Classification system, classification method, control program, recording medium

Abstract

To easily and quickly measure and classify the odor of objects produced by a manufacturing line that performs at least one manufacturing process. [Solution] The classification system of the present disclosure comprises: an acquisition unit (11) that acquires first data including detection signals output from each of the one or more odor sensor elements (311) of an odor measuring device (30); and a classification unit (12) that classifies the odor of an object of interest using the first data based on the results of a statistical analysis of second data corresponding to the odor of each of a plurality of objects, or an analysis based on feature quantities indicating the degree of similarity between each of the plurality of second data, and determines a criterion for classifying the object of interest based on the results of the analysis.

Description

[Technical Field]

[0001] The present disclosure relates to a classification system for classifying an odor of an object. [Background Art]

[0002] Conventionally, when analyzing the odor of an object, the type and content of odorant substances derived from the object are analyzed by analytical instruments such as gas chromatography (GC) and high performance liquid chromatography (HPLC). Such analysis has had problems that it requires labor such as collecting a sample to be analyzed and starting up the analytical instrument, and furthermore, it takes a long time to obtain an analysis result.

[0003] There is known a system that measures the odor of an object without performing analysis using analytical instruments such as GC and HPLC, and evaluates or predicts the quality of the object. For example, Patent Document 1 discloses a method for predicting quality based on scent (aroma). Further, Patent Document 2 discloses an odor identification device that identifies an odor factor of an object to be measured by comparing an interaction pattern between a sensor and the odor factor of the object to be measured with an interaction pattern of a known object to be measured. [Prior Art Documents] [Patent Documents]

[0004] [Patent Document 1] International Publication No. 2022 / 004828 [Patent Document 2] Japanese Unexamined Patent Application Publication No. 2020-012846 [Summary of the Invention] [Problems to be Solved by the Invention]

[0005] With conventional techniques, measuring and classifying the odor of intermediate and / or final products produced by a manufacturing line involving at least one manufacturing step is time-consuming and laborious.

[0006] One aspect of this disclosure aims to realize a classification system that can easily and quickly measure and classify the odor of an object produced by a manufacturing line including at least one manufacturing process. [Means for solving the problem]

[0007] A classification system according to one aspect of the present disclosure includes: an acquisition unit that acquires first data corresponding to the odor of an object of interest manufactured in at least one manufacturing process, the first data including detection signals output from each of one or more odor sensor elements of an odor measuring device; and a classification unit that classifies an object of interest using the first data based on the results of an analysis of a plurality of second data corresponding to the odor of each of a plurality of objects manufactured in the manufacturing process, the plurality of second data including detection signals output from each of the odor sensor elements, wherein the analysis is a statistical analysis or an analysis based on feature quantities indicating the degree of similarity between each of the plurality of second data, and the classification unit determines a criterion for classifying the object of interest based on the results of the analysis.

[0008] A classification method according to one aspect of the present disclosure is a classification method performed by one or more computers, comprising: an acquisition step in which the computer acquires first data corresponding to the odor of an object of interest manufactured in at least one manufacturing process, the first data including detection signals output from each of one or more odor sensor elements of an odor measuring device; and a classification step in which the computer classifies an object of interest using the first data based on the results of an analysis of a plurality of second data corresponding to the odor of each of a plurality of objects manufactured in the manufacturing process, the plurality of second data including detection signals output from each of the odor sensor elements. [Effects of the Invention]

[0009] According to one aspect of this disclosure, the odor of an object produced by a manufacturing line that performs at least one manufacturing process can be easily and quickly measured and classified. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing an example configuration of a classification system according to Embodiment 1 of this disclosure. [Figure 2] This is a functional block diagram showing an example configuration of an odor measuring device and an information processing device. [Figure 3] This is a top view showing an example of the configuration of an odor sensor element. [Figure 4] This figure shows an example of a detected signal. [Figure 5] This figure shows an example of the data structure of the first data output from the odor measuring device. [Figure 6] This figure shows an example of the data structure of the first data database. [Figure 7] This flowchart shows an example of the processing flow performed by an information processing device. [Figure 8] This figure shows an example of the classification results obtained by classifying the odor of an object. [Figure 9] A functional block diagram showing an example configuration of a classification system according to Embodiment 2 of this disclosure. [Figure 10] This is a block diagram showing another example configuration of a classification system. [Figure 11] This is a block diagram showing yet another example of a classification system configuration. [Figure 12] This figure shows an example of the classification results. [Figure 13] This figure shows an example of the classification results. [Figure 14] This figure shows an example of the classification results. [Modes for carrying out the invention]

[0011] [Embodiment 1] An embodiment of the present disclosure will be described in detail below.

[0012] (Overview of Classification System 100) First, the configuration of the classification system 100 according to the present embodiment will be described with reference to FIG. 1. FIG. 1 is a block diagram showing an example of the main configuration of the classification system 100. Hereinafter, a case where the classification system 100 according to the present disclosure is applied to classify the odor of an object generated by a production line L including at least one production process will be described as an example. Here, the object may be an intermediate product and / or a final product produced on the production line L. In this case, the intermediate product and / or the final product may be a chemical product, a drug, a food, or the like. Further, an object actually classified by the classification system 100 is also referred to as a "target object of interest". There may be only one target object of interest, or there may be a plurality of target objects of interest.

[0013] FIG. 1 shows an example in which the classification system 100 is applied to a production line L including a first production process for producing a first intermediate product, a second production process for producing a second intermediate product from the first intermediate product, and a third production process for producing a final product from the second intermediate product. That is, in the classification system 100 shown in FIG. 1, at least any one of the first intermediate product, the second intermediate product, and the final product can be the object.

[0014] FIG. 1 shows the classification system 100 having a configuration in which three odor measuring devices 30 disposed on or near the production line L measure the odors of the first intermediate product, the second intermediate product, and the final product, respectively. The number of the odor measuring devices 30 included in the classification system 100 only needs to be 1 or more, and may be 2 or 4 or more.

[0015] Each of the three odor measuring devices 30 shown in Figure 1 transmits first data, including a detection signal corresponding to the odor of the object of interest (odor of the first intermediate product, odor of the second intermediate product, and odor of the final product), to the information processing device 10. Therefore, by applying the classification system 100 to the production line L, the user does not need to take a sample containing part of the object or a sample containing the substance that causes the odor of the object (hereinafter referred to as "odor substance"), nor do they need to prepare analytical instruments such as GC and HPLC. In other words, the classification system 100 can easily and quickly measure and classify the odor of the object produced by the production line L.

