Electrolyte concentration measuring device and method for obtaining selectivity coefficient

Abstract

The present invention makes it possible to obtain a value corresponding to the selectivity coefficient of interfering ions or the concentration of the interfering ions without the need for a dedicated sample for calculating the selectivity coefficient of the interfering ions. Provided is an electrolyte concentration measuring device 100 for measuring the ion concentration of analyte ions contained in a liquid. The electrolyte concentration measuring device 100 comprises: ion-selective electrodes 111 to 113 to which the liquid is supplied; a comparative electrode 116 that provides a potential reference; and a control unit 120 that outputs the ion concentration of the analyte ions on the basis of the electromotive force between the comparative electrode 116 and the ion-selective electrodes 111 to 113. The control unit 120 inputs a measured potential of a solution containing the analyte ions and interfering ions, which affect the measurement of the concentration of the analyte ions, to a trained model, which outputs a value corresponding to the selectivity coefficient of the interfering ions or the concentration of the interfering ions on the basis of a measured potential of the solution containing the analyte ions and the interfering ions, and thus obtains a value corresponding to the selectivity coefficient of the interfering ions or the concentration of the interfering ions from the trained model.

Description

[Technical Field]

[0001] The present disclosure relates to an electrolyte concentration measuring device and a selectivity coefficient obtaining method, and more particularly to an electrolyte concentration measuring device and a selectivity coefficient obtaining method that are capable of calculating the selectivity of interfering ions. [Background technology]

[0002] Ion-selective electrodes are used in a wide range of fields, including water quality analysis and medicine, because they can selectively quantify the concentration of specific ions in a liquid. In particular, in the medical field, because there is a close relationship between metabolic reactions in living organisms and ion concentrations, quantifying specific ions contained in biological samples such as blood and urine can be used to diagnose conditions such as hypertension, kidney disease, and neurological disorders. Because clinical testing requires the continuous analysis of a large number of samples, high-throughput electrolyte concentration measurement devices equipped with ion-selective electrodes are routinely used.

[0003] Electrolyte concentration measuring devices primarily measure the concentrations of cations such as sodium ions and potassium ions, and anions such as chloride ions. Regarding cations, neutral carrier molecules such as crown ethers and valinomycin have been discovered that selectively capture specific cations but do not themselves carry an electric charge. Ion-sensitive membranes for cation-selective electrodes for sodium, potassium, and other ions generally contain these neutral carrier molecules, and these ion-sensitive membranes have high ion selectivity. On the other hand, for anions, there are no neutral carriers suitable for the ion-selective electrodes of electrolyte concentration measuring devices, which require high throughput and fast response, and ion-sensitive membranes of various configurations are therefore used.

[0004] In the case of an anion-selective electrode that uses a quaternary ammonium salt as a ligand, selectivity is determined by the degree of hydration of the ions, in the reverse order of the sequence called the elution sequence (see formula below). SCN - - <ClO3 - <NO3 - <Br - ​<Cl - <CH3COO - <SO4 2- <Tartaric acid <Citric acid

[0005] When measuring chloride ion concentration in biological samples such as serum or plasma, these samples often contain interfering ions other than the ions to be measured. Generally, human blood contains bicarbonate ions (HCO3), which account for approximately 25% of the chloride ion concentration. - ) and is the most highly concentrated of all interfering ions. For example, when measuring chloride ions, the selectivity for bicarbonate ions determines the impact on the measured chloride ion concentration. Changes in the selectivity of an anion selective electrode for bicarbonate ions over time and abnormalities in the ion selective electrode also affect the measured chloride ion concentration.

[0006] Patent Document 1 discloses a technology relating to an automatic analyzer that calculates the concentration of a target ion contained in a sample using the results of calculating a selectivity coefficient and the results of measuring the concentrations of coexisting ions contained in the sample, without adding any ion selective electrodes other than the ion selective electrode. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] International Publication No. 2019 / 163281 Summary of the Invention [Problem to be solved by the invention]

[0008] In Patent Document 1, in order to calculate the selectivity coefficient, a selectivity coefficient calculation sample 13 in which the sodium ion concentration and the coexisting ion (interfering ion) concentration are known is measured. Specifically, in Patent Document 1, two selectivity coefficient calculation samples 13 with different interfering ion concentrations are measured, and the selectivity coefficient of the interfering ion is calculated from these measurement results.

[0009] However, in Patent Document 1, it is necessary to prepare a dedicated selectivity coefficient calculation sample for calculating the selectivity coefficient of interfering ions and to measure the selectivity coefficient calculation sample.

[0010] Therefore, the present disclosure provides an electrolyte concentration measuring device and a selectivity coefficient acquisition method that can acquire a value corresponding to the selectivity coefficient of an interfering ion or the concentration of the interfering ion without requiring a dedicated sample for calculating the selectivity coefficient of the interfering ion. [Means for solving the problem]

[0011] The electrolyte concentration measuring device disclosed herein is an electrolyte concentration measuring device that measures the ion concentration of a target ion contained in a liquid, and includes an ion selective electrode that reacts to the target ion and outputs a potential corresponding to the ion concentration, a comparison electrode that outputs a reference potential that serves as a reference for the potential, and a control unit that outputs the ion concentration of the target ion based on the potential difference between the comparison electrode and the ion selective electrode.The control unit obtains the selectivity coefficient of the interfering ion or the value corresponding to the concentration of the interfering ion from the calculation unit by inputting the measured potential of the solution containing the target ion and the interfering ion to a calculation unit that outputs a value corresponding to the selectivity coefficient of the interfering ion or the concentration of the interfering ion from the measured potential of the solution containing the target ion and the interfering ion. [Effects of the Invention]

[0012] According to the present disclosure, a value corresponding to the selectivity coefficient of an interfering ion or the concentration of an interfering ion can be obtained without requiring a dedicated sample for calculating the selectivity coefficient of an interfering ion. [Brief explanation of the drawings]

