sensors that have different reactivity from one another according to the composition of the gas; and performing an abnormality determination regarding the target for determination by comparing determination data obtained from each of the measurement data of the plurality of gas sensors for the target for determination, which is determined according to the relationship between the measurement data, with reference data.

[0008] An abnormality determination program according to one aspect of the present invention causes a computer to function as an acquisition means for acquiring measurement data for a gas generated from a target for determination from a plurality of gas sensors that have different reactivity from one another according to the composition of the gas, and as a determination means for performing an abnormality determination regarding the target for determination by comparing determination data obtained from each of the measurement data of the plurality of gas sensors for the target for determination, according to the relationship between the measurement data, with reference data. [Effects of the Invention]

[0009] According to one aspect of the present invention, it is possible to perform highly accurate abnormality detection regarding the object to be judged. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing the configuration of an abnormality detection device according to exemplary embodiment 1 of the present invention. [Figure 2] This is a flowchart showing an example of the processing flow of the abnormality detection method according to exemplary embodiment 1 of the present invention. [Figure 3] It is a block diagram showing the configuration of the abnormality determination device according to the second exemplary embodiment of the present invention. [Figure 4] It is a figure showing an example of a measurement waveform of gas by the gas sensor according to the second exemplary embodiment of the present invention. [Figure 5] It is a figure showing an example of reaction waveforms for a plurality of different measurement targets in a plurality of gas sensors according to the second exemplary embodiment of the present invention. [Figure 6] It is a figure showing an example of the correspondence between the amplitude value of measurement data measured by the gas sensor according to the second exemplary embodiment of the present invention and the concentration of gas generated from the measurement target. [Figure 7] It is a figure showing an example of the position of each sample in the set of samples according to the second exemplary embodiment of the present invention. [Figure 8] It is a flowchart showing an example of the flow of processing of the abnormality detection method according to the second exemplary embodiment of the present invention. [Figure 9] It is a flowchart showing an example of the flow of processing of the threshold determination method executed by the threshold determination unit according to the second exemplary embodiment of the present invention. [Figure 10] It is a figure showing an example of the configuration of a computer that realizes functions as each device according to each exemplary embodiment of the present invention.

MODE FOR CARRYING OUT THE INVENTION

[0011] 〔Exemplary Embodiment 1〕 The first exemplary embodiment of the present invention will be described in detail with reference to the drawings. This exemplary embodiment is a basic form for the exemplary embodiments described later.

[0012] (Configuration of Abnormality Determination Device) The configuration of the abnormality determination device 1 according to this exemplary embodiment will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the abnormality determination device 1. As shown in FIG. 1, the abnormality determination device 1 includes an acquisition unit (acquisition means) 11 and a determination unit (determination means) 12.

[0013] The acquisition unit 11 acquires measurement data for the gas generated from the determination target from a plurality of gas sensors having different reactivities depending on the gas composition.

[0014] Here, the gas sensors having different reactivities are, for example, gas sensors that can obtain different patterns of measurement waveforms depending on the gas composition of the measurement target including the determination target.

[0015] Also, as an example, the measurement data is time-series data, but this does not limit the exemplary embodiment.

[0016] The determination unit 12 performs an abnormality determination regarding the determination target by comparing the determination data and the reference data. The determination data is determination data obtained from the measurement data of each of the plurality of gas sensors for the determination target, and is obtained according to the relationship between the measurement data.

[0017] Here, the determination data obtained according to the relationship between the measurement data is, for example, data obtained by normalizing each measurement data measured by a plurality of gas sensors using the measurement data for the determination target measured by a specific gas sensor among the plurality of gas sensors.

[0018] Also, the reference data may be, for example, data obtained from the measurement data of each of the plurality of gas sensors for a normal measurement target, and data obtained according to the relationship between the measurement data.

[0019] (Effect of the abnormality determination device) As described above, in the abnormality determination device 1 according to this exemplary embodiment, a configuration is adopted in which an abnormality determination regarding the determination target is performed using determination data obtained according to the relationship between the measurement data of a plurality of gas sensors. Therefore, according to the abnormality determination device 1 according to this exemplary embodiment, an effect that an abnormality determination with high accuracy regarding the determination target can be performed is obtained.

[0020] (Abnormality determination program) The functions of the abnormality determination device 1 described above can also be implemented by a program. The estimation program according to this exemplary embodiment functions as an acquisition means for acquiring measurement data for the gas generated from a target for determination from a plurality of gas sensors that have different reactivity from one another depending on the gas composition, and a determination means for performing abnormality determination regarding the target for determination by comparing the determination data obtained from each of the measurement data of the plurality of gas sensors for the target for determination, according to the relationship between the measurement data, with reference data. This abnormality determination program enables highly accurate abnormality determination regarding the target for determination.

[0021] (Abnormality detection procedure) The flow of the abnormality detection method according to this exemplary embodiment will be explained with reference to Figure 2. Figure 2 is a flowchart showing the flow of the abnormality detection method. Note that the entity executing each step in this abnormality detection method may be a processor provided in the abnormality detection device 1.

[0022] In S11, at least one processor acquires measurement data for the gas generated from the target for analysis from multiple gas sensors that have different reactivity to each other depending on the gas composition.

[0023] In S12, at least one processor performs an abnormality determination regarding the object to be judged by comparing the judgment data with the reference data. The judgment data is obtained from the measurement data of each of the multiple gas sensors for the object to be judged, and is obtained according to the relationships between the measurement data.

[0024] (Effectiveness of the abnormality detection method) As described above, the anomaly detection method according to this exemplary embodiment employs a configuration in which anomaly detection is performed on the target of detection using detection data obtained according to the relationships between the measurement data of multiple gas sensors. Therefore, the anomaly detection method according to this exemplary embodiment has the effect of being able to perform highly accurate anomaly detection on the target of detection.