[0016] There is a possibility that some abnormality may occur in the manufacturing process of the object in question, and the smell of the object produced by the abnormal manufacturing process may differ from the smell of the object produced when the manufacturing process is functioning normally. For example, in a manufacturing process that involves blending multiple chemical substances to produce an object, if an abnormality occurs in the manufacturing process, the blending ratio of the multiple chemical substances may become inaccurate, potentially leading to uneven quality of the object. In such cases, the smell of the object produced by the abnormal manufacturing process may reflect the inaccuracy in the blending ratio of the multiple chemical substances.

[0017] Furthermore, even objects manufactured using the same process may exhibit variations in quality. If there is a correlation between the odor of an object and its quality, it may be possible to estimate the quality of the object based on its odor.

[0018] The classification results of the classification system 100 regarding the odor of objects produced by the manufacturing line L can be used to determine whether or not an abnormality has occurred in the manufacturing line L, and / or to evaluate the quality of the objects. Furthermore, the odor measuring device 30 can also measure the odor of objects containing odor substances unsuitable for human sensory testing (e.g., biotoxic odor substances). Therefore, the classification system 100 can classify the odor of objects that conventionally cannot be subjected to sensory testing.

[0019] (Configuration of classification system 100) The classification system 100 includes an information processing device 10 that is communicatively connected to one or more odor measuring devices 30 located on or near the production line L. In one example, the information processing device 10 may be a computer used by the owner or manager of the production line L. The classification system 100 may also include a display device 20 capable of displaying information output from the information processing device 10. The display device 20 may be a display unit of a computer used by the owner or manager of the production line L, or a display device that is communicatively connected to said computer.

[0020] The configuration of the classification system 100 will be explained using Figure 2. Figure 2 is a block diagram showing an example configuration of the odor measuring device 30 and the information processing device 10. In the classification system 100, the odor measuring device 30 and the information processing device 10 may be connected via a wide-area communication network. The wide-area communication network is not particularly limited and may be a network capable of long-distance communication such as the Internet, telephone lines, mobile communication networks, CATV communication networks, and satellite communication networks. The wide-area communication network may be a local area network (LAN) connection that does not go through an ISP, etc., or it may be a serial communication, a mobile phone network including a 5G communication network, LPWA (Low Power Wide Area-network), Wi-Fi (registered trademark), PAN (Personal Area Network), Bluetooth (registered trademark), etc.

[0021] (Configuration of odor measuring device 30) Next, the configuration of the odor measuring device 30 will be described. The odor measuring device 30 is placed on or near the production line L and measures the odor of the object manufactured by the production line L. As shown in Figure 2, the odor measuring device 30 includes an odor sensor unit 31 having one or more odor sensor elements 311, and a communication unit 32.

[0022] The odor sensor unit 31 may further include a power supply (not shown) for supplying power to the metal wiring 313 (described later) and a voltmeter (not shown) for measuring the voltage applied to each odor sensor element 311. The odor sensor unit 31 may also include a mechanism for supplying air containing odor substances originating from the target object to the odor sensor elements 311, and a mechanism for supplying dry air or N2 gas, etc., that does not contain odor substances as a purge gas.

[0023] The odor sensor element 311 is an element that can adsorb / desorb odor substances when exposed to air containing such substances, and outputs a detection signal that shows the change in the electrical conductivity of the odor sensor element 311 over time before and after the odor substance is adsorbed onto the odor sensor element 311. The detection signal may include signals that show the change in the electrical conductivity of the odor sensor element 311 over time during the following first period P1, second period P2, and third period P3. • Period 1 P1: The period during which the odor sensor element 311 was exposed to a purge gas that did not contain odor substances before being exposed to air containing odor substances. • Second period P2: The period during which the odor sensor element 311 was exposed to air containing odor substances. • Third period P3: The period during which the odor sensor element 311 was again exposed to a purge gas that did not contain odor substances.

[0024] The odor sensor unit 31 of the odor measuring device 30 may have a function to transmit the first data to the information processing device 10. Alternatively, the odor measuring device 30 may have a configuration in which the odor sensor unit 31 and the main unit, which can receive the first data from the odor sensor unit 31 and transmit the received first data to the information processing device 10, are separate (or detachable). In these cases, only the function of the odor sensor unit 31 can be placed on or near the manufacturing line L.

[0025] The odor measuring device 30 transmits the first data to the information processing device 10 via the communication unit 32. The communication unit 32 may also acquire detection signals output from each odor sensor element 311 from the odor sensor unit 31 and transmit the first data, including the acquired detection signals, to the information processing device 10. Specific examples of the detection signals and the first data will be explained later.

[0026] As described above, the odor measuring device 30 includes one or more odor sensor elements 311 capable of outputting detection signals, and each odor sensor element 311 is equipped with an odor substance receiving layer 315 in which interactable odor substances are different from each other. The odor sensor elements 311 will be described below.

[0027] Figure 3 is a top view showing an example of the configuration of the odor sensor element 311. The odor sensor element 311 comprises an odor substance receiving layer 315 containing the resin composition described above, a first metal wiring 313A, and a second metal wiring 313B. In the following, when the first metal wiring 313A and the second metal wiring 313B are not distinguished, they may be referred to simply as metal wiring 313.

[0028] The first metal wiring 313A and the second metal wiring 313B are metal wirings that function as electrodes for measuring changes in the electrical conductivity of the odor substance receiving layer 315 (i.e., the resin composition). That is, the first metal wiring 313A and the second metal wiring 313B are spaced apart from each other, and the odor substance receiving layer 315 is in contact with at least a portion of the first metal wiring 313A and at least a portion of the second metal wiring 313B. In one example, the first metal wiring 313A and the second metal wiring 313B are metal wirings that are not in direct contact with each other, and may be substantially parallel to each other, as shown in Figure 3.

[0029] As shown in Figure 3, the metal wiring 313, including the first metal wiring 313A and the second metal wiring 313B, may be arranged on a substrate 314. The substrate 314 may be a substrate such as glass epoxy, which is commonly used in electronic circuits. The metal wiring 313 may be made of copper or gold. The thickness of the first metal wiring 313A and the second metal wiring 313B, as viewed from a direction perpendicular to the surface of the substrate 314, may be, for example, 10 μm to 2 mm.

[0030] The odor substance receiving layer 315 may be in contact with at least a portion of the first metal wiring 313A and at least a portion of the second metal wiring 313B. The odor substance receiving layer 315 may be arranged to fill the region sandwiched between the first metal wiring 313A and the second metal wiring 313B, for example, as shown in Figure 3.