[0013] [Figure 1A] 1 is a schematic diagram of an electrolyte concentration measuring device according to a first embodiment. [Figure 1B] FIG. 2 is a hardware block diagram of a computer system according to the first embodiment. [Figure 2]FIG. 1 is a schematic diagram showing a process of constructing a final model (learning phase) in Example 1 and a process of predicting an unknown result using the constructed final model (prediction phase). [Figure 3] 10 is a flowchart showing the flow of learning an initial model. [Figure 4] FIG. 1 illustrates multiple training data sets. [Figure 5] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the data set shown in FIG. 4 and applying multiple regression to an algorithm. [Figure 6] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the dataset shown in FIG. 4 and applying stochastic gradient descent regression to the algorithm. [Figure 7] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the data set shown in FIG. 4 and applying polynomial regression to an algorithm. [Figure 8] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the dataset shown in FIG. 4 and applying Ridge regression to the algorithm. [Figure 9] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the data set shown in FIG. 4 and applying Lasso regression to the algorithm. [Figure 10] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model for which machine learning was performed by applying Elastic Net regression to an algorithm using the data set shown in FIG. 4. [Figure 11] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the dataset shown in FIG. 4 and applying linear support vector regression to the algorithm. [Figure 12] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the data set shown in FIG. 4 and applying Gaussian kernel support vector regression to the algorithm. [Figure 13]FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model obtained by performing machine learning using the dataset shown in FIG. 4 and applying a decision tree to the algorithm. [Figure 14] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model in which machine learning was performed by applying a random forest to an algorithm using the dataset shown in FIG. 4. [Figure 15] FIG. 5 is a diagram showing a plot of test data and predicted values ​​for a model in which machine learning was performed using the dataset shown in FIG. 4 and XGBoost was applied to the algorithm. [Figure 16] FIG. 5 is a diagram showing a plot of test data and predicted values ​​of a model obtained by performing machine learning using the dataset shown in FIG. 4 and applying a multilayer perceptron to the algorithm. [Figure 17] 4 is a flowchart showing a deterioration determination process according to the first embodiment. [Figure 18] 10 is a flowchart showing a maintenance support process in response to a membrane contamination alarm in the first embodiment. [Figure 19] 10 is a flowchart showing the correction process for the measurement values ​​of the control sample in Example 2. DETAILED DESCRIPTION OF THE INVENTION

[0014] The embodiments of the present invention will be described in detail with reference to the drawings. In the following embodiments, it goes without saying that the components (including element steps, etc.) are not necessarily essential unless otherwise specified or considered to be obviously essential in principle.

[0015] Example 1 1A is a schematic diagram of an electrolyte concentration measurement device according to a first embodiment. The electrolyte concentration measurement device 100 according to the first embodiment is a device that measures the concentration of a target ion contained in a liquid sample. The configuration of the electrolyte concentration measurement device 100 will be described below.

[0016] (Electrolyte concentration measuring device 100) A sample container 101 contains a biological sample (hereinafter referred to as the sample) such as blood or urine. A sample dispensing nozzle 102 is immersed in the sample contained in the sample container 101. By operation of a sample dispensing nozzle syringe 103, the sample dispensing nozzle 102 aspirates a set amount of sample from the sample container 101 and dispenses it into a dilution tank 104. A dilution liquid bottle 105 contains a dilution liquid used to dilute the sample. The dilution liquid is sent to the dilution tank 104 by operation of a dilution liquid syringe 106 and a dilution liquid solenoid valve 107. The sample in the dilution tank 104 is diluted with the dilution liquid.

[0017] The sample diluted in the dilution tank 104 is supplied to a sodium ion selective electrode 111, a potassium ion selective electrode 112, and a chloride ion selective electrode 113 by operation of a sipper syringe 108, a sipper syringe solenoid valve 109, and a pinch valve 110. The sodium ion selective electrode 111 reacts to sodium ions and outputs a potential corresponding to the sodium ion concentration. The potassium ion selective electrode 112 reacts to potassium ions and outputs a potential corresponding to the potassium ion concentration. The chloride ion selective electrode 113 reacts to chloride ions and outputs a potential corresponding to the chloride ion concentration. The reference electrode solution contained in a reference electrode solution bottle 114 is supplied to a reference electrode 116 by operation of a reference electrode solution solenoid valve 115, a sipper syringe 108, and a sipper syringe solenoid valve 109. The reference electrode 116 outputs a reference potential that serves as a reference for the potentials output by each ion selective electrode. The control unit 120 acquires each electromotive force (each potential difference) between each of the ion selective electrodes 111, 112, and 113 to which the diluted sample has been supplied and the comparison electrode 116. Note that hereinafter, when simply referred to as electromotive force, this refers to the electromotive force between each of the ion selective electrodes 111, 112, and 113 and the comparison electrode 116.

[0018] In measuring the internal standard solution used to determine the concentration of a sample, the internal standard solution contained in the internal standard solution bottle 117 is sent to the dilution tank 104, from which the sample and dilution solution have been removed, by operation of the internal standard solution syringe 118 and the internal standard solution solenoid valve 119. The internal standard solution in the dilution tank 104 is supplied to the sodium ion selective electrode 111, the potassium ion selective electrode 112, and the chloride ion selective electrode 113 by operation of the sipper syringe 108, the sipper syringe solenoid valve 109, and the pinch valve 110. The reference electrode solution contained in the reference electrode solution bottle 114 is supplied to the reference electrode 116 by operation of the reference electrode solution solenoid valve 115, the sipper syringe 108, and the sipper syringe solenoid valve 109. The control unit 120 acquires the electromotive force between each of the ion selective electrodes 111, 112, and 113 to which the internal standard solution has been supplied and the reference electrode 116.

[0019] The sodium ion selective electrode 111, the potassium ion selective electrode 112, the chloride ion selective electrode 113, and the reference electrode 116 are connected to a control unit 120. The control unit 120 controls the overall operation of the electrolyte concentration measuring device 100, acquiring electromotive forces and controlling the operation of the syringes 103, 106, 108, and 118 and the solenoid valves 107, 109, 115, and 119. A storage unit 121, a display unit 122, and an input unit 123 are connected to the control unit 120. A user inputs various parameters and information about the sample to be measured (such as sample type information) via the input unit 123 based on a setting screen displayed on the display unit 122. The storage unit 121 stores the input information. The storage unit 121 also stores various programs used in measuring the sample, measurement results, and the like. A trained model that outputs a value corresponding to the selectivity coefficient of the interfering ions or the concentration of the interfering ions from the measured potential of a solution containing the target ions and the interfering ions measured by calibration, which will be described later, is stored in this storage unit 121. In other words, the trained model functions as a calculation unit that outputs a value corresponding to the selectivity coefficient of the interfering ions or the concentration of the interfering ions from the measured potential of a solution containing the target ions and the interfering ions.

[0020] (Computer Systems) 1B is a hardware block diagram of a computer system of the control unit of Example 1. A computer system 200 of the control unit 120 executes each process of the flowcharts described below. The computer system 200 has a processor 201, a main memory unit 202, an auxiliary memory unit 203, an input / output interface (hereinafter, interface will be abbreviated as I / F) 204, a communication I / F 205, and a bus 206 that communicatively connects the above-mentioned modules.

