Electrochemical measurement device

Abstract

To provide an electrochemical measurement device that allows for preconditioning and calibration in as short a time as possible and exhibits sufficient measurement accuracy.SOLUTION: An electrochemical measurement device includes: multiple types of sensors that are immersed in a test liquid and output signals according to the electrochemical characteristics of the test liquid, each of which outputs signals according to different electrochemical characteristics that are correlated with each other; and a calculation unit that calculates a measurement value related to the electrochemical characteristics of the test liquid, which is a measurement value calculated using each signal output from at least two or more types of sensors among the sensors. The calculation unit calculates the measurement value using a machine learning model obtained based on training data regarding the measurement value.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an electrochemical measuring device. [Background technology]

[0002] It is known that in electrochemical measurement devices equipped with a sensor for measuring the electrochemical properties of a test liquid, the measurement values ​​are unstable and gradually change (drift) for a while after the sensor is immersed in the test liquid.

[0003] Therefore, in order to improve the measurement accuracy of an electrochemical measurement apparatus, it is recommended to perform preconditioning by immersing the sensor in the test liquid for several minutes to several hours in advance to stabilize the measurement value.

[0004] On the other hand, for example, for users who use an electrochemical measurement device for daily health management, rather than professionally trained technicians, the time spent on preconditioning is said to be only about one minute, which includes the time required to calibrate the display value of the electrochemical measurement device. Furthermore, depending on the application of the electrochemical measurement device, there may be cases where it is not possible to secure sufficient time for preconditioning and calibration in the first place.

[0005] Therefore, there is a demand for an electrochemical measurement apparatus that can achieve sufficient measurement accuracy with preconditioning and calibration in as short a time as possible.

[0006] As a method for shortening the time required for preconditioning and calibration compared to conventional methods, for example, Patent Document 1 discloses a method in which, in an electrochemical measurement device equipped with two types of ion selective electrodes as sensors, when a test liquid with known electrochemical properties is measured, the change per unit time (drift velocity) of the deviation (drift amount) over time of the measured potential difference (measured value) between the two types of sensors is examined, and a drift correction equation using the Nernst equation is set based on this drift velocity, and then the measured potential difference is corrected. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2014-95675 A Summary of the Invention [Problem to be solved by the invention]

[0008] However, the present inventors have found that there are cases where the measured value cannot be corrected with sufficient accuracy due to the presence of factors that cannot be corrected by the Nernst equation, such as the state of the sensor.

[0009] The present invention has been completed based on an idea completely different from the conventional method described in Patent Document 1, and has as its main object to provide an electrochemical measurement device that can achieve sufficient measurement accuracy while shortening the preconditioning and calibration time significantly compared to conventional methods, even when there are factors that cannot be corrected by the Nernst equation, such as the state of the sensor. [Means for solving the problem]

[0010] That is, the electrochemical measurement device according to the present invention is a sensor that is immersed in a test liquid and outputs a signal corresponding to an electrochemical property of the test liquid, the sensor including a plurality of types of sensors that each output a signal corresponding to a different electrochemical property that is correlated with each other; a calculation unit that calculates a measurement value related to the electrochemical property of the test liquid using signals output from at least two or more types of sensors among the sensors, The calculation unit is characterized in that it calculates the measurement value using a machine learning model obtained based on teacher data related to the measurement value.

[0011] According to an electrochemical measurement apparatus configured in this manner, it is possible to achieve sufficient measurement accuracy while shortening the preconditioning and calibration time as much as possible, regardless of the state of the sensor included in the electrochemical measurement apparatus. In addition, because the machine learning model improves each time training data is accumulated, another major advantage is that it can continue to optimize measurement accuracy, unlike the conventional drift correction formula using the Nernst equation.

[0012] A specific example of a factor that cannot be corrected by the Nernst equation is the electrochemical state change of the sensor, as described above. A typical factor that affects the electrochemical state change of the sensor is the elapsed time from immersing the sensor in the test liquid to the time of measurement (measurement time). Furthermore, the measured values ​​calculated by the electrochemical measurement apparatus are influenced not only by the electrochemical property to be measured, such as the ion concentration, but also by other changes in the electrochemical potential of the test liquid. A typical factor that affects the change in the electrochemical potential of the test liquid is the temperature of the test liquid at the time of measurement.

[0013] Therefore, by accumulating training data that includes not only measurement values ​​but also components that cause changes in the electrochemical potential of the test liquid other than the measurement target and changes in the electrochemical state of the sensor, and calculating measurement values ​​using a machine learning model obtained as a result of machine learning based on this data, sufficient measurement accuracy can be ensured even when the preconditioning and calibration times are shortened.

[0014] Furthermore, the aforementioned components contained in the training data (such as the temperature of the test liquid and the measurement time) are information that can be obtained regardless of the number or state of the sensor, and can be widely applied to electrochemical measurement devices that immerse a sensor in a test liquid to measure the electrochemical properties of the test liquid.