[0031] If the electrical conductivity of the odor substance receiving layer 315 (i.e., the electrical conductivity of the odor sensor element 311) is low, it is desirable that the distance between the first metal wiring 313A and the second metal wiring 313B be less than or equal to a predetermined distance (for example, 500 μm).

[0032] The odor substance receiving layer 315 may contain a resin composition. The resin composition may contain a resin and further contain one or more selected from a surfactant and a filler (e.g., a conductive carbon material). In this specification, "odor substance receiving layer" means a layer capable of adsorbing and desorbing odor substances to be identified. The odor substance receiving layer 315 is formed from the above-described resin composition. The odor substance receiving layer 315 may be provided as part of an odor sensor element 311. The electrical resistance of this odor substance receiving layer 315 changes in response to the adsorption of odor substances, etc. That is, the odor sensor element 311 is an odor detection device equipped with such an odor substance receiving layer 315, and the odor measurement method of the odor sensor element 311 may be a chemiresistor type.

[0033] If the odor sensor element 311 is a chemisistor type containing a resin composition, the change in electrical conductivity over time differs depending on whether odor substance A is adsorbed or odor substance B, which is different from odor substance A, is adsorbed. Therefore, it is possible to detect and identify various odor substances. In the odor measuring device 30 described later, multiple odor sensor elements 311 are arranged on a substrate 314 provided with a configuration for detecting odor substances (metal wiring 313 and odor substance receiving layer 315). Each substrate 314 is provided with multiple sets of odor substance receiving layers 315, each containing odor substances that can be measured differently. Each of the multiple odor sensor elements 311 may be equipped with a constant voltage power supply and a voltmeter. In the odor measuring device 30, one configuration for detecting odor substances (metal wiring 313 and odor substance receiving layer 315) may be provided on each substrate 314. Alternatively, in the odor measuring device 30, multiple sets of configurations for detecting odor substances (metal wiring 313 and odor substance receiving layer 315) may be provided on a single substrate 314. In the latter case, a constant voltage power supply and a voltmeter are connected to each of the sets provided on the circuit board 314.

[0034] The resin compositions contained in the odor substance receiving layers 315 of the multiple odor sensor elements 311 in the odor measuring device 30 may be the same or different. If the odor substance receiving layers 315 contained in the multiple odor sensor elements 311 have the same composition, each of the multiple odor substance receiving layers 315 can detect the same odor substance. If the multiple odor sensor elements 311 each contain odor substance receiving layers 315 with different compositions, each of the multiple odor substance receiving layers 315 will respond differently to the odor substance. In this way, by providing multiple sets of configurations for detecting odor substances, the accuracy of odor substance identification in the odor measuring device 30 can be improved.

[0035] The odor measuring device 30 described above can output the change in the electrical conductivity of the odor sensor element 311 over time for each odor substance when various odor substances are adsorbed onto the odor sensor element 311. By applying this odor measuring device 30, the change in the electrical conductivity of the odor sensor element 311 over time when odor substance A is adsorbed onto the odor sensor element 311 can be compared with the change in the electrical conductivity of the odor sensor element 311 over time when odor substance B is adsorbed onto the odor sensor element 311.

[0036] [Detection signal and first data] Each of the odor sensor elements 311 is capable of outputting a detection signal. The detection signal may be, for example, the change over time in the voltage value indicating the electrical conductivity measured for the odor sensor element 311. The detection signal will be explained using Figure 4. Figure 4 is a diagram showing an example of a detection signal. Figure 4 shows the detection signal S output from a certain odor sensor element 311 (for example, an odor sensor element 311 whose odor sensor element ID is "CH1").

[0037] The signal in the first period P1 represents the voltage value (baseline) output when no odor substances are adsorbed on the odor substance receiving layer 315 of the odor sensor element 311. In the second period P2, the voltage value increases. This is because the adsorption of odor substances onto the odor substance receiving layer 315 of the odor sensor element 311 causes a volume change in the odor substance receiving layer 315, resulting in a change in the conductivity of the odor substance receiving layer 315. In the third period P3, the voltage value returns to approximately the baseline. This is because the supply of purge gas desorbs the odor substances adsorbed on the odor substance receiving layer 315 of the odor sensor element 311, and the volume and conductivity of the odor substance receiving layer 315 return to the same volume and conductivity as in the first period P1.

[0038] The amount of odor substances adsorbed onto the odor substance receiving layer 315 of the odor sensor element 311 may depend on the concentration (amount) of odor substances in the gas containing odor substances originating from the target object. Furthermore, the magnitude and rate of change in the electrical conductivity of the odor sensor element 311 differ depending on the type, concentration (amount), and relative abundance of odor substances adsorbed onto the odor substance receiving layer 315 of the odor sensor element 311. In other words, the detection signal reflects the type, concentration (amount), and relative abundance of odor substances contained in the measured odor.

[0039] The odor measuring device 30 may be equipped with multiple types of odor sensor elements 311, each having different specificities for various odor substances. For example, the odor sensor unit 31 may be equipped with multiple odor sensor elements 311, and the composition of the odor substance receiving layer 315 of each of the multiple odor sensor elements 311 may be different from each other. In this case, the odor measuring device 30 outputs data including detection signals output from each odor sensor element 311, with the detection signal corresponding to the odor of the object of interest being used as the first data. Figure 5 is a diagram showing an example of the data structure of the first data output from the odor measuring device 30. Figure 5 shows an example of the first data corresponding to the odor of an object of interest, measured by an odor measuring device 30 equipped with n (or n types) odor sensor elements 311.

[0040] The odor measuring device 30 is located on or near a manufacturing line that includes at least one manufacturing process, and outputs a plurality of second data corresponding to the odor of each of a plurality of objects manufactured in at least one of the at least one manufacturing process. The plurality of second data includes detection signals output from each of the odor sensor elements 311. For example, the odor measuring device 30 may output a plurality of second data corresponding to the odor of each of the first intermediate products manufactured in the first manufacturing process. Alternatively, the odor measuring device 30 may output a plurality of second data corresponding to the odor of each of the final products manufactured in the third manufacturing process.

[0041] Each of the multiple second data sets output from the odor measuring device 30 contains a detection signal that reflects the concentration (amount) and relative abundance of odor substances contained in the odor of the measured object. Therefore, the information processing device 10 can classify the odor of the object based on the first data acquired from the odor measuring device 30.