[0021] The processor 201 is a central processing unit that controls the operation of each unit communicatively connected to the control unit 120. The processor 201 is, for example, a central processing unit (CPU), a digital signal processor (DSP), or an application specific integrated circuit (ASIC). The processor 201 deploys a program (e.g., a deterioration determination program) stored in the storage unit 121 in an executable manner in a working area of ​​the main storage unit 202 and executes the program. The main storage unit 202 stores the program executed by the processor 201, data processed by the processor, and the like. The main storage unit 202 is, for example, a flash memory, a random access memory (RAM), or the like. The auxiliary storage unit 203 is, for example, a read-only memory (ROM), and stores a boot program and various setting values ​​of the control unit 120. The storage unit 121 stores an OS, a deterioration determination processing program (described later), and the like. The storage unit 121 is a silicon disk including a nonvolatile semiconductor memory (flash memory, EPROM (Erasable Programmable ROM)), a solid state drive device, a hard disk drive (HDD), or the like.

[0022] Input / output I / F 204 is communicably connected to input devices (e.g., input unit 123 and storage unit 121) and output devices (e.g., display unit 122 and storage unit 121). Communication I / F 205 is an interface for communicably connecting electrolyte concentration measuring device 100 to external devices via a network.

[0023] (calibration) A method for calculating the slope sensitivity SL through calibration will be described below.

[0024] Known low concentration (concentration C L ) into the dilution tank 104, and then the dilution liquid in the dilution liquid bottle 105 is further dispensed into the dilution tank 104 using the dilution liquid syringe 106. In this way, the standard solution (L) of known low concentration is diluted at a set ratio. Next, the sipper syringe 108 is operated to introduce the diluted standard solution (L) of known low concentration in the dilution tank 104 into the flow path of the ion selective electrodes 111-113. Furthermore, the reference electrode solution is introduced from the reference electrode solution bottle 114 into the flow path of the reference electrode 116. This brings the reference electrode solution into contact with the diluted standard solution (L) of known low concentration. Thereafter, the potential difference between the ion selective electrodes 111-113 and the reference electrode 116 (the electromotive force EMF of the standard solution (L) of known low concentration) L ) is measured by the control unit 120.

[0025] Next, the internal standard solution in the internal standard solution bottle 117 is newly dispensed into the dilution tank 104 using the internal standard solution syringe 118. Next, the sipper syringe 108 is operated to introduce the internal standard solution in the dilution tank 104 into the flow paths of the ion selective electrodes 111 to 113. Furthermore, the reference electrode solution is introduced from the reference electrode solution bottle 114 into the flow path of the reference electrode 116. As a result, the reference electrode solution and the internal standard solution come into contact with each other, and the potential differences between the ion selective electrodes 111 to 113 and the reference electrode 116 (the electromotive force EMF of the internal standard solution) increase, just as in the case of a known low concentration standard solution (L). IS ) is measured by the control unit 120.

[0026] Next, a known high concentration (C H) into the dilution tank 104, and then the dilution liquid in the dilution liquid bottle 105 is further dispensed into the dilution tank 104 using the dilution liquid syringe 106. In this way, the standard solution (H) of known high concentration is diluted at a set ratio. Next, the sipper syringe 108 is operated to introduce the diluted standard solution (H) of known high concentration in the dilution tank 104 into the flow path of the ion selective electrodes 111-113. Furthermore, the reference electrode solution is introduced from the reference electrode solution bottle 114 into the flow path of the reference electrode 116. This brings the reference electrode solution into contact with the diluted standard solution (H) of known high concentration. Thereafter, the potential difference between the ion selective electrodes 111-113 and the reference electrode 116 (the electromotive force EMF of the standard solution (H) of known high concentration) H ) is measured by the control unit 120.

[0027] As described above, the electromotive force EMF measured by the control unit 120 H and EMF L The slope sensitivity SL of the calibration curve is calculated using the following formula (1). SL=(EMF H -EMF L ) / (LogC H -LogC L )...Equation (1) Through the above process, calibration is performed.

[0028] The slope sensitivity SL corresponds to 2.303×(RT / zF) in the following equation (2) (Nernst equation). E = E0 + 2.303 × (RT / zF) × Log(f × C) Equation (2) E: potential of the selected electrode E0: Constant potential determined by the measurement system z: Charge number of the ion to be measured F: Faraday constant R: gas constant T: absolute temperature f: Activity coefficient C: Ion concentration

[0029] Concentration of internal standard solution C is is the measured electromotive force EMF of the internal standard solution ISTherefore, it can be calculated by the following equations (3) and (4). C is =C L x10 a ...Equation (3) a=(EMF IS -EMF L ) / SL...Formula (4) Although the specific calibration procedure has been described above, the calibration procedure is not limited to the above. In the above example, the measurement of a standard solution of known low concentration is performed after the measurement of a standard solution of known high concentration, but this order may be reversed.

[0030] (Measurement of sample liquid) The same procedure is followed when measuring a sample liquid. After the above calibration is performed, the sample liquid is measured. In the measurement of the sample liquid, the measurement of the sample liquid and the measurement of the internal standard solution are performed alternately and continuously.

[0031] After calibration, analysis is performed using serum, urine, or the like as a sample. Specifically, a sample dispensing nozzle (not shown) is used to dispense sample liquid into dilution tank 104, and then a diluent in diluent bottle 105 is dispensed into dilution tank 104 using diluent syringe 106, diluting the sample at a set ratio to produce diluted sample liquid. Next, sipper syringe 108 is operated to introduce the diluted sample liquid in dilution tank 104 into the flow paths of ion selective electrodes 111-113. Furthermore, reference electrode liquid is introduced from reference electrode liquid bottle 114 into the flow path of reference electrode 116. This brings the reference electrode liquid into contact with the diluted sample liquid. Thereafter, the potential differences between ion selective electrodes 111-113 and reference electrode 116 (electromotive force EMF of the diluted sample liquid) are calculated. S ) is measured by the control unit 120.

[0032] While the electromotive force of the diluted specimen solution is being measured, the diluted specimen solution remaining in the dilution tank 104 is discarded into a waste tank. Thereafter, the internal standard solution in the internal standard solution bottle 117 is newly dispensed into the dilution tank 104 using the internal standard solution syringe 118. Next, the sipper syringe 108 is operated to introduce the internal standard solution in the dilution tank 104 into the flow paths of the ion selective electrodes 111 to 113. Furthermore, the reference electrode solution is introduced from the reference electrode solution bottle 114 into the flow path of the reference electrode 116. As a result, the reference electrode solution and the internal standard solution come into contact with each other, and the potential differences between the ion selective electrodes 111 to 113 and the reference electrode 116 (the electromotive force EMF of the internal standard solution) are increased. IS ) is measured by the control unit 120.