[0015] Incidentally, it is known that the deviation (drift) of the measured value is affected by the response speed of the sensor used. It is believed that the response speed of the sensor is affected by external factors that are not directly related to the electrochemical properties of the test liquid, such as the manufacturing process of the sensor and the number of years since its manufacture. Therefore, it is preferable to include external factors such as information regarding the manufacturing process of the sensor and information regarding the time from the manufacture of the sensor to the time of measurement in the components of the training data. This configuration makes it possible to eliminate the need for complex fitting work by specialized technicians, which was previously necessary to eliminate the effects of external factors on response speed, thereby significantly reducing costs and effort.

[0016] In a specific embodiment of the present invention, the sensor is an ion-selective electrode.

[0017] If the sensor is provided with two types of ion selective electrodes that respond to two different types of ions and a common electrode that measures the potential of the test liquid, and calculates the concentration ratio of the two types of ions based on the signals detected and output from the two types of ion selective electrodes and the signal output from the common electrode without setting a constant reference potential, the signals output from the two types of ion selective electrodes are highly symmetrical compared to when a constant reference potential is set using a reference electrode, making it easier to offset the drift of the measurement calculated based on the output signals from these electrodes. As a result, there is a possibility that the measured value can be corrected with higher accuracy.

[0018] A specific example of the two types of ions is a combination of sodium and potassium. Effect of the Invention

[0019] According to the present invention, even when there are factors that cannot be corrected using the Nernst equation, which have not been taken into account in the past, such as the state of a sensor provided in an electrochemical measurement device, the measurement value is calculated using a machine learning model that can include these as teacher data, so that sufficient measurement accuracy can be achieved while keeping the preconditioning and calibration time as short as possible. [Brief description of the drawings]

[0020] [Figure 1] 1 is an overall schematic diagram of an electrochemical measurement apparatus according to one embodiment of the present invention; [Diagram 2] FIG. 2 is a schematic cross-sectional view of a sensor unit of the electrochemical measurement device according to the embodiment. [Diagram 3] 5A to 5C are schematic diagrams showing the measurement procedure and the measurement results when a measurement value is calculated using a machine learning model by the electrochemical measurement apparatus according to the embodiment. [Figure 4] 1A to 1C are schematic diagrams showing the measurement procedure and the measurement results when the measurement value is calculated without using a machine learning model by the electrochemical measurement apparatus according to the embodiment. [Diagram 5] FIG. 13 is a schematic diagram showing a machine learning device according to another embodiment of the present invention. [Figure 6] FIG. 13 is a schematic diagram showing an apparatus for electrochemical measurements according to another embodiment of the present invention. [Figure 7] FIG. 11 is a schematic diagram showing a measurement result using an electrochemical measurement apparatus according to another embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0022] <Basic configuration of the electrochemical measurement device according to this embodiment> The electrochemical measurement device 100 of this embodiment measures, for example, the sodium ion / potassium ion concentration ratio, which is an electrochemical characteristic of a test liquid, and includes a sensor unit 1 equipped with a sensor 11 that contacts the test liquid and detects an electrochemical signal corresponding to the electrochemical characteristic of the test liquid, and an information processing unit 2 that receives the electrochemical signal output from the sensor unit 1, processes it, and calculates the sodium / potassium ion concentration ratio.

[0023] The electrochemical measurement apparatus 100 according to this embodiment physically comprises, for example, a measuring device A having a sensor unit 1 at its tip, and an information processing device B including a computer or the like configured to be able to communicate with the measuring device A via wired or wireless communication, as shown in FIG. Measuring device A according to this embodiment is provided with, for example, a display / operation unit 3 having a display unit 31 and an operation unit 32, and part of an information processing unit including a power supply unit and a microcomputer, not shown.

[0024] In this embodiment, the sensor unit 1 includes, for example, two types of ion selective electrodes 111 each responding to a different ion as the sensor 11, a common electrode 112 outputting a potential for calculating a concentration ratio of two types of ions based on the potentials detected at these two types of ion selective electrodes 111, and a temperature sensor (not shown). In this embodiment, the sensor unit 1 is a composite type in which the above-mentioned multiple sensors 11 are formed on the same substrate such that the surfaces of the sensors 11 in contact with the sample solution are arranged on approximately the same plane.

[0025] In this embodiment, the two types of ion selective electrodes 111 described above are a sodium ion selective electrode 111N and a potassium ion selective electrode 111K. The sodium ion selective electrode 111N, the potassium ion selective electrode 111K, and the common electrode 112 used in common as a counter electrode for these electrodes have a structure as shown in FIG. 2, for example. The sodium ion selective electrode 111N comprises a substrate 111Na, an internal electrode 111Nb made of, for example, a silver electrode formed on the substrate 111Na, an ion-sensitive membrane 111Nc attached to the internal electrode 111Nb, and an ion-electron conversion layer 111Nd attached between the internal electrode 111Nb and the ion-sensitive membrane 111Nc so as to be in contact with both, and disposed between the working electrode and the ion-responsive membrane to electrically connect them. The potassium ion selective electrode 111K comprises an internal electrode 111Kb, for example a silver electrode, formed on the substrate 111Ka, an ion-sensitive membrane 111Kc attached to the internal electrode 111Kb, and an ion-electron conversion layer 111Kd attached between the internal electrode 111Kb and the ion-sensitive membrane 111Kc so as to be in contact with both, and disposed between the working electrode and the ion-responsive membrane to electrically connect them.