[0042] Here, the relationship between the multiple objects whose odors are measured in order to output multiple second data points and the object of interest whose odor is measured in order to output the first data point may be, for example, either of the following (1) and (2).

[0043] (1) Multiple objects and the object of interest are objects produced by the same manufacturing process. For example, the object of interest may be an object that will be produced in the future (or was produced in the past) by a manufacturing process using the same equipment as the multiple objects.

[0044] (2) Multiple objects and the object of interest are objects produced by manufacturing processes using similar equipment. For example, the object of interest may be an object produced by a manufacturing process using equipment installed in a different manufacturing plant from the multiple objects, but which is similar to the manufacturing process.

[0045] Furthermore, the odor measuring device 30 does not have a mechanism for supplying purge gas, and may instead have a configuration that continuously (or at regular time intervals) supplies air containing odor substances originating from the target object to the odor sensor element 311 of the odor sensor unit 31. In this case, air containing odor substances originating from the target object is continuously supplied to the odor sensor unit 31. During this time, the odor sensor unit 31 continues to measure odors and output detection signals.

[0046] For example, if the odor measuring device 30 (or odor sensor unit 31) is placed on or near the transport path for transporting the object of interest, the voltage value indicated by the detection signal included in the first data output from the odor measuring device 30 may change depending on the distance between the object of interest and the odor measuring device 30. As an example, the voltage value indicated by the detection signal will be lower when the object is not near the odor measuring device 30, and higher when the object passes near the odor measuring device 30. Therefore, the information processing device 10 can classify the odors of multiple objects based on a plurality of second data sets, each of which is measured when each of the multiple objects passes near the odor sensor unit 31, and which includes the detection signal output from each of the odor sensor elements 311.

[0047] (Configuration of the information processing device 10) Returning to Figure 2, the configuration of the information processing device 10 will be explained. The information processing device 10 acquires first data corresponding to the odor of the object of interest from the odor measuring device 30, and uses the acquired first data to classify the odor of the object of interest based on the results of the analysis of multiple second data corresponding to the odor of each of the multiple objects manufactured in the manufacturing process.

[0048] The information processing device 10 comprises a control unit 1 and a storage unit 2. In one example, the control unit 1 may be a CPU (Central Processing Unit). The control unit 1 reads the control program, which is software stored in the storage unit 2, and loads it into memory such as RAM (Random Access Memory) to execute various functions.

[0049] As shown in Figure 2, the control unit 1 includes an acquisition unit 11 and a classification unit 12. The storage unit 2 may store a second data DB (database) 21 and a classification model 22. Note that, for the sake of simplicity, the control program is not shown in the storage unit 2 shown in Figure 2.

[0050] The acquisition unit 11 acquires first data from the odor measuring device 30 that corresponds to the odor of the object of interest manufactured in at least one of the at least one manufacturing process. If there are multiple acquired first data, the acquisition unit 11 may associate each first data with identification information (e.g., object ID) indicating the object of interest corresponding to that object, and store it in the storage unit 2.

[0051] The acquisition unit 11 may acquire a plurality of second data sets corresponding to the smell of each of the plurality of objects manufactured in the manufacturing process, and include detection signals output from each of the odor sensor elements 311, and store them in the second data DB 21 of the storage unit 2. Figure 6 is a diagram showing an example of the data structure of the second data DB 21. Figure 6 shows the second data DB 21, which includes second data corresponding to the smell of an object with object ID "P1", and second data obtained by measuring the smell of an object with object ID "P2", etc.

[0052] The classification unit 12 classifies the odor of the object of interest using the first data based on the results of analyzing multiple acquired second data sets. The classification unit 12 may also be configured to output the results of analyzing multiple second data sets. The classification model 22 is a model for analyzing multiple second data sets. The classification model 22 and the analysis will be explained later.

[0053] The control unit 1 may further include an output control unit 14 that outputs the classification results from the classification unit 12 to the display device 20, and may also be configured to notify the owner or manager of the manufacturing line L of the classification results. This allows the owner or manager of the manufacturing line L to determine the quality of the object of interest and whether or not there are any defects in the manufacturing process that produced the object, based on the classification results for the odor of the object of interest output to the display device 20.

[0054] (Processing flow performed by the information processing device 10) Figure 7 is a flowchart showing an example of the processing (classification method) performed by the information processing device 10. As shown in Figure 7, first, the acquisition unit 11 acquires first data corresponding to the odor of the object of interest from the odor measuring device 30 located on or near the manufacturing line L, and includes detection signals output from each of the one or more odor sensor elements provided by the odor measuring device (step S1: acquisition step).

[0055] Next, the classification unit 12 uses the first data to classify the odor of the object of interest based on the results of the analysis of multiple second data (step S2: classification step).

[0056] Next, the output control unit 14 outputs the classification result to the display device 20 (step S3).

[0057] A classification system 100 equipped with an information processing device 10 that performs such processing can easily and quickly measure the odor of an object of interest manufactured by a manufacturing line L that includes at least one manufacturing process, and can classify the odor of the object based on the acquired first data.

[0058] [Classification Model 22] Classification model 22 is a model for analyzing second data (including one or more detection signals) corresponding to the smell of each of multiple objects. Here, the analysis may be statistical analysis or analysis based on feature quantities that indicate the degree of similarity between each of the multiple second data. Feature quantities may be, for example, statistics, differential and integral values, peak detection values, or autocorrelation values. Examples of statistics include the mean, variance, maximum, minimum, difference between the maximum and minimum, and standard deviation. Examples of differential and integral values ​​include the differential value (slope of a graph showing the change in measured values ​​over time) and the integral value (area of ​​the region defined by the curve showing the change in measured values ​​in a graph showing the change in measured values ​​over time and the horizontal axis (e.g., time axis)). Examples of peak detection values ​​include the number and height of peaks in the change in measured values ​​(e.g., change over time). Examples of autocorrelation values ​​include the difference in the change in measured values ​​(e.g., change over time). Extraction of these feature quantities from measured values ​​can be performed based on known methods. Examples of feature-based analysis methods include factor analysis, multivariate analysis, cluster analysis (K-means method or hierarchical cluster analysis), and multidimensional scaling.

[0059] In one embodiment, the statistical analysis may be a multivariate analysis. In this case, the classification unit 12 may perform a multivariate analysis to calculate the analysis results and determine a boundary line for classifying the object of interest or a function for classifying the object of interest based on the results of the analysis.