[0033] Slope sensitivity SL and concentration of internal standard solution C is From the above, the concentration of the sample C is calculated using the following equations (5) and (6). s Calculate. C s =C is x10 b ...Equation (5) b=(EMF IS -EMF S ) / SL...Formula (6)

[0034] The above formula is a basic one. In the device of this embodiment, a measurement of a constant concentration internal standard solution is also performed before and after the measurement of the diluted sample solution. The measurement value of the diluted sample solution is then corrected based on the measurement value of this internal standard solution. This allows accurate measurement to be achieved even if gradual potential fluctuations (potential drift phenomenon) occur due to changes in the surface of the sensitive membrane of the ion selective electrodes 111-113 or temperature changes.

[0035] FIG. 2 is a schematic diagram showing the process of constructing a final model (learning phase) in Example 1 and the process of predicting an unknown result using the constructed final model (prediction phase).

[0036] (Learning phase) In the training phase, machine learning is performed on the initial model using multiple training datasets selected from a library of collected data, and the final model (trained model) is constructed.

[0037] First, an initial model is set (step S201). Then, the initial model is trained using multiple training data sets selected from the library (step S202), and a final model (hereinafter, the final model will be referred to as the trained model as appropriate) is constructed (step S203). Input parameters are set in the initial model. The parameter settings affect the properties of the model. In the training process, calculations are performed to gradually adjust internal parameters such as the model's internal weighting coefficients so as to reduce as much as possible the risk that the trained model will give an incorrect answer when a measurement value from the electrolyte concentration measuring device 100 is input.

[0038] (Prediction phase) If the final model makes a prediction using target data other than the training data (step S204) and obtains a good predicted value, it can be said that the model functions well (step S205).

[0039] A supervised learning dataset is used in the learning in Example 1. This is a method in which a dataset consisting of a set of input items and correct values ​​for those input items is provided to the model as learning data, and the model learns rules for deriving correct answers.

[0040] FIG. 3 is a flowchart showing the flow of learning the initial model.

[0041] Step 1: The entire learning dataset is divided into training data and test data. The training data is used in Step 5, and the test data is used in Step 6. A cross-validation method is applied, using different data for training and evaluating the predictive model. The division ratio of training data to test data is set to 7:3 or 8:2 (Step S301).

[0042] Step 2: Standardization: When features are mixed, the different features are standardized to have the same scale and a uniform standard. Standardization makes the average value of the features zero and the standard deviation 1 (step S302).

[0043] Step 3: Specify an initial model. For the initial model, a mathematical formula for the initial model is prepared for each algorithm, and the preconditions for the model are determined. The output of the initial model is the calculation result of the input and parameters (step S303).

[0044] Step 4: Specify the loss function. The loss function is a function that evaluates the error and is used in learning the model for that stage. The definition of the error and the formula of the loss function differ for each algorithm, and the loss function is a function that determines the assumptions of the algorithm in combination with the initial model (step S304).

[0045] Step 5: Model training involves calculating the error of the training data using the initial model and loss function, and calculating the parameters that minimize the loss function. The loss function compares the initial model output with the correct value and calculates the error. The parameters are updated to minimize this error (step S305).

[0046] Step 6: Model evaluation uses the parameters obtained in step 5 and the test data to output predicted values ​​for the test data, compare the predicted values ​​with the correct values, and evaluate the model (step S306).

[0047] (How to determine the selectivity coefficient of a general interfering ion) Here, an example of how to determine the selectivity coefficient of a general interfering ion in an ion selective electrode will be described. The target ion concentration (C j ) to obtain the target ion concentration (C iThe selectivity coefficient can be calculated by subtracting the target ion concentration (C) from the target ion concentration (see equation (7) below) and dividing by the known target ion concentration (C). The target ion concentrations in the first solution and the second solution are set to the same ion concentration (C). This method is called the mixed solution method because it uses a solution containing the target ion and interfering ions. Selectivity coefficient of interfering ions = (C j -C i ) / C...Equation (7)

[0048] FIG. 4 shows multiple learning data sets. Part of 1000 learning data sets is shown in FIG. 4. The learning data set in FIG. 4 has the measured potential (EMF) of three types of solutions with the same bicarbonate ion concentration but different chloride ion concentrations as a feature. IS、 EMF L and EMF H ) and the bicarbonate ion selectivity coefficient (K Cl,HCO3 ) or the value corresponding to the bicarbonate ion concentration (hereinafter referred to as the predicted potential) (EMF HCO3 ) and the value corresponding to the bicarbonate ion concentration (EMF HCO3 ) is the measured potential of a solution with a bicarbonate ion concentration different from the three solutions measured to determine the selectivity coefficient, and the bicarbonate ion selectivity coefficient (K Cl,HCO3 ) is a value calculated from the measured potential. The final model optimized using the multiple learning data sets shown in FIG. 4 was applied with the measured potential (EMF) of three types of solutions with the same bicarbonate ion concentration and different chloride ion concentrations measured by the electrolyte concentration measuring device 100. IS、 EMF L and EMF H ) to calculate the selectivity coefficient. HCO3 ) or predicted bicarbonate selectivity coefficient (K Cl,HCO3 ) is output.

[0049] The predicted potential (EMF) for determining the selectivity coefficient mentioned above HCO3 ) is selected as the output, the potential (EMF IS、 EMF L and EMFH ) and predicted potential (EMF HCO3 ) to determine the bicarbonate ion selectivity coefficient (K Cl,HCO3 ) can be obtained.

[0050] As described above, the electrolyte concentration measuring device 100 of Example 1 measures three types of solutions with the same interfering ion concentration but different target ion concentrations in calibration. Therefore, in Example 1, the measured potentials of the three types of solutions measured in this calibration are input into the final model, and the bicarbonate ion selectivity coefficient (K Cl,HCO3 ) or the value corresponding to the bicarbonate ion concentration (EMF HCO3 ) is output.

[0051] Conventionally, it has been difficult to determine the selectivity coefficient of an anion selective electrode after its shipment, but in this embodiment, the interfering ion selectivity coefficient can be determined in real time, such as at the timing of calibration, making it possible to grasp the state of the anion selective electrode and improving the reliability of the estimated value.