[0025] [Exemplary Embodiment 2] A second exemplary embodiment of the present invention will be described in detail with reference to the drawings.

[0026] (Configuration of the anomaly detection device) Figure 3 is a block diagram showing the configuration of the anomaly detection device 2 according to this exemplary embodiment. The anomaly detection device 2 includes a sample supply unit 21, an air supply unit 22a, an air supply unit 22b, a heating unit (heating and combustion unit) 23, a temperature control unit 24, a recovery unit 25, a measurement unit 26, a control unit (anomaly determination device) 27, a storage unit 28, and a display unit 29.

[0027] <Sample Supply Unit 21> The sample supply unit 21 supplies the measurement sample to the heating unit 23. The sample supply unit 21 only needs to be configured to supply the measurement sample to the heating unit 23; its mechanism is not particularly limited. Here, the measurement sample includes the sample subject to abnormality determination (detection) by the abnormality detection device 2 (the object of determination), and samples that are normal. The sample supplied to the heating unit 23 can be any substance that produces gas through combustion by heating; for example, synthetic resins can be used.

[0028] Here, a "normal" sample is a sample that does not contain any foreign matter. For example, a normal sample may be a sample in which the purity of the material constituting the sample is higher than a predetermined value. In this specification, a normal sample is also referred to as a "learning target."

[0029] <Air supply unit 22a> The air supply unit 22a supplies air to the heating unit 23. The air supply unit 22a only needs to be configured to supply air to the heating unit 23; its mechanism is not particularly limited. The air supplied to the heating unit 23 by the air supply unit 22a is used for the combustion of the measurement sample.

[0030] <Heating section 23> The heating unit 23 generates gas by heating and burning the sample to be measured. For example, the heating unit 23 generates gas by heating and burning the object to be judged at a specific temperature. The heating unit 23 is controlled at this specific temperature by the temperature control unit 24. The heating unit 23 only needs to be configured to heat the sample to be measured, and its mechanism is not particularly limited.

[0031] <Temperature control unit 24> The temperature control unit 24 controls the heating temperature of the sample to be measured by the heating unit 23. For example, if the sample to be measured is a synthetic resin such as plastic, the temperature control unit 24 controls the heating unit 23 to heat the sample to at least 350°C. The temperature at which the heating unit 23 heats the sample to be measured may vary depending on the sample to be measured. The temperature control unit 24 controls the heating temperature of the sample to be measured by the heating unit 23 so that it reaches the temperature at which the sample burns.

[0032] <Recovery section 25> The recovery unit 25 recovers the gas generated from the measurement sample that has been burned by the heating unit 23. The recovery unit 25 supplies the recovered gas to the measurement unit 26.

[0033] In this exemplary embodiment, the recovery unit 25 is described as being configured to recover gases generated by the combustion of the measurement sample. As another example, the recovery unit 25 may be configured to recover gases generated by the measurement sample without combustion (for example, volatile substances, etc.).

[0034] <Measurement section 26> The measurement unit 26 is equipped with a plurality of gas sensors 261 that measure the gas supplied from the recovery unit 25. The plurality of gas sensors 261 have different reactivity to each other depending on the composition of the gas. The plurality of gas sensors 261 may be configured as, for example, a gas sensor array. In the example shown in Figure 3, the measurement unit 26 is equipped with gas sensor ch1, gas sensor ch2, gas sensor ch3, etc., as a plurality of gas sensors 261. Each gas sensor 261 outputs a signal indicating the measured value to the acquisition unit 271 of the control unit 27.

[0035] <Air supply unit 22b> The air supply unit 22b supplies air to the measuring unit 26. The air supply unit 22b only needs to be configured to supply air to the measuring unit 26; its mechanism is not particularly limited. The air supplied to the measuring unit 26 by the air supply unit 22b is used for gas measurement by the gas sensor 261.

[0036] (Gas measurement method) Here, an example of a method for measuring gas using multiple gas sensors 261 is described. Gas measurement by the gas sensors 261 is performed in two modes: sample gas supply mode and clean air supply mode. In sample gas supply mode, multiple gas sensors 261 measure the gas while gas is being supplied from the recovery unit 25 to the measurement unit 26. In clean air supply mode, multiple gas sensors 261 measure the gas while air is being supplied from the air supply unit 22b to the measurement unit 26. In sample gas supply mode, the gas supplied from the recovery unit 25 gradually fills the measurement unit 26. In clean air supply mode, the gas supplied from the recovery unit 25 that fills the measurement unit 26 is gradually exhausted from the measurement unit 26.

[0037] Figure 4 shows an example of a gas measurement waveform by the gas sensor 261 according to this exemplary embodiment. The vertical axis of Figure 4 shows the measurement value by the gas sensor 261 in "arbitrary unit (Arb. Unit)". The horizontal axis of Figure 4 shows "elapsed time (sec)". Figure 4 shows an example of a gas measurement waveform when the gas is measured in sample gas supply mode until elapsed time t1, and then the gas is measured in clean air supply mode from elapsed time t1 onwards.

[0038] In this specification, the "measured value" from the gas sensor 261 may also be referred to as the "measurement data" from the gas sensor 261. Furthermore, in this specification, the "measurement" from the gas sensor 261 may also be referred to as the "reaction" from the gas sensor 261. In addition, in this specification, the "measurement waveform" from the gas sensor 261 may also be referred to as the "reaction waveform" from the gas sensor 261.

[0039] Figure 5 shows an example of reaction waveforms for multiple different measurement targets (combustion gases of measurement targets A, B, and C) in multiple gas sensors 261 (gas sensors ch1, ch2, and ch3) with different reactivity depending on the gas composition according to this exemplary embodiment. As shown in Figure 5, the reaction waveforms in gas sensors ch1, ch2, and ch3 have different reaction waveform patterns depending on the composition of the measurement target.