[0060] In one embodiment, the analysis may be principal component analysis. In this case, the classification unit 12 may calculate one or more principal component values ​​corresponding to each of the acquired second data, and determine a function representing a boundary line BL that serves as a criterion for classifying the odors of the multiple objects based on the calculated one or more principal component values. The classification model 22 may be generated by supervised learning or by unsupervised learning.

[0061] The classification model 22 may be a classification model generated by unsupervised learning using multiple second data sets. In this case, the classification unit 12 uses the classification model 22 to classify the object of interest based on the first data. If the classification model 22 is generated by unsupervised learning, the classification unit 12 can classify the object of interest by only providing training data corresponding to a specific classification.

[0062] If the classification model 22 is generated by unsupervised learning using multiple second data sets, the classification model 22 may be subject to known algorithms for performing clustering of the second data (including one or more detection signals) measured for multiple targets. In this case, the algorithm for performing clustering may be a convolutional neural network (CNN), k-nearest neighbor method, Gaussian distribution (e.g., single Gaussian distribution, Gaussian mixture distribution), or support vector machine (SVM).

[0063] The training data used to generate the classification model 22 may include only a plurality of second data corresponding to the smell of each object that satisfies predetermined conditions among the plurality of objects manufactured in the manufacturing process. The objects that satisfy the predetermined conditions are not particularly limited. For example, if there are normal objects and abnormal objects among the plurality of objects manufactured in the manufacturing process, the normal objects may be selected as objects that satisfy the predetermined conditions. With the above configuration, for example, even if only objects that satisfy the predetermined conditions exist, the classification unit 12 can classify the object of interest.

[0064] The classification model 22 may be generated by supervised learning. If the classification model 22 is generated by supervised learning, it may be generated by machine learning using training data that includes the plurality of second data as explanatory variables and the classification result of the object corresponding to each of the plurality of second data as the target variable.

[0065] The machine learning algorithms that can be used to create the classification model 22 may be either supervised or unsupervised. Examples of machine learning algorithms include regression analysis, classification, trees, time series analysis, neural networks, and clustering. Examples of regression analysis include logistic regression, Lasso regression, elastic network regression, support vector regression (SVR), linear regression, Ridge regression, and ensemble regression. Examples of classification include k-nearest neighbor method, support vector classification (SVC), naive Bayes classifier, stochastic gradient descent (SGD), and kernel approximation. Examples of trees include decision trees, regression trees, random forests, boosting (LightGBM, XGBoost), and stacking. Examples of time series include AR, MA, ARIMA, and state space. Examples of neural networks include multilayer perceptrons (MLPs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), residual neural networks (ResNets), transformers, and graph neural networks (GNNs). Examples of clustering methods include Gaussian mixture models (GMMs), k-means algorithms, mini k-means algorithms, variational Gaussian mixture models (VBGMMs), and kernel approximations.

[0066] In the following, we will explain using an example configuration in which the classification unit 12 determines a function representing the boundary line BL, which serves as a criterion for classifying the odor of each of the multiple objects, by performing principal component analysis on the second data measured for multiple objects.

[0067] [Classification result] Figure 8 shows an example of a classification result for the odor of an object. For example, as shown in Figure 8, the results of principal component analysis of the second data may be plotted on a two-dimensional coordinate system defined by the first principal component value (PC1) and the second principal component value (PC2) for each of the second data measured for multiple objects. Note that the number of principal components calculated by principal component analysis is not limited to two; there may be one or three or more. In other words, the coordinate system showing the classification result is not limited to two dimensions.

[0068] For example, the classification unit 12 plots the second data corresponding to the smell of each of the multiple objects on a two-dimensional coordinate system defined by the first principal component value (PC1) and the second principal component value (PC2). In the example shown in Figure 8, the smells of each of the multiple objects are classified into a first group GR1 and a second group GR2. Furthermore, in the example shown in Figure 8, a boundary line BL is shown, which serves as the criterion for classifying the first group GR1 and the second group GR2.

[0069] If the classification result is shown in a two-dimensional coordinate system with the first principal component value (PC1) as the x-axis and the second principal component value (PC2) as the y-axis, and the boundary line BL is a straight line in that dimensional coordinate system, then the boundary line BL can be determined as the function y = ax + b. Note that the boundary line BL is not limited to a straight line but may also be a curve. In other words, the function determined by the classification unit 12 may be any function that can better classify the smell of each of the multiple objects into one or more groups, and is not limited to a linear function. For example, the function determined by the classification unit 12 may be a linear function of degree two or higher, a logarithmic function, or a trigonometric function. The function determined by the classification unit 12 may also be expressed in implicit function form.

[0070] [Embodiment 2] Other embodiments of this disclosure are described below. For the sake of clarity, components having the same function as those described in the above embodiments are denoted by the same reference numerals, and their descriptions are not repeated.

[0071] Here, we describe a classification system 100a that further includes a function to determine the quality of an object of interest based on first data obtained by measuring the odor of the object of interest.

[0072] (Configuration of classification system 100a) The configuration of the classification system 100a will be explained using Figure 9. Figure 9 is a block diagram showing an example configuration of the classification system 100a. The classification system 100a includes an information processing device 10a, which further includes a determination unit 13 that determines the quality of each object based on the classification results from the classification unit 12.

[0073] In the classification system 100a, the odor measuring device 30 and the information processing device 10a may be connected via a wide-area communication network. The wide-area communication network is not particularly limited and may be a network capable of long-distance communication such as the Internet, telephone lines, mobile communication networks, CATV communication networks, and satellite communication networks. The wide-area communication network may be a local area network (LAN) connection that does not go through an internet service provider, or it may be a serial communication network, a mobile phone network including a 5G communication network, LPWA (Low Power Wide Area-network), Wi-Fi (registered trademark), PAN (Personal Area Network), Bluetooth (registered trademark), etc.

[0074] (Configuration of information processing device 10a) Next, the configuration of the information processing device 10a will be described. The information processing device 10a acquires first data corresponding to the odor of the object of interest from the odor measuring device 30. Furthermore, it uses the acquired first data to classify the odors of multiple objects. In addition, the information processing device 10a determines the quality of the object of interest from the classification results based on the judgment criteria 23.

[0075] The information processing device 10a comprises a control unit 1a and a storage unit 2a. In one example, the control unit 1a may be a CPU (Central Processing Unit). The control unit 1a reads the control program, which is software stored in the storage unit 2a, and loads it into memory such as RAM (Random Access Memory) to execute various functions.