[0052] (Regarding the correlation between input and output of the final model) Below are the learning results of various algorithms using the multiple learning datasets shown in Figure 4. The closer the coefficient of determination, which indicates the correlation between the test data and its predicted value, to 1, the better the performance (higher the correlation) of the algorithm.

[0053] (Model 1) Algorithm: Multiple Regression Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 5(a) and (b) show plots of the test data and the predicted values. Table 1 shows the mean squared error and the coefficient of determination.

[0054] [Table 1]

[0055] (Model 2) Algorithm: Stochastic Gradient Descent Regression Parameters: Learning rate 0.01 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 6(a) and (b) show plots of the test data and the predicted values. Table 2 shows the mean squared error and the coefficient of determination.

[0056] [Table 2]

[0057] (Model 3) Algorithm: Polynomial regression Parameters: 3rd degree polynomial Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 7(a) and (b) show plots of the test data and the predicted values. Table 3 shows the mean squared error and the coefficient of determination.

[0058] [Table 3]

[0059] (Model 4) Algorithm: Ridge regression Parameters: 3rd degree polynomial, L2 regularization α=0.0001 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 8(a) and (b) show plots of the test data and the predicted values. Table 4 shows the mean squared error and the coefficient of determination.

[0060] [Table 4]

[0061] (Model 5) Algorithm: Lasso regression Parameters: 3rd degree polynomial, L1 regularization α=0.01 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 9(a) and (b) show plots of the test data and the predicted values. Table 5 shows the mean squared error and the coefficient of determination.

[0062] [Table 5]

[0063] (Model 6) Algorithm: Elastic Net Regression Parameters: 3rd degree polynomial, L1+L2 regularization α=0.001, l1_ratio=0.3 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 10(a) and (b) show plots of the test data and the predicted values. Table 6 shows the mean squared error and the coefficient of determination.

[0064] [Table 6]

[0065] (Model 7) Algorithm: Linear Support Vector Regression Parameters: Linear kernel, C=1.0, ε=0.3 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 11(a) and (b) show plots of the test data and the predicted values. Table 7 shows the mean squared error and the coefficient of determination.

[0066] [Table 7]

[0067] (Model 8) Algorithm: Gaussian Kernel Support Vector Regression Parameters: Gaussian kernel, C=1.0, ε=0.3, γ=001 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMFL、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 12(a) and (b) show plots of the test data and the predicted values. Table 8 shows the mean squared error and the coefficient of determination.

[0068] [Table 8]

[0069] (Model 9) Algorithm: Decision Tree Parameter: max_depth=8 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 13(a) and (b) show plots of the test data and the predicted values. Table 9 shows the mean squared error and the coefficient of determination.

[0070] [Table 9]

[0071] (Model 10) Algorithm: Random Forest Parameters: bootstrap, n_estimators=1000, max_depth=None Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 14(a) and (b) show plots of the test data and the predicted values. Table 10 shows the mean squared error and the coefficient of determination.

[0072] [Table 10]

[0073] (Model 11) Algorithm: XGBoost Parameter: learning_rate=1 Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 15(a) and (b) show plots of the test data and the predicted values. Table 11 shows the mean squared error and the coefficient of determination.

[0074] [Table 11]

[0075] (Model 12) Algorithm: Multilayer Perceptron Parameters: hidden layer 2, activation function = relu, output layer 1, optimizer = sgd Number of datasets: 1000 Training data / test data split ratio: 8:2 Feature: EMF IS、 EMF L、 EMF H Correct answer: EMF HCO3 or K Cl,HCO3 16(a) and (b) show plots of the test data and the predicted values. Table 12 shows the mean squared error and the coefficient of determination.

[0076] [Table 12]

[0077] As a result of machine learning using the various algorithms described above, a high coefficient of determination of 0.831 or higher was obtained. The above results indicate that there is a correlation between the measured potential of multiple solutions (three types in this example) with the same bicarbonate ion concentration and the bicarbonate ion selectivity coefficient or a value corresponding to the bicarbonate ion concentration. Furthermore, it was found that, regardless of the machine learning algorithm used, the bicarbonate ion selectivity coefficient can be estimated with high accuracy from the measured potential of multiple solutions with the same bicarbonate ion concentration.

[0078] In this embodiment, the function (trained model) obtained from the stochastic gradient descent regression, which is the algorithm that results in a coefficient of determination closest to 1, is recorded in the storage unit 121.

[0079] Note that the mean sum of squares error and coefficient of determination obtained will differ depending on the combination of the training dataset and batch size used for training. Furthermore, the mean sum of squares error and coefficient of determination obtained will also differ depending on the combination of parameters set in the algorithm. Therefore, the setting conditions for the machine learning algorithm, etc. are not limited to those described above.

[0080] (1) Deterioration determination process Fig. 17 is a flowchart showing the deterioration determination process of Example 1. The deterioration determination process determines the deterioration state of the anion selective electrode based on a prediction result using machine learning. This deterioration determination process will be described with reference to Fig. 17. A deterioration determination program that causes the computer system 200 to execute the deterioration determination process of Fig. 17 is stored in the storage unit 121. The processor 201 reads out and executes the deterioration determination program stored in the storage unit 121.

[0081] The following factors are assumed to be causes of deterioration, and by comparing the bicarbonate ion selectivity coefficient output from the final model with the reference value according to the flowchart in Figure 17, it is possible to identify the assumed causes of deterioration. 1. Membrane contamination 2. Peeling of membrane

[0082] In general, the performance of an anion selective electrode gradually deteriorates as time passes from the date of manufacture. Furthermore, electrode performance also deteriorates as the frequency of use, such as in measuring sample liquids, increases. The electrolyte concentration measuring device 100 incorporating an anion selective electrode is equipped with a mechanism for generating an alarm in response to fluctuations in the measured potential and slope sensitivity, but is not equipped with a function for generating an alarm in response to fluctuations in the interfering ion selectivity coefficient. Therefore, it is difficult to predict problems due to fluctuations in the interfering ion selectivity coefficient. In Example 1, the interfering ion selectivity coefficient can be calculated at the time of calibration or at any time, so the interfering ion selectivity coefficient can be calculated in real time. This makes it possible to immediately identify problems due to fluctuations in the interfering ion selectivity coefficient, which is expected to significantly improve the reliability of measurement data.

[0083] In the first embodiment, the interfering ion selectivity coefficient of the anion selective electrode is automatically calculated during calibration, and the deterioration determination process is performed based on the calculated interfering ion selectivity coefficient. Even when calibration is not performed, the control unit 120 may calculate the interfering ion selectivity coefficient and perform the deterioration determination process when an instruction to start the deterioration determination process is input by an operator or at a preset start time of the deterioration determination process.

