Analysis abnormality diagnosis system and analysis abnormality diagnosis method

Abstract

The present invention is provided with: a learning model holding unit (8) that holds a learning model created using learning model creation data, the learning model creation data includes a plurality of parameters of analysis data acquired by normal analysis, and a plurality of parameters of each of a plurality of analysis data acquired by analysis in a state in which a plurality of types of abnormalities are individually present or mixed. The learning model trends to learn the influence of each of the plurality of anomalies on the plurality of parameters; an analysis data holding unit (10) that holds reference analysis data obtained by normally analyzing a reference sample containing a specific component under specific analysis conditions, and analysis data to be diagnosed; the analysis data of the diagnosis object is obtained by analyzing a sample containing at least the specific component under the specific analysis conditions; and an abnormality estimation unit (12) configured to apply the plurality of parameters of each of the reference analysis data and the analysis data to be diagnosed to the learning model, thereby estimating an abnormality type existing in an analysis performed when the analysis data to be diagnosed is acquired.

Description

Technical Field

[0001] This invention relates to an analytical anomaly diagnosis system and method, used to infer the type of anomaly from analytical data obtained by analytical devices such as liquid chromatographs. Background Technology

[0002] When using analytical instruments such as liquid chromatographs, various anomalies may occur that affect the analytical data. Among these anomalies, some can be automatically detected during the analysis process, while others can only be discovered when unexpected analytical data is obtained during actual analysis, and these anomalies require time to verify and determine their cause.

[0003] For example, column degradation or blockage can be automatically detected by comparing the values ​​of instruments such as pressure sensors with preset thresholds (see, for example, Patent Document 1). On the other hand, the concentration of the target component in the sample can change due to incorrect stock solution or improper instrument operation. Furthermore, the sample injection volume can change not only due to the deterioration of the autosampler leading to decreased accuracy of the aspiration volume, but also due to incorrect analytical condition settings by the operator. These anomalies all manifest as changes in peak area values ​​in the chromatogram, but even if an anomaly in the peak area value is known, it is difficult to identify what caused the anomaly.

[0004] Furthermore, if the concentration of the standard sample is incorrect due to factors such as using the wrong stock solution, the calibration curve made using that standard sample will also be inaccurate. Therefore, the target component concentration calculated using the calibration curve will differ from the actual concentration. When both the sample and the standard sample use the same stock solution, it is difficult for the user to know that the sample concentration differs from the specifications.

[0005] Existing technical documents

[0006] Patent documents

[0007] Patent Document 1: International Publication No. 2020 / 170378 Summary of the Invention

[0008] Technical problems to be solved

[0009] As mentioned above, even if anomalies are found during analysis, it is often difficult to identify the cause of the anomaly simply by confirming the analyzed data. There are problems such as difficulty in detecting the existence of anomalies or the need to spend a lot of labor and time to determine what kind of anomaly has occurred.

[0010] Therefore, the purpose of this invention is to be able to accurately deduce what kind of anomalies occurred during the analysis from the analytical data.

[0011] Methods for solving problems

[0012] The first analytical anomaly diagnostic system according to the present invention comprises:

[0013] The learning model maintenance unit maintains the learning model, which is created using monthly data. The data used to create the learning model includes multiple parameters of analytical data obtained through normal analysis, as well as the multiple parameters of multiple analytical data obtained through analysis under the condition that multiple anomalies exist individually or in a mixed manner. The learning model learns the influence of each of the multiple anomalies on the multiple parameters in a trend-based manner.

[0014] An analytical data holding unit holds baseline analytical data and analytical data for diagnostic purposes. The baseline analytical data is obtained by normal analysis of a reference sample containing a specific component under specific analytical conditions. The analytical data for diagnostic purposes is obtained by analyzing an analytical sample containing at least the specific component under the same specific analytical conditions.

[0015] An anomaly estimation unit is configured to apply the respective plurality of parameters of the baseline analysis data and the analysis data as the diagnostic object held in the analysis data holding unit to the learning model held in the learning model holding unit, thereby estimating the types of anomalies present during the analysis performed to obtain the analysis data as the diagnostic object.