[0040] <Control Unit 27> As shown in Figure 3, the control unit 27 includes an acquisition unit (acquisition means) 271, a specificity determination unit 272, a normalization unit (normalization means) 273, an abnormality determination unit (determination means) 274, a threshold determination unit (threshold determination means) 275, and a display control unit (display control means) 276.

[0041] <Acquisition part 271> The acquisition unit 271 acquires measurement data for the gas generated from the target for determination from a plurality of gas sensors 261, each of which has a different reactivity depending on the gas composition.

[0042] In detail, the acquisition unit 271 acquires signals indicating the measured values ​​of each gas sensor 261 for the gas generated from the object to be judged. The acquisition unit 271 uses the acquired measured values ​​to update the object to be judged measurement data (measurement data) 282 stored in the storage unit 28. The object to be judged measurement data 282 is data indicating the measured values ​​of the gas generated from the object to be judged measured by each gas sensor 261 (gas sensor ch1, gas sensor ch2, and gas sensor ch3, etc.). For example, the object to be judged measurement data 282 may be data indicating the reaction waveforms of each gas sensor 261 for the object to be judged, as shown in Figures 4 and 5. In addition, the object to be judged measurement data 282 may include data indicating the measured values ​​of multiple objects to be judged that each gas sensor 261 has measured so far.

[0043] Furthermore, the acquisition unit 271 may acquire signals indicating the measured values ​​of each gas sensor 261 for the gas generated from the learning target. The acquisition unit 271 uses the acquired measured values ​​to update the learning target measurement data (measurement data) 281 stored in the storage unit 28. The learning target measurement data 281 is data indicating the measured values ​​of the gas generated from the learning target measured by each gas sensor 261 (gas sensor ch1, gas sensor ch2, and gas sensor ch3, etc.). For example, the judgment target measurement data 282 may be data indicating the reaction waveform of each gas sensor 261 for the learning target, as shown in Figures 4 and 5. Note that the learning target measurement data 281 only needs to be the measured values ​​of each gas sensor 261 for the gas generated from a normal sample. For example, measurement data measured by an external device equipped with gas sensors similar to the gas sensors 261 provided by the measurement unit 26 may be applied.

[0044] When the acquisition unit 271 updates the measurement data 282 to be judged or the measurement data 281 to be learned stored in the storage unit 28, it outputs a signal indicating the update to the specificity determination unit 272.

[0045] <Specificity determination unit 272> When the specificity determination unit 272 receives a signal from the acquisition unit 271 indicating an update to the measurement data 282 to be determined or the measurement data 281 to be learned, it determines the specificity of the gas sensor 261. Specifically, the specificity determination unit 272 determines whether the specificity of each gas sensor 261 (gas sensor ch1, ch2, ch3, etc.) is lower than a predetermined standard for multiple measurement targets measured so far.

[0046] Here, "low specificity" can also be expressed as having low selectivity for multiple previously measured targets. Conversely, "low specificity" can also be expressed as having high reactivity for multiple previously measured targets. For example, a gas sensor with low specificity is one that reacts to any of the previously measured targets.

[0047] (Method for determining the specificity of a gas sensor) An example of a method for determining the specificity of each gas sensor 261 by the specificity determination unit 272 will be described below.

[0048] In this example, the specificity determination unit 272 determines that a gas sensor 261 has low specificity if the amplitude values ​​of the measurement data of all gases (targets of measurement) measured by that gas sensor 261 so far are greater than a predetermined value.

[0049] For example, if the amplitude values ​​of the measurement data for all measurement targets measured by gas sensor ch1 so far are greater than 10 times the standard deviation of the amplitude values ​​of the measurement data of gas sensor ch1 in the blank state, the specificity of gas sensor ch1 is determined to be lower than a predetermined standard. Here, "blank" refers to a state in which no measurement target is present, for example, a state in which only air is present. The specificity determination unit 272 may use measurement data obtained during noise measurement as the measurement data in the blank state. Also, the value of "10 times" can be changed as appropriate.

[0050] The specificity determination unit 272 may use at least one of the data shown in the learning target measurement data 281 and the data shown in the determination target measurement data 282 as the "measurement data of the gas (target of measurement) that each gas sensor 261 has measured so far." Alternatively, the specificity determination unit 272 may use all the data shown in the learning target measurement data 281 and the determination target measurement data 282 as the "measurement data of the gas (target of measurement) that each gas sensor 261 has measured so far." The specificity determination unit 272 may make a determination about the specificity of each gas sensor 261 each time the learning target measurement data 281 and the determination target measurement data 282 are updated.

[0051] The specificity determination unit 272 may determine that the specificity of a gas sensor 261 that satisfies the following equation is lower than a predetermined standard.

[0052]

number

number

[0053] Also,

number

[0054] Also, gas g i In contrast, gas sensor k performs multiple measurements, and j is gas g i The measurement number is shown to identify a specific measurement from the multiple measurements mentioned above.

[0055] Also, gas g i In contrast, when the gas sensor k performs m measurements:

number

number

[0056] In other words, the specificity determination unit 272 determines that a gas sensor k whose specificity is lower than a predetermined standard is determined to satisfy the following equation.