[0026] The sodium ion sensitive film 111Nc is formed by adding a plasticizer and a sodium ionophore to polyvinyl chloride (PVC), dissolving the resultant in an organic solvent such as tetrahydrofuran (THF), applying the solution by potting or inkjet printing, and then heating to evaporate the organic solvent to form a solid sodium ion sensitive film 111Nc. An example of the sodium ionophore is Bis(12-crown-4). The solution for forming the sodium ion sensitive film 111Nc may contain polyvinyl chloride (PVC) containing a plasticizer, sodium ionophore, and an ionic additive such as a phenyl borate ionic additive.

[0027] The potassium ion sensitive film 111Kc is formed in the same manner as the sodium ion sensitive film 111Nc, except that a potassium ionophore is used, such as Bis (benzo-15-crown-5).

[0028] The ion-electron conversion layers 111Nd and 111Kd contain an ion-electron conversion material and an adhesive.

[0029] As the ion-electron conversion material, for example, a carbon microstructure can be used. The carbon microstructure may include, for example, one or more selected from carbon nanotubes, carbon nanofibers, carbon nanowalls, fullerenes, graphite, graphene, etc. In the present embodiment, carbon nanotubes are used as an example of the ion-electron conversion material. The content of the carbon microstructure in the ion-electron conversion layer 121c is preferably 0.001 mass % or more and 12.0 mass % or less, more preferably 0.001 mass % or more and 0.02 mass % or less, and even more preferably 0.003 mass % or more and 0.01 mass % or less. The adhesive preferably contains one or more selected from polyvinyl chloride (PVC), polystyrene, acrylate, polyvinyl butyral, polyamide, polyimide, polyurethane, polytetrafluoroethylene (PTFE), polysiloxane, copolymer of vinylidene fluoride and hexafluoropropylene (PVDF-HFP), and fluoropolysiloxane. These adhesives may be used alone or in combination. The adhesive may contain a plasticizer in addition to the above-mentioned components.

[0030] The common electrode 112 functions as a reference electrode for the sodium ion selective electrode 111N and the potassium ion selective electrode 111K, and includes a substrate 112a, an internal electrode 112b made of, for example, a silver electrode arranged on the substrate 112a, a salt bridge layer 112c attached on the internal electrode 112b, and an ion-electron conversion layer 112d arranged between the internal electrode 112b and the salt bridge layer 112c to electrically connect them. In this embodiment, the ion-electron conversion layer 112d has the same composition as the above-mentioned ion-electron conversion layers 111Nd and 111Kd. The compositions of the ion-electron conversion layers 111Nd, 111Kd, and 112d do not all need to be the same, and may be different from each other. The salt bridge layer 112c contains, for example, a hydrophobic ionic liquid and an adhesive. The hydrophobic ionic liquid may be any known one, for example, any one described in WO2021 / 177178A1, particularly P 444 MOEBETI (tributyl(2-methoxyethyl)phosphonium bis(pentafluoroethanesulfonyl)imide) and the like can be used. The adhesive preferably contains one or more selected from polyvinyl chloride (PVC), polystyrene, acrylate, polyvinyl butyral, polyamide, polyimide, polyurethane, polytetrafluoroethylene (PTFE), polysiloxane, copolymer of vinylidene fluoride and hexafluoropropylene (PVDF-HFP), and fluoropolysiloxane. These adhesives may be used alone or in combination. The adhesive may contain a plasticizer in addition to the above-mentioned components.

[0031] In this embodiment, the substrates 111Na, 111Ka, and 112a constituting the sodium ion selective electrode 111N, the potassium ion selective electrode 111K, and the common electrode 112 are one and the same substrate, but the present invention is not limited to this.

[0032] Each of the sensors 11 is electrically connected to an electric circuit board or the like that is built into the measuring device A and forms part of the information processing unit 2, for example, via lead wires or the like that are electrically connected to the internal electrodes 111Nb, 111Kb, 112b of each sensor 11, respectively.

[0033] The sensor unit 1 is formed at the tip of the measuring device A so as to surround from the side the entire area in which the surfaces of all the sensors 11 mentioned above that come into contact with the sample solution are formed, and is equipped with a test liquid holder 12 that holds the test liquid so that the test liquid comes into contact with the surfaces of all the sensors 11.

[0034] The information processing unit 2 includes an analog electric circuit having a buffer, an amplifier, etc., a digital electric circuit having a CPU, a memory, a DSP, etc., and an A / D converter and the like interposed therebetween.

[0035] The information processing unit 2 functions as a data acquisition section 21 that acquires signals output from each sensor by the CPU and its peripheral devices working together in accordance with a predetermined program stored in memory, and a calculation section 22 that calculates measurement values ​​such as potential difference and ion concentration ratio between the sensors 11 based on the acquired data.

[0036] <Basic Electrochemical Measurement Method Using the Electrochemical Measurement Device According to the Present Embodiment> The procedure for measuring the sodium ion / potassium ion concentration ratio in a sample solution using the electrochemical measurement device 100 configured in this manner is as follows. First, an appropriate amount of test liquid is dropped onto the test liquid holder 12 so that the test liquid comes into contact with the common electrode 112, the sodium ion sensitive film 111Nc, and the potassium ion sensitive film 111Kc.