[0084] First, the computer system 200 controls the anion selective electrode to be calibrated (step S401). This calibration allows the measurement results (e.g., EMF IS、 EMF L and EMF H) can be obtained. Next, the computer system 200 inputs the calibration measurement results into the final model and obtains the selectivity coefficient of bicarbonate ions from the final model (step S402). Note that the value obtained in step S402 may be a value corresponding to the bicarbonate ion concentration.

[0085] As a result, the selectivity coefficient of bicarbonate ion can be obtained from the final model. That is, the method for obtaining the selectivity coefficient of interfering ions is as follows: Calibration allows the measurement of three solutions with the same bicarbonate ion concentration but different chloride ion concentrations, and the measurement results (e.g., EMF IS、 EMF L and EMF H ) to obtain The final model is fitted with measurements (e.g., EMF IS、 EMF L and EMF H ), and obtaining a selectivity coefficient for the interfering ion or a value corresponding to the concentration of the interfering ion from the final model.

[0086] Next, the computer system 200 determines whether the obtained selectivity coefficient for bicarbonate ions is equal to or greater than a reference value (step S403). If the selectivity coefficient for bicarbonate ions is less than the reference value (step S403: No), the process ends. If the selectivity coefficient for bicarbonate ions is equal to or greater than the reference value (step S403: Yes), the computer system 200 determines whether the slope sensitivity is outside a predetermined range (step S404). If the computer system 200 determines that the slope sensitivity is not outside the predetermined range (step S404: No), it issues a membrane fouling alarm (step S405). If the computer system 200 determines that the slope sensitivity is outside the predetermined range (step S404: Yes), it issues a membrane peeling alarm (step S406).

[0087] Even when no fluctuations are observed in the measured potential or slope sensitivity of the anion-selective electrode, the selectivity coefficient for bicarbonate ion may fluctuate. If the selectivity for bicarbonate ion fluctuates, the measured value for chloride ion will not be accurate unless calibration is performed. In Example 1, the selectivity coefficient for bicarbonate ion can be predicted, so a membrane fouling alarm is triggered when the selectivity coefficient for bicarbonate ion fluctuates, even when no fluctuations are observed in the slope sensitivity. Cleaning the membrane improves the selectivity coefficient for bicarbonate ion, making it possible to immediately identify problems caused by fluctuations in the selectivity coefficient for bicarbonate ion, and is expected to significantly improve the reliability of the measurement data.

[0088] (2) Maintenance support processing The procedure for processing when any of the degradation factors is reported in the degradation determination processing of FIG. 17 will be described.

[0089] (2-1) Maintenance support process for membrane fouling alarms 18 is a process flow showing a maintenance support process corresponding to a membrane fouling alarm. When the computer system 200 issues a membrane fouling alarm, it displays on the display unit 122 a message urging the operator to clean the flow path (step S501). Next, the computer system 200 determines whether the cleaning has been completed (step S502). Specifically, the computer system 200 displays on the display unit 122 a screen for confirming whether the cleaning has been completed, and if the operator inputs information indicating that the cleaning has been completed via the input unit 123 such as a keyboard, it determines that the cleaning has been completed; otherwise, it determines that the cleaning has not been completed.

[0090] If it is determined that the cleaning is completed (step S502: Yes), the computer system 200 controls to execute calibration (step S503). This calibration allows the measurement results (for example, EMF IS、 EMF L and EMF H) can be obtained. Next, the computer system 200 inputs the calibration measurement results into the final model and obtains the selectivity coefficient for bicarbonate ion from the final model (step S504).

[0091] Next, the computer system 200 determines whether the obtained selectivity coefficient for bicarbonate ions is equal to or greater than a reference value (step S505). If the selectivity coefficient for bicarbonate ions is equal to or greater than the reference value (step S505: Yes), the computer system 200 issues an electrode failure alarm (step S506) and terminates the process. On the other hand, if the selectivity coefficient for bicarbonate ions is less than the reference value (step S505: No), the computer system 200 determines that the membrane has been cleaned by cleaning, issues a message indicating that there is no abnormality, and terminates the process of this flowchart.

[0092] (2-2) Maintenance support process for membrane peeling alarm As described above, the electrolyte concentration measuring device 100 incorporating an anion selective electrode is equipped with a mechanism for issuing alarms in response to fluctuations in the measured potential and slope sensitivity. When membrane peeling occurs, these alarms are likely to be issued, but there is a possibility that the alarm will not be issued if the fluctuation in slope sensitivity is small. In Example 1, by predicting the interfering ion selectivity coefficient, it is possible to detect a malfunction earlier than the issuance of an alarm for the measured potential or slope sensitivity, which is expected to significantly improve reliability. When a membrane peeling alarm is issued, the computer system 200 issues an electrode malfunction alarm and terminates processing.

[0093] Next, the measured values ​​(EMF IS、 EMF L and EMF H ) into the final model, the output (electromotive force of bicarbonate ions or selectivity coefficient of bicarbonate ions) obtained is compared with the reference value. Here, four examples, Cases A to D, are explained.

[0094] (Case A) In case A, the following measurements (EMF IS、 EMF L and EMF H ) into the function (trained model) of the algorithm recorded in the storage unit 121 to obtain the predicted value EMF HCO3 Next, the selectivity coefficient K for bicarbonate ion was calculated using the following equation (8). Cl,HCO3 was calculated and compared with the standard value. K Cl,HCO3 =10 (EMFHCO3-EMFIS) / SL-1 ...Equation (8)

[0095] [Table 13]

[0096] As a result, the selectivity coefficient for bicarbonate ions was 0.35, which was 0.25 higher than the reference value of 0.1. Since no decrease in slope sensitivity SL was observed, a membrane fouling alarm was issued according to the flowchart in Figure 17. Then, a screen prompting the operator to clean the flow path was displayed on the display unit 122.

[0097] (Case B) In case B, the following measurements (EMF IS、 EMF L and EMF H ) into the function of the algorithm recorded in the memory unit 121 to obtain the selectivity coefficient K Cl,HCO3 was calculated and compared with the standard value.

[0098] [Table 14]

[0099] As a result, the selectivity coefficient for bicarbonate ions was 0.30, which was 0.20 greater than the reference value of 0.1. Since no decrease in slope sensitivity SL was observed, a membrane fouling alarm was issued according to the flowchart in Figure 17. Then, a screen prompting the operator to clean the flow path was displayed on the display unit 122.