[0057]

number

[0058] Here, the correspondence between the amplitude values ​​of the measurement data measured by the gas sensor 261 and the concentration of the gas produced from the measurement target will be explained using Figure 6. Figure 6 is a diagram showing an example of the correspondence between the amplitude values ​​of the measurement data measured by the gas sensor 261 and the concentration of the gas produced from the measurement target. The vertical axis of the graph shown in Figure 6 represents the amplitude values ​​of the measurement data measured by the gas sensor 261. The horizontal axis of the graph shown in Figure 6 represents the concentration of the gas produced from the measurement target. As shown in Figure 6, in the region where the amplitude value is smaller than W1 and in the region where it is larger than W2, the relationship between the amplitude value and the gas concentration is nonlinear. Also, in the region where the amplitude value is between W1 and W2, the relationship between the amplitude value and the gas concentration is linear.

[0059] In the example shown in Figure 6, the specificity determination unit 272 selects gas sensors 261 whose amplitude values ​​of measurement data for all measurement targets measured so far are between W1 and W2.

[0060] The specificity determination unit 272 updates the specificity data 283 stored in the storage unit 28 according to the determination result and selection. The specificity data 283 is data indicating a gas sensor 261 with a specificity lower than a predetermined standard.

[0061] The specificity determination unit 272 may combine the above-described "gas sensor specificity determination method" and "gas sensor selection" to select only one gas sensor 261 that is determined to have lower specificity than a predetermined standard.

[0062] Even after combining the above-described "method for determining the specificity of gas sensors" and "selection of gas sensors," if there are multiple gas sensors 261 that are determined to have lower specificity than a predetermined standard, the specificity determination unit 272 may perform the following additional selection process on the multiple gas sensors 261.

[0063] For example, the specificity determination unit 272 is as described in the "Method for Determining the Specificity of a Gas Sensor" above.

number

[0064] Furthermore, the specificity determination unit 272 may calculate the correlation coefficients between each gas sensor defined by the following formula and select the sensor whose sum of correlation coefficients is the largest.

[0065]

number

[0066] When the specificity determination unit 272 has completed the determination of the specificity of the gas sensor 261, it outputs a signal to the standardization unit 273 indicating the completion of the determination.

[0067] <Standardization Section 273> When the standardization unit 273 receives a signal from the specificity determination unit 272 indicating the completion of the determination, it generates determination data 285 and reference data 284.

[0068] The standardization unit 273 normalizes the judgment target measurement data 282 measured by the multiple gas sensors 261 to generate judgment data 285. The standardization unit 273 performs the above normalization on multiple measurement targets measured by the multiple gas sensors 261 so far, using the amplitude values ​​of the measurement data for the judgment target measured by the gas sensor among the multiple gas sensors 261 that has a lower specificity than a predetermined standard.

[0069] For different objects of measurement composed of the same composition, the composition of the gas produced from each object will be the same. However, if the concentrations of the gases produced from each object differ, the gas sensor 261 will measure different values ​​depending on the concentration of the gas produced by each object. To suppress the influence of the gas concentration from each object on the measurement, the normalization unit 273 normalizes the measured values ​​measured by the gas sensor 261 and generates judgment data 285.

[0070] The standardization performed by the standardization unit 273 will be explained using Figure 5. The standardization unit 273 refers to the specificity data 283 stored in the storage unit 28 and identifies gas sensors with specificity lower than a predetermined standard. In the example shown in Figure 5, gas sensor ch2 is a gas sensor with specificity lower than a predetermined standard. The standardization unit 273 performs the standardization by dividing the measured values ​​measured by gas sensors 261 (gas sensor ch1, gas sensor ch2, gas sensor ch3) by the amplitude value (maximum amplitude) of the measured value of gas sensor ch2.

[0071] For example, as shown in Figure 5, the amplitude value of the measurement taken by gas sensor ch2 for the combustion gas of the object A1 is S1. The normalization unit 273 performs this normalization by dividing the measurement values ​​taken by gas sensors ch1, ch2, and ch3 for the combustion gas of the object A1 by S1.

[0072] Furthermore, S2 is the amplitude value of the measurement value of gas sensor ch2 for the combustion gas of the object A2 being measured. The normalization unit 273 performs the normalization by dividing the measurement values ​​measured by gas sensors ch1, ch2, and ch3 for the combustion gas of the object A2 being measured by S2.

[0073] Furthermore, S3 is the amplitude value of the measurement value of gas sensor ch2 for the combustion gas of the object A3 being measured. The normalization unit 273 performs the normalization by dividing the measurement values ​​measured by gas sensors ch1, ch2, and ch3 for the combustion gas of the object A2 being measured by S3.

[0074] The standardization unit 273 stores the judgment data 285 generated by the standardization process in the storage unit 28.

[0075] Furthermore, the standardization unit 273 normalizes the learning target measurement data 281, which is the measurement data of the learning target measured by the multiple gas sensors 261, and generates reference data 284. The standardization unit 273 performs the above normalization using the amplitude values ​​of the measurement data for the learning target measured by the gas sensor among the multiple gas sensors 261 that has a lower specificity than a predetermined standard, for the multiple measurement targets that the multiple gas sensors 261 have measured so far.

[0076] Here, the reference data refers to data obtained based on the relationships between measurement data from multiple gas sensors for a gas produced from a normal training target.

[0077] The generation of reference data 284 by the standardization unit 273 is the same as the generation of determination data 285 described above, so it will not be explained again here. The standardization unit 273 stores the generated reference data 284 in the storage unit 28.

[0078] When the standardization unit 273 generates the judgment data 285 and the reference data 284, it outputs a signal indicating the generation to the abnormality determination unit 274.

[0079] <Abnormality determination unit 274> When the abnormality determination unit 274 receives a signal from the standardization unit 273 indicating the generation of determination data 285 and reference data 284, it performs an abnormality determination regarding the target for determination.

[0080] In detail, the abnormality determination unit 274 performs an abnormality determination regarding the target for determination by comparing the determination data, which is obtained from the measurement data of each of the multiple gas sensors 261 for the target for determination and is obtained according to the relationship between the measurement data, with reference data.