[0076] As shown in Figure 9, the control unit 1a differs from the control unit 1 of the information processing device 10 in that it further includes a determination unit 13. The storage unit 2a differs from the storage unit 2 of the information processing device 10 in that it further stores the determination criteria 23. Note that, for the sake of simplicity, the control program is not shown in the storage unit 2a shown in Figure 9.

[0077] The determination unit 13 determines the quality of the object of interest based on the classification result from the classification unit 12, using the determination criteria 23 described later.

[0078] [Judgment Criteria 23] Criterion 23 is a standard for the quality of multiple objects, defined based on the results of the analysis of multiple second data sets. For example, it may be confirmed in advance by the owner or manager of the manufacturing line L (or the person checking the quality of the objects) that objects belonging to the first group GR1 shown in Figure 8 are of first quality, and objects belonging to the second group GR2 shown in Figure 8 are of second quality. In this case, Criterion 23 can be generated by associating first quality with the first group GR1, while associating a second quality, which is different from first quality, with the second group GR2 shown in Figure 8.

[0079] Furthermore, either the first quality or the second quality may be a high-quality product, or a quality that conforms to laws and / or standards regarding the quality of the subject object, while the other may be a low-quality product, or a quality that does not conform to laws and / or standards regarding the quality of the subject object. Furthermore, if the classification result includes three groups, criterion 23 may define each group as "high quality," "medium quality," and "low quality."

[0080] The control unit 1a may further include an output control unit 14 that outputs the determination result from the determination unit 13 to the display device 20, and may also be configured to notify the owner or manager of the manufacturing line L of the determination result. This allows the owner or manager of the manufacturing line L to determine, based on the determination result regarding the quality of the object of interest output to the display device 20, whether the quality of the object of interest is good or bad, and whether or not there is a defect in the manufacturing process that produced the object of interest.

[0081] [Embodiment 3] Other embodiments of this disclosure are described below. For the sake of clarity, components having the same function as those described in the above embodiments are denoted by the same reference numerals, and their descriptions are not repeated.

[0082] (Configuration of classification system 100b) Figure 10 is a block diagram showing an example of the main components of the classification system 100b. As shown in Figure 10, the classification system 100b includes an information processing device 10 that is communicatively connected to a communication device 40a located in the manufacturing plant FA and a communication device 40b located in the manufacturing plant FB via a wide-area communication network 50. In this case, the information processing device 10 may be a server device installed in a facility that manages multiple manufacturing plants (for example, a central management facility), or it may be a cloud server configured using cloud computing.

[0083] The communication device 40a is a computer used by the owner or manager of the manufacturing line L at the manufacturing plant FA. Additionally, an odor measuring device 30a is located on or near the manufacturing line L at the manufacturing plant FA. The odor measuring device 30a measures the odor of a target object generated by the manufacturing process included in the manufacturing line La (not shown) at the manufacturing plant FA. The odor measuring device 30a transmits first data to the communication device 40a. The communication device 40a transmits the first data obtained from the odor measuring device 30a to the information processing device 10.

[0084] The communication device 40b is a computer used by the owner or manager of the production line L of the manufacturing plant FB. Additionally, an odor measuring device 30b is located on or near the production line L of the manufacturing plant FB. The odor measuring device 30b measures the odor of a target object generated by the manufacturing process included in the production line Lb (not shown) of the manufacturing plant FB. The odor measuring device 30b transmits first data to the communication device 40b. The communication device 40b transmits the first data obtained from the odor measuring device 30b to the information processing device 10.

[0085] The classification unit 12 of the information processing device 10 analyzes the first data received from the communication device 40a and classifies the odor of the object of interest manufactured at the manufacturing plant FA. The information processing device 10 then transmits the classification result to the communication device 40a. Meanwhile, the classification unit 12 of the information processing device 10 uses the first data received from the communication device 40b to classify the odors of multiple objects manufactured at the manufacturing plant FB. The information processing device 10 then transmits the classification result to the communication device 40b.

[0086] With this configuration, the classification system 100b can classify the odor of objects produced by production lines equipped with facilities located in each of multiple manufacturing plants. This allows managers overseeing multiple manufacturing plants to centrally manage the production status of objects at each plant.

[0087] In addition, in the classification system 100b, the information processing device 10a shown in Figure 9 may be used instead of the information processing device 10 shown in Figure 10. In this case, the information processing device 10a transmits the quality judgment results of multiple objects manufactured at the manufacturing plant FA to the communication device 40a, and transmits the quality judgment results of multiple objects manufactured at the manufacturing plant FB to the communication device 40b. This allows a manager overseeing multiple manufacturing plants to centrally manage the quality of objects at each manufacturing plant.

[0088] [Embodiment 4] Other embodiments of this disclosure are described below. For the sake of clarity, components having the same function as those described in the above embodiments are denoted by the same reference numerals, and their descriptions are not repeated.

[0089] (Configuration of classification system 100c) Figure 11 is a block diagram showing an example of the main components of the classification system 100c. As shown in Figure 11, the classification system 100c includes an odor measuring device 30 and an information processing device 10. The objects to be classified by the classification system 100c are at least one of the first intermediate product, the second intermediate product, and the final product of the manufacturing line L. Note that the classification system 100c may also be configured by applying the information processing device 10a shown in Figure 9 instead of the information processing device 10.

[0090] The odor measuring device 30 may include a containment section capable of holding an object of interest (e.g., a solid or liquid), and may be capable of measuring the odor originating from the object of interest within the containment section. If the odor measuring device 30 further includes a containment section, the temperature inside the containment section may be adjusted, for example, to room temperature to 60°C. In one embodiment, the odor measuring device 30 may be capable of adjusting the concentration of odor substances originating from the object of interest within the containment section by adjusting the temperature inside the containment section.

[0091] In one embodiment, the concentration of the odor substance may be adjusted, for example, by concentrating the odor substance. When concentrating the odor substance, the odor measuring device 30 may further include an adsorption section. After the odor substance emitted from the object of interest stored in the containment section is adsorbed onto the adsorption section, the odor substance can be concentrated by heating the adsorption section. The temperature of the adsorption section may be, for example, room temperature to 250°C. In one example, the odor measuring device 30 may be installed at a location away from the manufacturing line L (for example, in an inspection room). In this case, the object to be tested by the odor measuring device 30 is transported from the manufacturing line L to the inspection room and stored in the containment section of the odor measuring device 30.

[0092] [Examples of implementation using software] The functions of the classification systems 100, 100a, 100b, and 100c (hereinafter referred to as "the systems") are programs that cause a computer to function as the system, and these programs can be realized by programs that cause a computer to function as each control block of the system (particularly the control unit 1 of the information processing device 10 and each part included in the control unit 1a of the information processing device 10a).