[0016] Furthermore, the second analytical anomaly diagnostic system according to the present invention includes:

[0017] The learning model maintenance unit maintains the learning model, which is created using learning model creation data. The learning model creation data includes parameters of multiple analysis data obtained by continuously executing multiple normal analyses, and parameters of multiple analysis data obtained by continuously executing multiple analyses under the condition that multiple anomalies exist individually or in a mixed manner. The learning model learns the influence of each of the multiple anomalies on the parameters of the multiple analysis data in a trend-based manner.

[0018] An analytical data holding unit holds multiple analytical data points used for diagnosis, said multiple analytical data points being obtained by performing multiple analyses on the same sample; and

[0019] An anomaly estimation unit is configured to apply the parameters of each of the multiple analytical data as diagnostic objects held in the analytical data holding unit to the learning model held in the learning model holding unit, thereby estimating the types of anomalies present during the multiple analyses performed to obtain the multiple analytical data as diagnostic objects.

[0020] Furthermore, the first analytical anomaly diagnosis method according to the present invention includes:

[0021] The production steps involve using multiple parameters from analytical data obtained through normal analysis, as well as the multiple parameters from multiple analytical data obtained through analysis under various abnormalities, either individually or in a mixed state, to create a learning model that trend-learns the influence of each of the various abnormalities on the multiple parameters.

[0022] The preparation steps include preparing baseline analysis data and diagnostic analysis data. The baseline analysis data is obtained by analyzing a reference sample containing a specific component under normal conditions within specific analytical conditions. The diagnostic analysis data is obtained by analyzing a sample containing at least the specific component under the same specific analytical conditions.

[0023] The estimation step involves applying the respective plurality of parameters of the baseline analysis data and the analysis data as the diagnostic object to the learning model after the production step and the preparation step, thereby estimating the types of anomalies present during the analysis performed to obtain the analysis data as the diagnostic object.

[0024] Furthermore, the second analytical anomaly diagnosis method according to the present invention includes:

[0025] The production steps involve using the parameters of multiple analytical data obtained by continuously executing multiple normal analyses, as well as the parameters of multiple analytical data obtained by continuously executing multiple analyses under the conditions of multiple anomalies existing alone or mixed, to create a learning model that trend-learns the influence of each of the multiple anomalies on the parameters of the multiple analytical data.

[0026] Preparation steps include preparing multiple analytical data sets for the diagnostic object, wherein the multiple analytical data sets are obtained by performing multiple analyses on the same sample; and

[0027] The estimation step involves applying the parameters of each of the multiple analytical data points used as diagnostic objects to the learning model, thereby estimating the types of anomalies present during the multiple analyses performed to obtain the multiple analytical data points used as diagnostic objects.

[0028] That is, the present invention uses machine learning algorithms to create a learning model. This learning model learns what kind of trend influence the parameters of the analysis data have when performing analysis under abnormal conditions. The created learning model is used to analyze how the parameters of the analysis data as the diagnostic object change relative to the parameters of the analysis data under normal conditions, thereby inferring what kind of abnormality exists in the analysis when obtaining the analysis data as the diagnostic object.

[0029] Invention Effects

[0030] According to the first analytical anomaly diagnostic system of the present invention, since a learning model that has learned the influence of each of a variety of anomalies on multiple parameters of the analytical data in a trend-based manner is maintained, and the multiple parameters of the benchmark analytical data obtained by normally acquiring a benchmark sample containing a specific component under specific analytical conditions and the analytical data as the diagnostic object obtained by analyzing a sample containing the specific component under said specific analytical conditions are applied to the maintained learning model, the types of anomalies present during the analysis performed in order to acquire the analytical data as the diagnostic object can be estimated. Therefore, it is possible to estimate with high accuracy what kind of anomaly occurred during the analysis based on the analytical data.