[0081] (Method for detecting anomalies: Local outlier factor method) An example of an abnormality determination method by the abnormality determination unit 274 will be described. In this example, the abnormality determination unit 274 performs abnormality determination regarding the target for determination using the local outlier factor method.

[0082] In this example, abnormality detection uses one-dimensional concatenated data generated from the measured values ​​(reaction waveforms) of all gas sensors (gas sensor ch1, gas sensor ch2, and gas sensor ch3) provided by the measurement unit 26, as shown in Figure 5. This one-dimensional concatenated data is concatenated data obtained by sequentially concatenating the measured values ​​of each gas sensor (gas sensor ch1, gas sensor ch2, and gas sensor ch3) that have been normalized by the normalization unit 273.

[0083] As an example, the above one-dimensional concatenated data can be obtained by concatenating the reaction waveforms of gas sensor ch1, gas sensor ch2, and gas sensor ch3 in series in that order.

[0084] In this description of the anomaly detection method, one one-dimensional data point is referred to as a sample. Furthermore, in this description of the anomaly detection method, Sample A is a sample based on the detection data 285, and anomaly detection will be performed on Sample A. That is, Sample A can be referred to as the detection data. Additionally, samples other than Sample A are samples based on the reference data 284, and samples other than Sample A can be referred to as reference data.

[0085] Let k_dis(A) be the distance to the k-th nearest neighbor of sample A, and the set of k nearest neighbor samples is N. k Let (A) be the distance between sample A and sample B. The anomaly detection unit 274 calculates the reachability distance and local reachability density as shown below.

[0086] Reachability distance:

[0087]

number

[0088]

number

[0089]

number

[0090]

number

[0091] Also, the abnormality determination unit 274 may perform an abnormality determination regarding the determination target using the local outlier factor method with the threshold value described later determined by the threshold value determination unit 275. For example, the calculated LOF K When the value of (A) is larger than the threshold value determined by the threshold value determination unit 275, the abnormality determination unit 274 determines that there is an abnormality in the sample A.

[0092] In the above-described abnormality determination method, the distance between samples is defined for each time slice of the measurement data, and it can also be expressed that the LOF is defined using the integration of the distances.

[0093] When the abnormality determination unit 274 determines that there is an abnormality in the determination target, it outputs a signal indicating that there is an abnormality in the determination target to the display control unit.

[0094] 〈Threshold Value Determination Unit 275〉 The threshold value determination unit 275 determines a threshold value used for the abnormality determination regarding the determination target using the local outlier factor method by the abnormality determination unit 274 from the distribution of the reference data. For example, the threshold value determination unit 275 may determine an appropriate threshold value from the LOF value distribution of the reference data 284. The threshold value determination unit 275 stores threshold value data 286 indicating the set threshold value in the storage unit 28.

[0095] (Abnormality Determination Method: Method Using Principal Component Analysis and Neural Network) The abnormality determination method by the abnormality determination unit 274 is not limited to the above-described example. As an example, the abnormality determination unit 274 may be configured to perform an abnormality determination using principal component analysis (PCA) using the determination data 285 generated by the normalization unit 273 through normalization or a neural network using the determination data 285 as input data.

[0096] <Display Control Unit 276> If the abnormality determination unit 274 determines that there is an abnormality in the object to be determined, the display control unit 276 causes the display unit 29 to display an indication that there is an abnormality in the object to be determined.

[0097] <Storage section 28> The memory unit 28 stores the learning target measurement data 281, the judgment target measurement data 282, the specificity data 283, the reference data 284, the judgment data 285, and the threshold data 286.

[0098] <Display section 29> The display unit 29 displays an indication that there is an abnormality in the object to be judged. In this exemplary embodiment, the configuration in which the abnormality detection device 2 displays on the display unit 29 that there is an abnormality in the object to be judged has been described, but for example, the abnormality detection device 2 may also be configured to output an audio message from an audio output unit or the like that indicates that there is an abnormality in the object to be judged. The configuration in which the abnormality detection device 2 notifies the user that there is an abnormality in the object to be judged is not particularly limited.

[0099] (Flowchart of the anomaly detection method using anomaly detection device 2) The processing flow of the anomaly detection method executed by the anomaly detection device 2 according to this exemplary embodiment will be explained with reference to Figure 8. Figure 8 is a flowchart showing an example of the processing flow of the anomaly detection method executed by the anomaly detection device 2.

[0100] In S21, the acquisition unit 271 acquires measurement data for the gas generated from the target for determination from multiple gas sensors 261 (gas sensor ch1, gas sensor ch2, gas sensor ch3, etc.) which have different reactivity with respect to the gas composition. The acquisition unit 271 updates the target for determination measurement data 282 stored in the storage unit 28.

[0101] In S22, the specificity determination unit 272 determines the specificity of the gas sensors 261 (gas sensor ch1, gas sensor ch2, gas sensor ch3, etc.). Specifically, the specificity determination unit 272 determines whether the specificity of each gas sensor 261 (gas sensor ch1, gas sensor ch2, gas sensor ch3, etc.) is lower than a predetermined standard for multiple measurement targets measured so far.

[0102] Following S22, in S23, the normalization unit 273 normalizes the measurement data 282 to be judged and generates judgment data 285. The normalization unit 273 also normalizes the learning target measurement data 281, which is the measurement data to be learned, and generates reference data 284. In detail, the normalization unit 273 performs the above normalization using the amplitude values ​​of the measurement data measured by the gas sensor among the gas sensors 261 that has a lower specificity than a predetermined standard for the multiple measurement targets that the multiple gas sensors 261 have measured so far.