[0100] (Case C) In case C, the following measurements (EMF IS、 EMF L and EMF H ) into the function of the algorithm stored in the storage unit 121 to obtain the predicted value EMF HCO3 Next, the selectivity coefficient K for bicarbonate ion was calculated using equation (8). Cl,HCO3 was calculated and compared with the standard value.

[0101] [Table 15]

[0102] As a result, the bicarbonate ion selectivity coefficient was 0.41, which was 0.31 greater than the reference value of 0.1. In addition, a decrease in slope sensitivity SL was observed, so a membrane peeling alarm was issued according to the flowchart in Figure 17. Then, a screen indicating an electrode failure alarm was displayed on the display unit 122.

[0103] (Case D) In case D, the following measurements (EMF IS、 EMF L and EMF H ) into the function of the algorithm recorded in the memory unit 121 to obtain the selectivity coefficient K Cl,HCO3 was calculated and compared with the standard value.

[0104] [Table 16]

[0105] As a result, the bicarbonate ion selectivity coefficient was 0.45, which was 0.35 greater than the reference value of 0.1. In addition, a decrease in slope sensitivity SL was observed, so a membrane peeling alarm was issued according to the flowchart in Figure 17. Then, a screen indicating an electrode failure alarm was displayed on the display unit 122.

[0106] In the above cases A to D, the reference value is set to 0.1, but the reference value is not limited to 0.1. The reference value may be set arbitrarily by the user.

[0107] In the flow of FIG. 17, the cause of the abnormality is identified by a combination of comparing the selectivity coefficient of bicarbonate ions with a reference value (S403 in FIG. 17) and comparing the slope sensitivity with a predetermined range (S404 in FIG. 17). An abnormality may be detected only by comparing the selectivity coefficient of bicarbonate ions with a reference value when the selectivity coefficient of bicarbonate ions is equal to or less than the reference value. In other words, the presence or absence of an abnormality may be determined by executing the processes of S401 to S403 in the flow chart of FIG. 17. Identifying the cause of the abnormality allows for more appropriate response to the abnormality, but determining only the presence or absence of an abnormality is feasible even when the processing capacity of the control unit 120 is low.

[0108] (Effects of Example 1) By inputting the measured potentials of three types of solutions containing chloride ions and bicarbonate ions into the final model, the selectivity coefficient for bicarbonate ions or a value corresponding to the concentration of bicarbonate ions can be obtained from the final model. In this way, in Example 1, a dedicated sample for calculating the selectivity coefficient for interfering ions, as in Patent Document 1, is not required, and it is possible to output the selectivity coefficient for bicarbonate ions or a value corresponding to the concentration of bicarbonate ions.

[0109] By training the initial model with a supervised training dataset, the selectivity coefficient for bicarbonate ion or a value corresponding to the concentration of bicarbonate ion can be obtained from the final model by inputting the measured potentials of three solutions containing chloride ion and bicarbonate ion into the final model.

[0110] By providing the electrolyte concentration measuring device 100 with a memory unit 121 that stores a trained model, it is possible to obtain the selectivity coefficient of bicarbonate ions or a value corresponding to the concentration of bicarbonate ions in a local device.

[0111] <Example 2> Electrolyte concentration measuring devices periodically measure control samples for quality control. If the bicarbonate ion concentration in the control sample differs from the bicarbonate ion concentration in the internal standard solution, high-concentration calibration standard solution, and low-concentration calibration standard solution, the displayed value during control sample measurement will differ from the package insert value. When a discrepancy occurs between the control sample and the package insert value, no decrease in the slope sensitivity of the anion selective electrode is observed. Furthermore, even if the interfering ion selectivity coefficient appears normal, the anion selective electrode's quality control may be deemed inadequate. In Example 2, the interfering ion selectivity coefficient can be predicted, allowing for correction of measured values ​​even when the bicarbonate ion concentration in the control sample used differs from the bicarbonate ion concentrations in the internal standard solution, high-concentration calibration standard solution, and low-concentration calibration standard solution. This is expected to significantly improve the reliability of the measurement data.

[0112] To correct this deviation, the computer system 200 executes the following process in accordance with the flowchart of FIG.

[0113] The operator inputs the concentration value of bicarbonate ions contained in the control sample using the input unit 123 or the like (step S601). Then, the computer system 200 executes calibration (step S602) and calculates the measured value (EMF IS、 EMF L and EMF H ) is obtained. The computer system 200 inputs the calibration measurement values ​​shown in Table 17 into the function of the algorithm stored in the storage unit 121 to obtain the predicted value EMF HCO3 Next, the bicarbonate ion selectivity coefficient K was calculated using equation (8). Cl,HCO3 was calculated (step S603).

[0114] [Table 17]

[0115] As a result, the bicarbonate ion selectivity coefficient was 0.10. Next, the computer system 200 calculates the deviation from the package insert value of the control sample by multiplying the difference between the bicarbonate ion concentrations contained in the internal standard solution, the high-concentration standard solution for calibration, and the low-concentration standard solution and the bicarbonate ion concentration contained in the control sample by the calculated bicarbonate ion selectivity coefficient (step S604). The control sample is then measured, and a corrected value is obtained (step S605).

[0116] Example 3 In the second embodiment, the predicted value EMF HCO3 In Example 3, the selectivity coefficient K Cl,HCO3 Ask for.

[0117] The operator inputs the concentration value of bicarbonate ions contained in the control sample using the input unit 123 or the like (step S601). Then, the computer system 200 executes calibration (step S602) and calculates the measured value (EMF IS、 EMF L and EMF H Then, the computer system 200 inputs the calibration measurement values ​​shown in Table 18 into the function of the algorithm recorded in the memory unit 121 to obtain the bicarbonate ion selectivity coefficient K Cl,HCO3 is calculated (step S603).

[0118] [Table 18]

[0119] As a result, the bicarbonate ion selectivity coefficient was 0.12. Next, the computer system 200 calculates the deviation from the package insert value of the control sample by multiplying the difference between the bicarbonate ion concentrations contained in the internal standard solution, the high-concentration standard solution for calibration, and the low-concentration standard solution and the bicarbonate ion concentration contained in the control sample by the calculated bicarbonate ion selectivity coefficient (step S604). The control sample is then measured, and a corrected value is obtained (step S605).

[0120] The present invention is not limited to the above-described embodiments and includes various modifications. The above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment, or to add the configuration of another embodiment to the configuration of one embodiment. Furthermore, it is also possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

[0121] In Example 1, the measured potentials of three solutions with the same bicarbonate ion concentration were used as the training data set, but the measured potentials of two solutions with the same bicarbonate ion concentration may be used as the training data set to train the model, or the measured potentials of four or more solutions may be used as the training data set to train the model. Note that the more types of solutions whose measured potentials are used for training, the higher the accuracy will be.