[0031] According to the second analytical anomaly diagnostic system of the present invention, since a learning model is maintained that learns the influence of each of a variety of anomalies on the parameters of multiple analytical data obtained from multiple consecutive analyses, and the parameters of each of the multiple analytical data obtained by performing multiple analyses on a sample as diagnostic objects are applied to the maintained learning model, the types of anomalies present in the multiple analyses performed to obtain the multiple analytical data as diagnostic objects can be estimated with high precision based on the analytical data.

[0032] According to the first analytical anomaly diagnosis method of the present invention, since a learning model that has been prepared in advance to learn the influence of each of a variety of anomalies on multiple parameters of the analytical data in a trend-based manner is prepared, and the multiple parameters of the baseline analytical data obtained through normal analysis and the multiple parameters of the analytical data as the diagnostic object obtained by analyzing the same sample as the sample for which the baseline analytical data was obtained under the same analytical conditions are applied to the prepared learning model, the types of anomalies present during the analysis performed in order to obtain the analytical data as the diagnostic object can be estimated. Therefore, it is possible to estimate with high accuracy what kind of anomaly occurred during the analysis based on the analytical data.

[0033] According to the second analytical anomaly diagnosis method of the present invention, since a learning model that has been prepared in advance to learn the influence of each of a variety of anomalies on the parameters of multiple analytical data obtained from multiple consecutive analyses is prepared, and the parameters of the multiple analytical data obtained by performing multiple analyses on a sample as diagnostic objects are applied to the prepared learning model, the types of anomalies present in the multiple analyses performed to obtain the multiple analytical data as diagnostic objects can be estimated. Therefore, it is possible to estimate with high accuracy what kind of anomaly occurred in the multiple consecutive analyses based on the analytical data. Attached Figure Description

[0034] [ Figure 1 This diagram provides a summary view of an embodiment of an analysis anomaly diagnosis system.

[0035] [ Figure 2 [A concept diagram used to illustrate an example of creating a learning model.]

[0036] [ Figure 3 [A conceptual diagram used to illustrate an example of a presumed anomaly.]

[0037] [ Figure 4 A graph used to illustrate the effect of sample concentration on the chromatogram.

[0038] [ Figure 5 This is a graph used to illustrate the effect of sample injection volume on the chromatogram.

[0039] [ Figure 6 A distribution of chromatographic peak retention times obtained under conditions of leakage.

[0040] [ Figure 7 The distribution of peak symmetry coefficients in a chromatogram obtained under conditions where dead volume exists in the piping connection section.

[0041] [ Figure 8 This is a flowchart showing an example of the steps to infer anomalies from analyzed data.

[0042] [ Figure 9 This figure illustrates an example of the effect of air bubbles being introduced into the sample aspiration path on the peak area values ​​of the chromatograms obtained in each analysis during sequential analysis of the same sample.

[0043] [ Figure 10 This figure illustrates an example of the effect of an excess of sample (more than a specified amount) on the peak area values ​​of the chromatograms obtained in each analysis during sequential analysis of the same sample.

[0044] [ Figure 11 [Conceptual diagram used to illustrate other examples of presumed anomalies]

[0045] [ Figure 12 The flowchart shows other examples of steps to infer anomalies from analyzed data. Detailed Implementation

[0046] Hereinafter, an embodiment of the analytical anomaly diagnosis system and analytical anomaly diagnosis method of the present invention will be described with reference to the accompanying drawings.

[0047] like Figure 1As shown, the analytical anomaly diagnostic system 1 includes a processing unit 2 and a database 4. The processing unit 2 is used to parse and process the analytical data obtained in the analytical apparatus 100, and is implemented via a personal computer or similar device with dedicated software installed for the analytical apparatus 100. The analytical apparatus 100 is, for example, a liquid chromatograph. The database 4 is used to store data for creating learning models and to provide this data to the processing unit 2. The database 4 can be an optical storage medium, non-volatile memory, or other device directly connected to the processing unit 2 and providing data to it, or it can be a device that communicates with the processing unit 2 via the Internet to provide data to it.