[0103] Following S23, in S24, the abnormality determination unit 274 performs an abnormality determination regarding the object to be determined by comparing the determination data 285 with the reference data 284. The determination data 285 is obtained from the measurement data of each of the multiple gas sensors 261 (gas sensor ch1, gas sensor ch2, gas sensor ch3, etc.) for the object to be determined. Furthermore, the determination data 285 is obtained according to the relationships between the measurement data. Note that "obtained according to the relationships between the measurement data" means, for example, in S23, the normalization unit 273 normalizes the measurement data 282 of the object to be determined to generate the determination data 285. If the abnormality determination unit 274 determines that there is an abnormality in the object to be determined (YES in S24), the process continues to S25. If the abnormality determination unit 274 determines that there is no abnormality in the object to be determined (NO in S24), the process ends.

[0104] In S25, the display control unit 276 causes the display unit 29 to display an indication that there is an abnormality in the object to be judged. Then, the process ends. Alternatively, if the abnormality determination unit 274 determines that there is no abnormality in the object to be judged (NO in S24), it may be configured to output a signal to the display control unit 276 indicating that there is no abnormality in the object to be judged. In this configuration, the display control unit 276 may display an indication on the display unit 29 that there is no abnormality in the object to be judged.

[0105] (Flowchart of the threshold determination method by the anomaly detection device 2) The processing flow of the threshold determination method executed by the anomaly detection device 2 according to this exemplary embodiment will be explained with reference to Figure 9. Figure 9 is a flowchart showing an example of the processing flow of the threshold determination method executed by the threshold determination unit 275 of the anomaly detection device 2.

[0106] In S31, the threshold determination unit 275 determines a threshold from the distribution of normalized measurement data of the learning target shown in the reference data 284. For example, the threshold determination unit 275 may determine the threshold after the normalization unit 273 generates the reference data 284 (normalization of the measurement data of the learning target). Alternatively, after the threshold determination unit 275 has determined the threshold, the anomaly determination unit 274 may use the threshold to perform an anomaly determination regarding the target for determination.

[0107] (Effectiveness of anomaly detection devices) As described above, in the anomaly detection device 2 according to this exemplary embodiment, a normalization unit 273 is employed to normalize the measurement data measured by the multiple gas sensors 261 and generate preliminary determination data by using the amplitude value of the measurement data for the determination target measured by the gas sensor among the multiple gas sensors 261 that has a lower specificity than a predetermined standard for the multiple measurement targets measured so far by the multiple gas sensors 261. For this reason, the anomaly detection device 2 according to this exemplary embodiment has the effect of suppressing the influence of the gas concentration generated from the determination target on the anomaly determination, and enabling highly accurate anomaly determination, in addition to the effects of the anomaly determination device 1 according to exemplary embodiment 1.

[0108] Furthermore, in the anomaly detection device 2 according to this exemplary embodiment, the reference data 284 is data obtained according to the relationship between the measurement data of multiple gas sensors 261 for gases generated from a normal learning target. A normalization unit 273 is employed to generate the reference data 284 by normalizing the measurement data of the learning target measured by multiple gas sensors 261 using the amplitude value of the measurement data for the learning target measured by the gas sensors among the multiple gas sensors 261 that have a lower specificity than a predetermined standard for multiple measurement targets measured by the multiple gas sensors 261 so far. Therefore, the anomaly detection device 2 according to this exemplary embodiment has the effect of suppressing the influence of the gas concentration generated from the learning target on the anomaly determination, and enabling highly accurate anomaly determination, in addition to the effects of the anomaly determination device 1 according to exemplary embodiment 1.

[0109] Furthermore, in the anomaly detection device 2 according to this exemplary embodiment, an anomaly determination unit 274 is employed that performs an anomaly determination regarding the target for determination using the local outlier factor method. Therefore, in addition to the effects of the anomaly determination device 1 according to the exemplary embodiment 1, the anomaly detection device 2 according to this exemplary embodiment provides the effect of being able to perform highly accurate anomaly determination using the local outlier factor method.

[0110] Furthermore, in the anomaly detection device 2 according to this exemplary embodiment, the reference data 284 is data obtained according to the relationship between the measurement data of multiple gas sensors 261 for gas generated from a normal learning target, and a threshold determination unit 275 is employed that determines a threshold from the distribution of the reference data 284. Therefore, in addition to the effects of the anomaly determination device 1 according to the exemplary embodiment 1, the anomaly detection device 2 according to this exemplary embodiment can be made to perform highly accurate anomaly determination using a local outlier factor method that applies a threshold determined from the distribution of reference data.

[0111] Furthermore, the abnormality detection device 2 according to this exemplary embodiment employs a configuration comprising a heating unit 23 that generates gas by heating and burning the object to be determined, a plurality of gas sensors 261 that measure the gas and have different reactivity with respect to each other depending on the gas composition, a display unit 29, and a control unit 277 that further includes a display control unit 276 that displays an indication on the display unit 29 that there is an abnormality in the object to be determined when an abnormality is determined to exist in the object to be determined. Therefore, the abnormality detection device 2 according to this exemplary embodiment can obtain the same effects as the abnormality determination device 1 according to exemplary embodiment 1.

[0112] [Examples of implementation using software] Some or all of the functions of the anomaly determination device 1 and the anomaly detection device 2 may be implemented by hardware such as integrated circuits (IC chips) or by software.

[0113] In the latter case, the anomaly determination device 1 and the anomaly detection device 2 are implemented, for example, by a computer that executes instructions for a program, which is software that implements each function. An example of such a computer (hereinafter referred to as computer C) is shown in Figure 10. Computer C comprises at least one processor C1 and at least one memory C2. The memory C2 stores a program P that causes computer C to operate as an anomaly determination device 1 and anomaly detection device 2. In computer C, the processor C1 reads program P from memory C2 and executes it, thereby implementing the functions of the anomaly determination device 1 and the anomaly detection device 2.