[0037] Then, in the sodium ion sensitive membrane 111Nc and the potassium ion sensitive membrane 111Kc, an electromotive force corresponding to each ion concentration is generated between the test liquid and the ion-electron conversion layers 111Nd, 111Kd arranged in place of the internal liquid in the sodium ion sensitive membrane 111Nc and the potassium ion sensitive membrane 111Kc. This electromotive force is detected as a potential difference (voltage) between the internal electrode 111Nb of the sodium ion selective electrode 111N and the internal electrode 111Kb of the potassium ion selective electrode 111K and the potential of the test liquid detected by the internal electrode 112b of the common electrode 112, and the calculation section 22 included in the information processing unit 2 calculates these potential differences and the ion concentration ratio based on this potential difference based on a known method based on the detected signal.

[0038] In the electrochemical measurement method according to this embodiment, before performing a measurement using a sample solution with an unknown concentration of the ion to be measured as the test liquid, a standard solution, which is a buffer solution with a known concentration of the ion to be measured, is measured to calibrate the measurement value of the electrochemical measurement apparatus 100. After this calibration, actual measurements are carried out using a sample solution with an unknown ion concentration as a test liquid.

[0039] Normally, in order to obtain a reference potential, the common electrode 112 also requires an ion-electron conversion layer and an internal liquid. However, for example, by using the common electrode 112 having the internal electrode 112b made of Ag or the like as in this embodiment and by making the chloride ion activity of the standard solutions for zero calibration and span calibration the same, it is possible to make the ion-electron conversion layer and the internal liquid unnecessary components of the common electrode 112.

[0040] <Characteristic configuration of the electrochemical measurement device according to this embodiment> The information processing unit 2 in this embodiment further includes a memory unit 23 that stores and accumulates the teacher data received by the data acquisition unit 21, and a machine learning unit 24 that generates a machine learning model by performing machine learning based on the teacher data accumulated in the memory unit 23. The calculation unit 22 is configured to calculate the measurement value using the machine learning model generated by the machine learning unit 24.

[0041] In this embodiment, the measured value is the potential difference between two types of ion selective electrodes 111N, 111K. Therefore, for example, as shown in FIG. 3, first, during calibration when a standard solution is measured and during measurement when a sample solution is measured, the potential difference between the two ion selective electrodes is suppressed from fluctuating, thereby improving the calculation accuracy of the ion concentration ratio calculated for the test solution, which is the sample solution to be actually measured.

[0042] <Features of the measurement value calculation method according to this embodiment> The method in this embodiment in which the machine learning unit 24 generates the machine learning model will be described below. First, it is necessary to accumulate training data in the memory unit 23 for machine learning. In order to accumulate training data, a standard solution is prepared which is a buffer solution with known ion concentrations for sodium ions and potassium ions. The standard solution is then dropped onto the test liquid holder 12 as the test liquid so that the test liquid comes into contact with all of the dry sensor 11, and measurements are performed to calibrate the electrochemical measurement device 100 using a single-point calibration.

[0043] At this time, the time for calibration measurement after dropping the standard liquid is set to 60 seconds, and immediately after the 60 seconds have elapsed, each sensor 11 is rinsed with tap water for 5 seconds, and the tap water is immediately removed, after which the test liquid to be measured (if teacher data is to be obtained, a standard liquid with a known ion concentration) is dropped onto the test liquid holder 12 and the measurement is performed for 60 seconds. This series of operations is repeated for standard solutions of various concentrations within the measurement range of the electrochemical measurement apparatus 100 .

[0044] Furthermore, the temperature sensor provided in the sensor unit 1 is used to measure the temperature of the test liquid while the ion concentration ratio is being measured.

[0045] Then, the data acquisition unit 21 receives signals output from each sensor 11 that has come into contact with the test liquid, and the calculation unit 22 calculates a measurement value without using a machine learning model based on the signals from each sensor 11 received by the data acquisition unit 21. If the calculation unit 21 calculates a measurement value without using a machine learning model in this manner, as shown in Figure 4, it can be seen that the potential difference while the standard solution is being measured at the time of calibration changes over time, and it is difficult to stabilize it in the short time of 60 seconds, and the potential difference also fluctuates over time while the sample solution is being measured.