[0122] In addition, in Example 1, the measured potentials of multiple solutions with the same bicarbonate ion concentration were used as the training data set, but the measured potentials of multiple solutions with different bicarbonate ion concentrations may also be used as the training data set to train the model. This training makes it possible to obtain values ​​corresponding to the selectivity coefficients of interfering ions and the concentrations of interfering ions from the measured potentials of multiple solutions with different bicarbonate ion concentrations.

[0123] In Example 1, the training data set was a combination of the measured potentials of multiple solutions and the correct values, but the training data set may be a combination of the measured potentials of a single solution containing bicarbonate ions and the correct values. By learning using this training data set, it becomes possible to obtain values ​​corresponding to the selectivity coefficients of interfering ions and the concentrations of interfering ions from the measured values ​​of a single solution.

[0124] Furthermore, in the first embodiment, the trained model is stored in the storage unit 121, but the trained model may also be stored in an on-premise server or a cloud server that is communicatively connected to the electrolyte concentration measuring device 100. In this case, the measurement values ​​measured by the electrolyte concentration measuring device 100 are transmitted via a communication line to the server that stores the trained model, and the selectivity coefficient of the interfering ions and values ​​corresponding to the concentrations of the interfering ions are obtained as return values.

[0125] Furthermore, in Example 1, the trained model output the selectivity coefficient of the interfering ion and a value for the concentration of the interfering ion, but the present invention may not use the trained model, but may instead select a value that is approximate to the measurement value measured by the electrolyte concentration measuring device 100 from the data shown in Figure 4, and obtain the selectivity coefficient associated with the selected data. [Explanation of symbols]

[0126] 100: electrolyte concentration measuring device, 101: sample container, 102: sample dispensing nozzle, 103: syringe for sample dispensing nozzle, 104: dilution tank, 105: dilution solution bottle, 106: dilution solution syringe, 107: dilution solution solenoid valve, 108: sipper syringe, 109: solenoid valve for sipper syringe, 110: pinch valve, 111: sodium ion selective electrode, 112: potassium ion selective electrode, 113: chloride ion selective electrode, 114 : Reference electrode solution bottle, 115: Reference electrode solution solenoid valve, 116: Reference electrode, 117: Internal standard solution bottle, 118: Internal standard solution syringe, 119: Internal standard solution solenoid valve, 120: Control unit, 121: Memory unit, 122: Display unit, 123: Input unit, 200: Computer system, 201: Processor, 202: Main memory unit, 203: Auxiliary memory unit, 204: Input / output interface, 205: Communication interface, 206: Bus

Claims

1. An electrolyte concentration measurement device for measuring the ion concentration of a target ion contained in a liquid, an ion selective electrode that reacts to the target ion and outputs a potential corresponding to the ion concentration; and a comparison electrode that outputs a reference potential that serves as a reference for the potential. a control unit that outputs the ion concentration of the target ion based on the potential difference between the reference electrode and the ion selective electrode, the control unit acquires the selectivity coefficients of the interfering ions from a calculation unit that has been trained to output selectivity coefficients of the interfering ions from measured potentials of a plurality of solutions containing the target ions and the interfering ions, the plurality of solutions containing the target ions and the interfering ions, the plurality of solutions having the same concentration of the interfering ions and different concentrations of the target ions; The selectivity coefficient of the interfering ions output by the calculation unit is compared with a reference value, and a slope sensitivity calculated from the measured potential of the solution containing the target ion and the interfering ions is compared with a predetermined range, and an alarm is issued based on the comparison result between the selectivity coefficient of the interfering ions and the reference value and the comparison result between the slope sensitivity and the predetermined range. An electrolyte concentration measuring device characterized by:

2. 2. The electrolyte concentration measuring device according to claim 1, An electrolyte concentration measuring device characterized in that the calculation unit is a trained model trained using multiple supervised learning data sets that are sets of measured potentials of the multiple solutions and selectivity coefficients of the interfering ions that are correct values ​​for combinations of the measured potentials.

3. 2. The electrolyte concentration measuring device according to claim 1, The electrolyte concentration measuring device further comprises the calculation unit.

4. 2. The electrolyte concentration measuring device according to claim 1, The control unit performs calibration to determine slope sensitivity, and acquires the selectivity coefficient of the interfering ions from the calculation unit by inputting the measured potentials of the plurality of solutions measured in the calibration to the calculation unit.

5. 2. The electrolyte concentration measuring device according to claim 1, The control unit corrects the measurement value of a control sample measured for accuracy control of the electrolyte concentration measuring device using the selectivity coefficient of the interfering ions obtained from the calculation unit.

6. A method for acquiring a selectivity coefficient of an interfering ion that affects the measurement of the concentration of a target ion contained in a liquid, comprising: measuring the measured potential of a plurality of solutions containing the target ions and the interfering ions, the solutions having the same concentration of the interfering ions and different concentrations of the target ions; a calculation unit that has been trained to output the selectivity coefficient of the interfering ions from the measured potentials of a plurality of solutions containing the target ions and the interfering ions, the solutions having the same concentration of the interfering ions but different concentrations of the target ions; acquiring a selectivity coefficient of the interfering ions from the calculation unit; comparing the selectivity coefficient of the interfering ions output by the calculation unit with a reference value; comparing a slope sensitivity calculated from the measured potential of the solution containing the target ion and the interfering ion with a predetermined range; and issuing an alarm based on a comparison result between the selectivity coefficient of the interfering ions and the reference value and a comparison result between the slope sensitivity and the predetermined range; have A method for obtaining a selection coefficient.

7. The selection coefficient acquisition method according to claim 6, A method for acquiring a selectivity coefficient, further comprising: training the calculation unit with a plurality of supervised learning data sets each including a set of the measured potentials of the plurality of solutions and the selectivity coefficients of the interfering ions that are correct values ​​for the combinations of the measured potentials.

8. The selection coefficient acquisition method according to claim 6, A method for acquiring a selectivity coefficient, characterized in that measuring the measured potentials of the plurality of solutions is performed during calibration to determine slope sensitivity.

9. The selection coefficient acquisition method according to claim 6, A method for acquiring a selectivity coefficient, further comprising correcting the measurement value of a control sample measured for quality control using the selectivity coefficient of the interfering ion acquired from the calculation unit.

Images