[0114] Processor C1 can include, for example, a CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), microcontroller, or a combination thereof. Memory C2 can include, for example, flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof.

[0115] Computer C may also be equipped with RAM (Random Access Memory) for loading program P at runtime and for temporarily storing various data. Furthermore, computer C may be equipped with communication interfaces for sending and receiving data with other devices. Additionally, computer C may be equipped with input / output interfaces for connecting input / output devices such as keyboards, mice, displays, and printers.

[0116] Furthermore, program P can be recorded on a non-temporary, tangible recording medium M that is readable by computer C. Such a recording medium M could be, for example, tape, disk, card, semiconductor memory, or programmable logic circuitry. Computer C can acquire program P via such a recording medium M. Program P can also be transmitted via a transmission medium. Such a transmission medium could be, for example, a communication network or broadcast waves. Computer C can also acquire program P via such a transmission medium.

[0117] [Additional Note 1] The present invention is not limited to the embodiments described above, and various modifications are possible within the scope of the claims. For example, embodiments obtained by appropriately combining the technical means disclosed in the embodiments described above are also included in the technical scope of the present invention.

[0118] [Additional Note 2] Some or all of the embodiments described above may also be described as follows. However, the present invention is not limited to the embodiments described below.

[0119] (Note 1) An acquisition means for acquiring measurement data for the gas generated from a target for determination from multiple gas sensors that have different reactivity to each other depending on the gas composition, The system includes determination means for performing an abnormality determination regarding the target to be determined by comparing determination data obtained from the measurement data of each of the plurality of gas sensors for the target to be determined, the determination data obtained according to the relationship between the measurement data, and reference data. An abnormality detection device characterized by the following features.

[0120] According to the above configuration, since the abnormality detection uses judgment data obtained according to the relationships between the measurement data, highly accurate abnormality detection can be performed.

[0121] (Note 2) The aforementioned abnormality detection device further includes a standardization means, The standardization means uses the amplitude values ​​of the measurement data for the determination target measured by the gas sensor among the multiple gas sensors that has a lower specificity than a predetermined standard for the multiple measurement targets measured by the multiple gas sensors so far, to standardize the measurement data measured by the multiple gas sensors and generate preliminary determination data. An abnormality detection device as described in Appendix 1, characterized by the features described herein.

[0122] With the above configuration, the judgment data is standardized, so it is possible to perform highly accurate anomaly detection by suppressing the influence of the gas concentration generated from the object being judged on the anomaly detection.

[0123] (Note 3) The aforementioned abnormality detection device further includes a standardization means, The aforementioned reference data is data obtained according to the relationship between the measurement data of the multiple gas sensors for the gas generated from a normal learning target, The standardization means uses the amplitude values ​​of the measurement data for the learning target measured by the gas sensors, which have a lower specificity than a predetermined standard among the gas sensors, to standardize the measurement data of the learning target measured by the multiple gas sensors and generate reference data for the first half of the measurement period. An abnormality detection device as described in Appendix 1 or 2, characterized by the above.

[0124] With the above configuration, the reference data is standardized, which has the effect of suppressing the influence of gas concentration from the training target on anomaly detection, resulting in highly accurate anomaly detection.

[0125] (Note 4) The determination means performs an abnormality determination regarding the target for determination using the local outlier factor method. An abnormality detection device as described in any one of the appendices 1 to 3, characterized by the above.

[0126] With the above configuration, highly accurate anomaly detection can be performed using the local outlier factor method.

[0127] (Note 5) The aforementioned reference data is data obtained according to the relationship between the measurement data of the multiple gas sensors for the gas generated from a normal learning target, The anomaly detection device further comprises threshold determination means for determining a threshold from the distribution of the reference data, The determination means performs an abnormality determination regarding the target for determination using the local outlier factor method with the threshold value. An abnormality detection device as described in Appendix 4, characterized by the features described herein.

[0128] The above configuration allows for highly accurate anomaly detection using a local outlier factorization method that applies a threshold determined from the distribution of reference data.

[0129] (Note 6) A heating and combustion unit that generates gas by heating and burning the object to be judged, A plurality of gas sensors, each having different reactivity depending on the gas composition, for measuring the aforementioned gas, Display unit and An abnormality detection device, characterized by comprising: an abnormality determination device according to any one of appendices 1 to 5, further comprising a display control means for causing the display unit to display an indication that there is an abnormality in the object to be determined when it is determined that there is an abnormality in the object to be determined.

[0130] With the above configuration, the same effects as those achieved by the anomaly detection device described above can be obtained.

[0131] (Note 7) Obtaining measurement data for the gas generated from the target for evaluation from multiple gas sensors with different reactivity depending on the gas composition, An abnormality determination regarding the target for determination is performed by comparing the determination data obtained from the measurement data of each of the plurality of gas sensors for the target for determination, the determination data obtained according to the relationship between the measurement data, and reference data. An anomaly detection method, including the above.

[0132] The method described above produces the same effect as the anomaly detection device mentioned above.

[0133] (Note 8) Computers, An acquisition means for acquiring measurement data for the gas generated from a target for determination from multiple gas sensors that have different reactivity to each other depending on the gas composition, and Determination means for performing an abnormality determination regarding the target to be determined by comparing determination data obtained from the measurement data of each of the plurality of gas sensors for the target to be determined, the determination data obtained according to the relationship between the measurement data, and reference data. An anomaly detection program that functions as such.

[0134] The above configuration achieves the same effect as the anomaly detection device described above.