[0046] A set of data containing the measured value calculated in this manner (in this embodiment, the actual measured value of the potential difference between each sensor 11), the temperature of the test liquid at the time the measured value was obtained, and the elapsed time from dropping the standard liquid to the time the measured value was obtained (including the time spent rinsing the sensor with tap water) as components is stored and accumulated in memory unit 23 as teacher data. The above-mentioned elapsed time can be expressed by, for example, the following formula 1 when the sensor 11 is disposable or when a dried sensor 11 is used. (Elapsed time) = t1 (immersion time in standard solution) + t2 (rinsing time with tap water) + t3 (immersion time in sample solution until measurement value is obtained) (1) As shown in this formula 1, the elapsed time means the total time that the sensor 11 is in contact with liquids such as the standard solution, cleaning liquid such as tap water, and sample solution from the time the sensor 11 is immersed in the standard solution to the time a measured value is obtained. Here, immersing the sensor includes the case where the sensor unit 1 is immersed in a sample solution stored in a beaker or the case where liquid is supplied by dropping the liquid onto the surface of the sensor 11, and the surface of the sensor 11 is immersed in the liquid. The above-mentioned function as a timer is performed by the information processing device B, for example. For example, when the sensor 11 is immersed in a liquid such as a standard solution, a signal is output from the sensor 11 to the information processing device B. This signal is output periodically at preset intervals, for example, every second, while the sensor 11 is immersed in the liquid, and the information processing device B is configured to receive this signal and calculate the elapsed time from the number of signals received up to the point in time when a measured value that is output as teacher data together with the elapsed time is obtained. The immersion time in the standard solution (t1) refers to, for example, the time from the point when sensor 11 begins to output a signal as a result of being immersed in the standard solution to the point when a preset calibration time (e.g., 60 seconds) has elapsed after information processing device B is notified that the user has pressed the calibration button on measuring device A, or to the point when the deviation in the output value from sensor 11 falls within the acceptable range after information processing device B is notified that the user has pressed the calibration button on measuring device A. The point at which the measurement value is obtained means the point at which a preset measurement time (e.g., 60 seconds) has elapsed since the information processing device B was notified that the user had pressed the measurement button on measuring device A, or the point at which the deviation in the output value from sensor 11 has fallen within the acceptable range after the information processing device B was notified that the user had pressed the measurement button on measuring device A. When the sensor 11 is used for repeated measurements with its surface (ion-sensitive membrane) remaining wet, the total time from when the sensor is first immersed in the standard solution to when a measurement is obtained means the total time during which the sensor is rinsed with tap water or the like and immersed in an aqueous solution such as the standard solution or sample solution. These elapsed times may be counted automatically by a timer included in the electrochemical measurement device or a separately prepared timer. For example, when the sensor 11 is immersed in a liquid such as a standard solution, a signal is output from the sensor 11 to the information processing device B. This signal is output periodically at a preset interval, for example, every second, while the sensor 11 is immersed in the liquid or the surface of the sensor 11 is covered with the liquid, and the information processing device B is configured to receive this signal and calculate the elapsed time from the number of signals received up to the point in time when a measured value that is output as teacher data together with the elapsed time is obtained.

[0047] Next, based on the teacher data stored in the memory unit 23, the machine learning unit 24 generates a machine learning model that indicates how to correct the potential difference of each sensor 11 in order to minimize the fluctuation in the potential difference between each sensor 11.

[0048] It is expected that the accuracy of the machine learning model generated in this way will improve as more training data is accumulated, but it is preferable to conduct verification work to determine how much training data needs to be accumulated to achieve sufficient accuracy.

[0049] Therefore, in this embodiment, for example, when a certain amount of teacher data has been accumulated, the calculation unit uses a machine learning model to calculate measurement values ​​for standard solutions with known ion concentrations in exactly the same procedure, and if it is confirmed that the difference between the actual measured value and the theoretical value in the measurement value when using the machine learning model is within a predetermined range, it is determined that it is possible to measure a sample with unknown ion concentrations for each ion that is the actual measurement target.

[0050] This determination may be made by the person performing the checking operation, or the information processing unit 2 may further include a determination section 25, which may make the determination.

[0051] <Effects of this embodiment> The electrochemical measurements apparatus 100 according to this embodiment configured as above can provide the following advantages. The calculation unit 22 uses a machine learning model generated based on teacher data that contains not only the measurement value but also the temperature of the test liquid at the time the measurement value was obtained and the elapsed time from the start of measurement, which could not be used to correct the measurement value in the past.This makes it possible to minimize the error between the measurement value and the theoretical value while shortening the preconditioning and calibration time compared to the past.

[0052] By improving the structural symmetry of the two types of ion-selective electrodes 111N, 111K by differing only the ionophores used in the ion-sensitive membranes 111Nc, 111Kc, the reaction rates of these electrodes 111N, 111K can be made closer to each other, so that the fluctuations in potential difference can be reduced by correcting to offset the reaction rates, and the measurement accuracy of the ion concentration ratio by the electrochemical measurement device 100 can be further improved.

[0053] Since no internal liquid is required for the common electrode 112, and furthermore, solid ion-electron conversion layers 111Nd, 111Kd that perform the function of the internal liquid instead of an internal liquid are used for the ion selective electrodes 111N, 111K, a housing for holding the internal liquid is not required, and the sensor unit 1 can be made thinner than conventional ones.

[0054] In the above embodiment, the sodium ion / potassium ion concentration ratio is calculated, but it is also possible to calculate the sodium ion concentration and the potassium ion concentration separately. Moreover, for the internal electrode 112b of the common electrode 112, a wide variety of electrodes can be used, such as electrodes made of metals other than Ag electrodes, Ag / AgCl electrodes, Ag / AgBr electrodes, Ag / AgI electrodes, and the like.