[0093] In this case, the system includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., memory) as hardware for executing the program. By executing the program using this control device and storage device, the functions described in each of the embodiments are realized.

[0094] The above program may be recorded on one or more computer-readable recording media, not temporary ones. These recording media may or may not be provided by the above device. In the latter case, the program may be supplied to the above device via any wired or wireless transmission medium.

[0095] Furthermore, some or all of the functions of each of the above control blocks can also be implemented by logic circuits. For example, an integrated circuit in which logic circuits functioning as each of the above control blocks are formed is also included in the scope of this disclosure. In addition, it is also possible to implement the functions of each of the above control blocks by, for example, a quantum computer.

[0096] This disclosure is not limited to the embodiments described above, 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 this disclosure. Furthermore, unless otherwise specified herein, "A to B" representing a numerical range means "greater than or equal to A (including A and greater than A) and less than or equal to B (including B and less than B)."

[0097] 〔summary〕 A classification system according to Embodiment 1 of the present disclosure includes: an acquisition unit that acquires first data corresponding to the odor of an object of interest manufactured in at least one manufacturing process, which includes detection signals output from each of one or more odor sensor elements of an odor measuring device; and a classification unit that classifies an object of interest using the first data based on the results of an analysis of a plurality of second data corresponding to the odor of each of a plurality of objects manufactured in the manufacturing process, which includes detection signals output from each of the odor sensor elements, wherein the analysis is a statistical analysis or an analysis based on feature quantities indicating the degree of similarity between each of the plurality of second data, and the classification unit determines a criterion for classifying the object of interest based on the results of the analysis.

[0098] In the classification system according to aspect 2 of this disclosure, in aspect 1 above, the analysis is a multivariate analysis, and the classification unit may perform the multivariate analysis on the plurality of second data to calculate the analysis results, and determine a boundary line for classifying the object of interest or a function for classifying the object of interest based on the results of the analysis.

[0099] In the classification system according to aspect 3 of this disclosure, in aspect 1 or 2 described above, the classification unit may classify the object of interest based on the first data using a classification model generated by unsupervised learning using the plurality of second data.

[0100] In the classification system according to aspect 4 of this disclosure, in aspect 3 described above, the training data used for training to generate the classification model may include only the second data corresponding to the smell of each of the multiple objects manufactured in the manufacturing process that satisfy predetermined conditions.

[0101] In the classification system according to aspect 5 of this disclosure, in any of aspects 1 to 4 described above, the classification unit classifies the object of interest from the first data using a classification model, and the classification model may be generated by machine learning using training data that includes the plurality of second data as explanatory variables and the classification result of the object corresponding to each of the plurality of second data as the objective variable.

[0102] In the classification system according to aspect 6 of the present disclosure, in any of aspects 1 to 5 described above, the system may further include a storage unit for storing criteria for determining the quality of the object of interest based on the classification result by the classification unit, and a determination unit that determines the quality of the object of interest from the classification result of the object of interest based on the criteria.

[0103] A classification method according to aspect 7 of the present disclosure is a classification method performed by one or more computers, comprising: an acquisition step in which the computer acquires first data corresponding to the odor of an object of interest manufactured in at least one manufacturing process, the first data including detection signals output from each of one or more odor sensor elements of an odor measuring device; and a classification step in which the computer classifies an object of interest using the first data based on the results of an analysis of a plurality of second data corresponding to the odor of each of a plurality of objects manufactured in the manufacturing process, the plurality of second data including detection signals output from each of the odor sensor elements.

[0104] The control program according to aspect 8 of this disclosure is a control program for causing a computer to function as a classification system in any of aspects 1 to 6 described above, and is a control program for causing the computer to function as the acquisition unit and the classification unit.

[0105] The storage medium according to aspect 9 of this disclosure is a computer-readable recording medium on which the control program described in aspect 8 is recorded. [Examples]

[0106] [Example 1] The present disclosure will be further illustrated by the following examples, but will not be limited thereto. In Example 1, five samples M1 to M5 were prepared, mimicking substances produced using xylene and ethylbenzene, and their odors were classified using the classification system (classification system 100 or classification system 100a) according to the present disclosure. Figure 12 shows an example of the classification results.

[0107] Five samples, M1 to M5, were prepared, each containing the following amounts of xylene and ethylbenzene. Sample M1: Xylene: 1.0% (w / v), Ethylbenzene: 0% (w / v). Sample M2: Xylene: 0.9% (w / v), Ethylbenzene: 0% (w / v). Sample M3: Xylene: 0% (w / v), Ethylbenzene: 0.9% (w / v). Sample M4: Xylene: 0.29% (w / v), Ethylbenzene: 0.32% (w / v). Sample M5: Xylene: 0% (w / v), Ethylbenzene: 0% (w / v).

[0108] As shown in Figure 12, the classification system according to this disclosure successfully classified the odors of each of the samples M1 to M5.

[0109] Furthermore, by applying the assumption that a judgment criterion has been set that a product will be rejected if the xylene content is 1.0% (w / v) or higher, the classification system relating to this disclosure succeeded in determining the boundary line BLa shown in Figure 12 based on the classification results of classifying the odor of each of the samples M1 to M5.

[0110] [Example 2] When coffee beverages containing milk are manufactured, an odor derived from hexanal may occur. Even a small amount of hexanal (for example, about 63 ppm) can impair the flavor of the coffee beverage. Therefore, in Example 2, a normal coffee beverage (Sample M6) and a coffee beverage to which a small amount (63 ppm) of hexanal was added (Sample M7) were prepared, and the odors of these were classified using the classification system according to this disclosure (Classification System 100 or Classification System 100a). Figure 13 shows an example of the classification results.

[0111] As shown in Figure 13, the classification system according to this disclosure successfully classified the odors of sample M6 and sample M7, respectively.

[0112] Furthermore, by applying the assumption that a criterion has been set that coffee beverages containing 63 ppm or more of hexanal are deemed unacceptable, the classification system according to this disclosure succeeded in determining the boundary line BLb shown in Figure 13 based on the classification results of the odors of samples M6 and M7, respectively.

[0113] [Example 3] In Example 3, the target product was manufactured by the following steps 1 to 3. 1. Xylene and ethylbenzene were added to mineral oil to obtain a mixture. 2. The mixture was heated to approximately 40°C while being stirred. 3. When the mixture had a uniform appearance, stirring was stopped and it was used as the target material.