[0135] [Additional Note 3] Some or all of the embodiments described above can also be expressed as follows:

[0136] An abnormality determination device comprising at least one processor, the processor performing an acquisition process to acquire measurement data for a gas generated from a target for determination from a plurality of gas sensors that differ in reactivity to each other according to the gas composition, and a determination process to perform an abnormality determination regarding the target for determination by comparing determination data obtained from each of the measurement data of the plurality of gas sensors for the target for determination, the determination data obtained according to the relationship between the measurement data, with reference data.

[0137] Furthermore, this abnormality detection device may also be equipped with memory, and this memory may store a program that causes the processor to execute the acquisition process and the determination process. This program may also be recorded on a computer-readable, non-temporary, tangible recording medium. [Explanation of symbols]

[0138] 1...Abnormality determination device 11,271... Acquisition unit (acquisition means) 12...Judgment unit (judgment means) 23...Heating section (heating and combustion section) 261,k ···Gas sensor 27. Control Unit (Anomaly Detection Device) 273 ···Standardization department (standardization means) 274 ... Abnormality judgment unit (judgment means) 275 ···Threshold determination unit (threshold determination means) 276 ···Display control unit (display control means) 281 ···Training target measurement data (measurement data) 282 ···Measurement data subject to judgment (measurement data) 285 ···Data for judgment 284 ···Reference data 29...Display section

Claims

1. An acquisition means for acquiring measurement data for the gas generated from a target for determination from multiple gas sensors that have different reactivity to each other depending on the gas composition, A determination means for performing an abnormality determination regarding the target to be determined by comparing determination data obtained from the measurement data of each of the plurality of gas sensors for the target to be determined, the determination data obtained according to the relationship between the measurement data, and reference data. A standardization means for generating determination data by normalizing the measurement data measured by the plurality of gas sensors using the amplitude value of the measurement data for the determination target measured by the gas sensor among the plurality of gas sensors that has a lower specificity than a predetermined standard for the plurality of measurement targets measured by the plurality of gas sensors so far, It is equipped with An abnormality detection device characterized by the following features.

2. Acquisition means for acquiring measurement data for the gas generated from a target for determination from a plurality of gas sensors whose reactivity differs from that of the gas depending on the gas composition, A determination means for performing an abnormality determination regarding the target to be determined by comparing determination data obtained from the measurement data of each of the plurality of gas sensors for the target to be determined, the determination data obtained according to the relationship between the measurement data, and reference data. Equipped with, The aforementioned reference data is data obtained according to the relationship between the measurement data of the multiple gas sensors for the gas generated from a normal learning target, The system further includes a standardization means for generating reference data by normalizing the measurement data of the learning target measured by the multiple gas sensors using the amplitude values ​​of the measurement data of the learning target measured by the gas sensors among the multiple gas sensors that have a lower specificity than a predetermined standard for multiple measurement targets measured by the multiple gas sensors to date. An abnormality detection device characterized by the following features.

3. The determination means performs an abnormality determination regarding the target for determination using the local outlier factor method. An abnormality detection device according to claim 1 or 2.

4. The aforementioned reference data is data obtained according to the relationship between the measurement data of the multiple gas sensors for the gas generated from a normal learning target, The anomaly detection device further comprises threshold determination means for determining a threshold from the distribution of the reference data, The determination means performs an abnormality determination regarding the target for determination using the local outlier factor method with the threshold value. The abnormality detection device according to feature 3.

5. A heating and combustion unit that generates gas by heating and burning the object to be judged, A plurality of gas sensors, each having different reactivity depending on the gas composition, for measuring the aforementioned gas, Display unit and An abnormality detection device comprising: an abnormality determination device according to any one of claims 1 to 4, further comprising a display control means for causing the display unit to display an indication that there is an abnormality in the object to be determined when it is determined that there is an abnormality in the object to be determined.

6. Obtaining measurement data for the gas generated from the target for evaluation from multiple gas sensors with different reactivity depending on the gas composition, An abnormality determination regarding the target for determination is performed by comparing the determination data obtained from the measurement data of each of the plurality of gas sensors for the target for determination, the determination data obtained according to the relationship between the measurement data, and reference data. The process involves using the amplitude values ​​of the measurement data for the target to be determined, measured by the gas sensor among the multiple gas sensors that has a lower specificity than a predetermined standard, to normalize the measurement data measured by the multiple gas sensors and generate the determination data. An anomaly detection method, including the above.

7. Obtaining measurement data for the gas generated from the object to be determined from a plurality of gas sensors that have different reactivity from each other depending on the composition of the gas, The determination of an abnormality concerning the target for determination is performed by comparing the determination data obtained from the measurement data of each of the plurality of gas sensors for the target for determination, the determination data obtained according to the relationship between the measurement data, and reference data. Including, The aforementioned reference data is data obtained according to the relationship between the measurement data of the multiple gas sensors for the gas generated from a normal learning target, The process further includes generating the reference data by normalizing the measurement data of the learning target measured by the multiple gas sensors using the amplitude values ​​of the measurement data of the learning target measured by the gas sensor among the multiple gas sensors that has a lower specificity than a predetermined standard for multiple measurement targets measured by the multiple gas sensors to date. Abnormality determination method.

8. Computers, An acquisition means for acquiring measurement data on the gas generated from a target for determination from multiple gas sensors that have different reactivity to each other depending on the gas composition. A determination means for performing an abnormality determination regarding the target to be determined by comparing determination data obtained from the measurement data of each of the plurality of gas sensors for the target to be determined, the determination data obtained according to the relationship between the measurement data, and reference data, and A standardization means that generates the determination data by normalizing the measurement data measured by the multiple gas sensors using the amplitude values ​​of the measurement data for the determination target measured by the gas sensor among the multiple gas sensors that has a lower specificity than a predetermined standard for the multiple measurement targets measured by the multiple gas sensors so far. An anomaly detection program that functions as such.

Images