[0055] <Other embodiments of the present invention> The present invention is not limited to the above-described embodiment. For example, in the above-described embodiment, an ion selective electrode was given as an example of a sensor equipped in an electrochemical measurement device, but the present invention is not limited to this. It is a sensor that can detect the electrochemical properties of a test liquid, and is known to have measurement values ​​that fluctuate for a while after being immersed in the test liquid, and can be widely applied to electrochemical measurement devices equipped with two or more types of sensors that output signals that are correlated with each other. In addition, the counter electrode, reference electrode, and the like used with these sensors can be appropriately selected depending on the purpose. Therefore, it goes without saying that the objects to be measured are not limited to the above-mentioned sodium ions and potassium ions. In addition, the number of sensors may be three, four, or more types, as long as there are two or more types of sensors.

[0056] In addition, measuring devices for measuring teacher data components other than electrochemical properties, such as temperature sensors and timers, may be provided in the measuring instrument, or independent measuring devices may be used. The components of the teacher data are not limited to those described above, and may further include, for example, one or more types selected from the group consisting of information regarding the manufacturing process of the sensor, information regarding the time from when the sensor was manufactured to the time of measurement, sample information, calibration information regarding the measurement value, and information regarding the measurement location where the measurement value was obtained.

[0057] Examples of information relating to the manufacturing process of the sensor and information relating to the time from the manufacture of the sensor to the time of measurement include the lot number of each sensor. Examples of sample information include the manufacturer, product number, lot number, and storage state of the test liquid. Examples of calibration information include information about when and how the most recent multi-point calibration was performed before measurements were obtained for the electrochemical geodesic device. Information about the measurement location may include, for example, the room temperature and humidity at the time of measurement.

[0058] In the above-mentioned embodiment, the potential difference has been described as an example of the measured value, but the measured value may be any value calculated by the calculation unit using the four basic arithmetic operations using multiple types of signals output simultaneously from multiple sensors included in the sensor unit, and may be, for example, an ion concentration ratio or an ion concentration in the above-mentioned example. Depending on the device configuration, the multiple types of signals do not need to be output simultaneously from multiple sensors, and may be, for example, signals measured separately for the same sample solution.

[0059] In the above-described embodiment, the machine learning model was used to improve the calculation accuracy of the measurement value, but it is also possible to use the machine learning model to detect abnormalities such as membrane rupture.

[0060] A part of the information processing unit may be, for example, a machine learning device composed of an independent server device or the like capable of communicating with multiple measuring devices via the Internet, and may collect teacher data from measuring devices used by an unspecified number of users, and distribute measurement values ​​or the machine learning model calculated using a machine learning model to each of the multiple measuring devices.

[0061] In this case, the machine learning device may, for example, be one that includes a data acquisition unit that accepts teacher data, a memory unit, and a machine learning unit (machine learning model generation unit), as shown in Figure 5, accumulates teacher data output from multiple measuring instruments, generates a machine learning model, and outputs the machine learning model generated for each measuring instrument.

[0062] In the above-mentioned embodiment, an electrochemical measurement device equipped with two types of ion selective electrodes and a common electrode common to these electrodes has been described, but an electrochemical measurement device that is widely used to measure the ion concentration in a sample solution is an ion concentration measurement device that is equipped with two types of sensors, an ion selective electrode and a reference electrode, and measures the ion concentration from the potential difference between these electrodes. In such a conventional ion concentration measurement device, the ion selective electrode and the reference electrode have completely different structures. For example, ion-selective electrodes with liquid membranes have a plasticized PVC membrane (the ion-sensitive membrane) to which a chemical (ionophore) sensitive to a particular ion has been added, while the reference electrode typically used with these electrodes has a porous salt bridge that generates a stable potential.

[0063] When these ion-selective electrodes and a reference electrode are immersed in a sample solution, the ion-selective electrode usually requires a preconditioning time of several minutes to several hours until the ion-sensitive membrane is hydrated and the potential becomes stable, whereas the reference electrode equipped with a porous salt bridge can stabilize its potential within a few seconds. Therefore, in an ion concentration measuring device equipped with a general ion selective electrode and a reference electrode, a preconditioning time of several minutes to several hours is required.

[0064] In such an ion concentration measuring device, one possible way to shorten the preconditioning time as much as possible is to make the structures of the ion selective electrode and the reference electrode, which are both immersed in the sample solution, as similar as possible, to improve the symmetry of the signals output from each of these electrodes, and to make the potential difference between these electrodes as constant as possible.

[0065] However, it is considered difficult to realize such an ion concentration measuring device because, in order to make the reference electrode closer to the structure of an ion-selective electrode, a solvent is needed that can simultaneously dissolve PVC and a hydrophilic salt such as KCl, which is necessary to form a salt bridge, but no such solvent has been known so far.

[0066] The present inventors have conceived of using a reference electrode that uses a salt bridge containing an ionic liquid as a means for solving such conventional problems.

[0067] Since ionic liquids can dissolve both hydrophilic compounds and hydrophobic polymers, salt bridges containing ionic liquids (also called ionic liquid salt bridges) can be produced using various polymer matrices. For example, ionic liquid salt bridges can be produced by mixing ionic liquids with PVC and plasticizers in a procedure similar to that used to produce ion-sensitive membranes for ion-selective electrodes. In addition, a method for producing a comparative electrode equipped with an ionic liquid salt bridge made by mixing ionic liquids with silicone rubber is described in WO2021 / 177178A1. Furthermore, since the ionic liquid salt bridge itself can function as a conventional internal liquid, there is no need to provide a separate internal liquid. Instead of providing an internal liquid, it is possible to incorporate a solid internal layer made of a carbon material with a relatively large surface area or a polymer with various new properties such as PEDOT:PSS into the reference electrode.