[0114] The solvent content of the obtained samples was determined using a GC-2014S (Shimadzu Corporation). A DB-5 capillary column (length 30 m, inner diameter 0.25 mm, film thickness 0.25 μm) was used. As samples, an internal standard mother liquor (1.0 wt% 4-methylcyclohexanone / hexane solution) and a calibration curve mother liquor (0.5 wt% ethylbenzene and m-xylene / hexane solution) were used. The internal standard mother liquor was obtained by accurately weighing 0.1 g of 4-methylcyclohexanone into a screw tube or bottle, then accurately weighing hexane to a total volume of 10.0 g (recorded to the nearest 0.1 mg), and mixing. The calibration curve mother liquor was obtained by accurately weighing 0.25 g of ethylbenzene and 0.25 g of m-xylene into a screw tube or bottle, then accurately weighing hexane to a total volume of 50.0 g (recorded to the nearest 0.1 mg), and mixing. Calibration curve samples were prepared by quantifying the internal standard mother liquor in a screw-top tube or mayonnaise bottle, adding varying amounts of hexane as the solvent, weighing accurately (recorded to the nearest 0.1 mg), and using the resulting mixed solution. The composition of the solutions used to prepare the calibration curve samples is as follows: (1) Calibration mother liquor 2.5g, internal standard mother liquor 0.5g, hexane 47.0g (2) Calibration mother liquor 1.0g, internal standard mother liquor 0.5g, hexane 48.5g (3) Calibration mother liquor 0.2g, internal standard mother liquor 0.5g, hexane 49.3g Furthermore, the measurement samples were prepared by mixing the following components in the following ratios in a container such as a screw-top tube.

[0115] Object: 0.5000g Internal standard mother liquor: 0.5000g Hexane: 49.0000g In the sample being measured, the concentration of the target substance was 1.0000% by weight, and the concentration of 4-methylcyclohexanone was 0.0100% by weight.

[0116] The instrument conditions were set as follows, and measurements were performed on the prepared calibration curve sample and the measurement sample. The peak near 8 min in the retention time was identified as the peak for ethylbenzene, and the peak near 9 min was identified as the peak for m-xylene.

[0117] Carrier gas: He Evaporation chamber temperature: 200℃ Control mode: Pressure Inlet pressure (kPa): 100kPa Split ratio: 10 Injection volume: 1.0μL Column temperature: After holding at 40°C for 15 minutes, the temperature is increased to 180°C at a heating rate of 20°C / min, and then held for 10 minutes. Detector temperature: 200℃ Makeup flow rate: Recommended value for the device The obtained samples were classified into groups No. 1 to 12 according to their solvent (xylene and ethylbenzene) content. The classification results are shown in Table 1. In Table 1, the ethylbenzene and xylene content was measured by selecting one sample from each group.

[0118] [Table 1]

[0119] From the groups shown in Table 1, Nos. 5, 8-11, which had low solvent content, were selected as normal objects. A classification model was created for the solvent content of these normal objects by performing unsupervised learning based on a Gaussian mixing distribution. The obtained classification model was applied to the classification system of this disclosure (classification system 100 or classification system 100a), and objects Nos. 1-4, 6-7, and 12 were classified using this classification system. The results are shown in Figure 14.

[0120] In Figure 14, "Validation" refers to any object selected from the objects included in Nos. 5, 8-11. From Figure 14, it was successful to determine that objects Nos. 1-4, 6-7, and 12, which were selected as "Test" and had high solvent content, were all abnormal. Furthermore, objects with lower solvent content were plotted closer to the measurement results of normal objects (Train, Validation). Therefore, according to the classification system of one embodiment of the present invention, objects can be classified based only on the data corresponding to the odor of each object that meets predetermined conditions. [Explanation of Symbols]

[0121] 100, 100a, 100b, 100c classification system 2, 2a storage section 11 Acquisition Department 12 Classification section 13 Judgment section 30, 30a, 30b Odor measuring device 311 Odor sensor element BL boundary line L Production Line S1 Acquisition Steps S2 Classification Step

Claims

1. An acquisition unit acquires first data corresponding to the odor of an object of interest manufactured in at least one manufacturing process, which includes detection signals output from each of the one or more odor sensor elements of an odor measuring device. The manufacturing process includes a classification unit that classifies the object of interest using the first data based on the results of analyzing a plurality of second data sets, each of which corresponds to the smell of a plurality of objects manufactured in the manufacturing process, and which includes detection signals output from each of the odor sensor elements. The analysis is a statistical analysis, or an analysis based on feature quantities that indicate the degree of similarity between each of the multiple second data sets. The classification unit determines criteria for classifying the object of interest based on the results of the analysis. Classification system.

2. The aforementioned analysis is a multivariate analysis, The aforementioned classification unit is The multivariate analysis is performed on the aforementioned multiple second data sets to calculate the analysis results. Based on the results of the above analysis, a boundary line for classifying the object of interest or a function for classifying the object of interest is determined. The classification system according to claim 1.

3. The classification unit classifies the object of interest based on the first data using a classification model generated by unsupervised learning using the plurality of second data. The classification system according to claim 1.

4. The classification system according to claim 3, wherein the training data used for training to generate the classification model includes only the plurality of second data corresponding to the smell of each of the plurality of objects manufactured in the manufacturing process that satisfy predetermined conditions.

5. The classification unit classifies the object of interest based on the first data using a classification model. The classification model is generated by machine learning using training data that includes the plurality of second data as explanatory variables and the classification result of the object corresponding to each of the plurality of second data as the dependent variable. The classification system according to claim 1.

6. A storage unit that stores criteria for determining the quality of the object of interest based on the classification results of the classification unit, A determination unit that determines the quality of the object of interest from the classification result of the object of interest based on the aforementioned determination criteria, Furthermore, A classification system according to any one of claims 1 to 5.

7. A classification method performed by one or more computers, An acquisition step in which a computer acquires first data corresponding to the smell of an object of interest manufactured in at least one manufacturing process, the first data including detection signals output from each of one or more odor sensor elements of an odor measuring device, A computer classifies an object of interest using first data based on the results of analyzing a plurality of second data sets, each of which corresponds to the smell of a plurality of objects manufactured in the manufacturing process, and which includes detection signals output from each of the odor sensor elements. Classification method.

8. A control program for causing a computer to function as a classification system according to claim 1, wherein the computer functions as the acquisition unit and the classification unit.

9. A computer-readable recording medium that stores the control program described in claim 8.

Images