[0068] Specific examples of such combinations of an ion selective electrode and a reference electrode are described below, but it goes without saying that the combinations of an ion selective electrode and a reference electrode are not limited to these.

[0069] The ion selective electrode 111' can be produced, for example, as shown in Figures 6(a) and (b) by printing a silver electrode 111b' on a substrate 111a' made of PET, forming a solid internal layer C containing a carbon material such as carbon nanotubes on this silver electrode 111b', and forming an ion-sensitive membrane 111c' on this carbon layer C. Figure 6(a) is a view of the ion selective electrode 111' and the reference electrode R as viewed from the side in contact with the sample solution. Figure 6(b) is a cross-sectional view taken along the line A-A' in Figure 6(a). The ion-sensitive membrane 111c' can be formed, for example, by dropping onto the solid inner layer C a mixture of PVC premixed with a plasticizer, an ionophore, and an ionic additive in a solvent. As the plasticizer, for example, n-DOP can be used. As the ionophore, any compound can be used as long as it can selectively bind to ions in the sample solution. For example, 4-Nonadecylpyeidine for H + , BIS[(12-crown-4)methyl]dodecyl methylmalonate for Na+ or 4,5-Bis-[N'-(butyl)thioureido]-2,7-di-tert-butyl-9,9-dimethylxanthene for Cl - The ionic additive is not particularly limited as long as it assists the function of the ionophore so as to keep the charge of the ion-sensitive membrane 111c' neutral during measurement, but for example, potassium tetraphenylborate or the like can be used as a compound for the cation membrane, and tridodecyl(methyl)ammonium chloride or the like can be used as a compound for the anion membrane. For example, tetrahydrofuran (THF) or the like can be used as a solvent.

[0070] The comparison electrode R can be produced, for example, as shown in FIG. 6, by printing a silver electrode Rb having the same composition as the ion selective electrode 111′ on a substrate Ra made of PET, forming an internal layer C having the same composition as the internal layer C of the ion selective electrode 111′ on this silver electrode Rb, and forming an ionic liquid salt bridge Rc on this internal layer C. The ionic liquid salt bridge Rc can be produced, for example, by mixing PVC, which has been premixed with a plasticizer like that used for the ion-sensitive membrane 111c' of the ion selective electrode 111', with an ionic liquid in a solvent and dripping the mixture onto the solid inner layer (carbon base layer, C) to form an ionic liquid salt bridge. The hydrophobicity and resistance of the ionic liquid salt bridge can be adjusted to the same values ​​as those of the ion selective electrode 111' by changing the concentration of the ionic liquid.

[0071] By using such an ionic liquid salt bridge Rc, the ion-selective electrode 111' and the comparison electrode R can be constructed using the same polymer matrix and the same internal layer, as shown in Figure 6. By making the ion-selective electrode 111' and the comparison electrode R almost identical in structure in this way, when the ion-selective electrode 111' and the comparison electrode R are immersed in a sample, the ion-sensitive membrane 111c' and the salt bridge Rc, which are also based on PVC and have the same resistance value, can be hydrated at as close to the same pace as possible, and the degree of deterioration due to the number of times of use can also be made the same between the two electrodes, so that the potential difference measured by these electrodes 111' and R can be kept as constant as possible, as shown in Figure 7. Fig. 7(a) is a diagram showing the potential difference between the ion selective electrode 111' shown in Fig. 6 and a conventional comparison electrode having a porous salt bridge, measured when these electrodes are immersed in a sample solution. Fig. 7(b) is a diagram showing the potential difference between the comparison electrode R shown in Fig. 6 and a conventional comparison electrode, measured when these electrodes are immersed in a sample solution. Fig. 7(c) is a diagram showing the potential difference between the ion selective electrode 111' shown in Fig. 6 and the comparison electrode R, measured when these electrodes are immersed in a sample solution.

[0072] As described above, as shown in FIG. 6, an ion concentration measuring device equipped with an ion selective electrode 111′ and a comparison electrode R that are structurally symmetrical to each other can shorten the time for preconditioning and calibration while ensuring sufficient measurement accuracy, and is therefore considered to be particularly effective as a health and medical device used by users for daily health management at home, etc.

[0073] The ion selective electrode 111' and the reference electrode R, which use PVC as the base polymer, can have various functions in addition to those described above depending on the purpose, and can also be mass-produced. Therefore, it is possible that they can be applied to even more diverse fields in the future. Furthermore, since the type of polymer matrix used is not particularly limited, it is fully possible to use a base polymer other than PVC depending on the purpose.

[0074] Furthermore, by using a combination of the ion selective electrode 111' and the comparison electrode R, which are structurally symmetrical with each other as described above, as a sensor for an electrochemical measurement device that calculates measurement values ​​using the machine learning model of the present invention, it is expected that the time required for preconditioning and calibration can be further shortened and the measurement accuracy can be further improved.

[0075] In addition, some or all of the above-described embodiments and modified embodiments may be combined as appropriate, and it goes without saying that various modifications are possible without departing from the spirit of the invention. [Explanation of symbols]

[0076] 100 Electrochemical measurement device 1. Sensor unit 11 Sensors 2. Information Processing Unit 21 Calculation section 24 Machine Learning Department

Claims

1. a plurality of types of sensors that are immersed in a test liquid and output signals corresponding to electrochemical properties of the test liquid, each of which outputs a signal corresponding to a different electrochemical property that is correlated with each other; a calculation unit that calculates a measurement value related to the electrochemical property of the test liquid using signals output from at least two or more types of sensors among the sensors, The electrochemical measurement apparatus, characterized in that the calculation unit calculates the measurement value using a machine learning model obtained based on teacher data related to the measurement value.

2. The electrochemical measurement apparatus according to claim 1 , wherein the calculation unit calculates the measurement value by correcting the output value from each of the sensors using the machine learning model.

3. 2. The electrochemical measurement device of claim 1, wherein the teaching data includes measurement values ​​previously measured for various test liquids, the temperature of the test liquid at the time a signal for calculating the measurement value is acquired, and the elapsed time from when the sensor is immersed in the test liquid to when the signal is acquired.

4. 4. The electrochemical measurement apparatus according to claim 3, wherein the teacher data further includes at least one of information regarding a manufacturing process of the sensor, information regarding the time from manufacturing of the sensor to the time of measurement, sample information, calibration information regarding the measurement value, and information regarding the measurement location where the measurement value was obtained.

5. 2. The electrochemical measurement apparatus according to claim 1, further comprising a machine learning unit that generates a machine learning model based on training data including the measurement value, the temperature of the test liquid at the time when a signal for calculating the measurement value is acquired, and the elapsed time from when the sensor is immersed in the test liquid to when the signal is acquired.

6. 6. The electrochemical measurement apparatus according to claim 1, wherein at least one of the two types of sensors is an ion-selective electrode.

7. a common electrode for measuring the potential of the test liquid; 7. The electrochemical measurement device according to claim 6, wherein the two types of sensors are both ion selective electrodes, and the concentration ratio of the two types of ions is calculated based on signals detected and output from the two types of ion selective electrodes and a signal output from the common electrode.

8. 8. The electrochemical measurement device according to claim 7, wherein the two types of ion selective electrodes are a sodium selective electrode and a potassium selective electrode.

9. 1. An electrochemical measurement method for calculating a measurement value relating to the electrochemical properties of a test liquid based on signals output from at least two or more types of sensors among signals output from a plurality of types of sensors that are immersed in the test liquid and output signals corresponding to the electrochemical properties of the test liquid, the sensors respectively outputting signals corresponding to different electrochemical properties that are correlated with each other, the method comprising: An electrochemical measurement method characterized by calculating a measurement value related to the electrochemical property based on the signal output from the sensor using a machine learning model obtained based on training data including the measurement value, the temperature of the test liquid at the time a signal for calculating the measurement value was acquired, and the elapsed time from when the sensor was immersed in the test liquid to when the signal was acquired.

10. A program for calculating a measurement value relating to the electrochemical properties of a test liquid based on signals output from at least two or more types of sensors among signals corresponding to the electrochemical properties of the test liquid output from a plurality of types of sensors that are immersed in the test liquid and output signals corresponding to different electrochemical properties that are correlated with each other, the program comprising: An electrochemical measurement program characterized by causing a computer to function as a calculation unit that calculates a measurement value related to the electrochemical property based on the signal output from the sensor using a machine learning model obtained based on teacher data including the measurement value, the temperature of the test liquid at the time a signal for calculating the measurement value was acquired, and the elapsed time from when the sensor was immersed in the test liquid to when the signal was acquired.

11. A machine learning method used to calculate a measurement value relating to the electrochemical properties of a test liquid based on signals output from at least two or more types of sensors, each of which is a sensor immersed in a test liquid and outputs a signal corresponding to the electrochemical properties of the test liquid, the sensors outputting signals corresponding to different electrochemical properties that are correlated with each other, the method comprising: acquiring teacher data including the measurement value, the temperature of the test liquid at the time when the signal for calculating the measurement value was acquired, and the elapsed time from when the sensor was immersed in the test liquid to when the signal was acquired; A machine learning method for generating a machine learning model based on the acquired training data.

12. A program for a machine learning device used in an electrochemical measurement device that calculates a measurement value relating to the electrochemical properties of a test liquid based on signals output from at least two or more types of sensors, each of which is a sensor immersed in a test liquid and outputs a signal corresponding to the electrochemical properties of the test liquid, the sensors outputting signals corresponding to different electrochemical properties that are correlated with each other, the program comprising: a data acquisition unit that acquires teacher data including the measurement value, the temperature of the test liquid at the time when a signal for calculating the measurement value is acquired, and the elapsed time from when the sensor is immersed in the test liquid to when the signal is acquired; A machine learning program that causes a computer to function as a machine learning model generation unit that generates a machine learning model based on the training data acquired by the data acquisition unit.

Images