[0048] The processing unit 2 includes a learning model creation unit 6, a learning model storage unit 8, an analysis data storage unit 10, and an anomaly prediction unit 12. The learning model creation unit 6 and the anomaly prediction unit 12 are functions implemented by the CPU (Central Processing Unit) executing specific programs. The learning model storage unit 8 and the analysis data storage unit 10 are functions implemented by a portion of the storage area of ​​the data storage device.

[0049] The learning model creation unit 6 is configured to create learning models using learning model creation data provided by database 4. For example... Figure 2 As shown, the data for creating the learning model includes analysis data obtained under normal conditions, i.e., by performing analysis in a state where no abnormalities occur in the analysis device (normal analysis data), and analysis data obtained by performing analysis in states where various abnormalities occur (abnormal analysis data 1 to n). The learning model creation unit 6 uses a machine learning algorithm to create a learning model based on the learning model creation data composed of these analysis data, which learns how the presence of a certain abnormality has a trend-like impact on the parameters of the analysis data when a certain abnormality occurs during analysis.

[0050] When the analytical data is a chromatogram, parameters such as retention time, peak area, peak symmetry coefficient, theoretical plate number, peak area / height, peak start time, and peak end time can be used as parameters for the analytical data used to create a learning model.

[0051] The learning model holding unit 8 holds the learning model created by the learning model creation unit 6. Alternatively, the learning model holding unit 8 can also hold a learning model provided externally. In this case, since the externally provided learning model can be used in the anomaly estimation performed by the anomaly estimation unit 12 (described later), the anomaly analysis and diagnosis system 1 may not need to include the learning model creation unit 6.

[0052] The analytical data holding unit 10 holds analytical data acquired by analyzing a sample using the analytical apparatus 100. The analytical data held by the analytical data holding unit 10 may include baseline analytical data and diagnostic target analytical data. The baseline analytical data is acquired by analyzing a sample under conditions where no abnormalities occur in the analytical apparatus 100. The diagnostic target analytical data is acquired by analyzing the same sample under the same analytical conditions as those used to acquire the baseline analytical data. However, when diagnosing abnormalities such as air bubbles during sample injection or excessive sample in the sample vial from multiple analytical data obtained through continuous analysis by repeatedly injecting the same sample into the analytical apparatus 100, the analytical data held by the analytical data holding unit 10 may not include the baseline analytical data. The diagnosis of abnormalities such as air bubbles during sample injection and excessive sample in the sample vial will be described later.

[0053] like Figure 3 As shown, the anomaly estimation unit 12 is configured to: apply the parameters of the baseline analysis data and the diagnostic object analysis data held in the analysis data holding unit 10 to the learning model held in the learning model holding unit 8, thereby estimating anomalies present in the analysis when acquiring the diagnostic object analysis data.

[0054] As for the types of abnormalities that the abnormality presumption section 12 diagnoses, examples include abnormal sample concentration and abnormal sample injection volume.

[0055] like Figure 4 As shown, in liquid chromatography, if the sample injection volume is the same, the higher the sample concentration, the larger the peak area of ​​the chromatogram; however, the retention time (peak position) is often not affected by the sample concentration. On the other hand, as... Figure 5 As shown, if the sample concentration is the same, the larger the sample injection volume, the larger the peak area, and the later the retention time tends to be. By preparing a learning model that has learned this trend, the anomaly estimation unit 12 can estimate whether there are abnormalities in sample concentration and sample injection volume in the analysis when acquiring the diagnostic subject analysis data, based on the differences in parameters such as peak area, peak height, and retention time between the diagnostic subject analysis data and the benchmark analysis data.

[0056] In addition, among the types of abnormalities that the abnormality estimation unit 12 diagnoses, examples include leakage caused by poor piping connections and the presence of dead volume.

[0057] Figure 6 This is a chromatogram showing the retention time distribution of peaks obtained under conditions of leakage. The range between the two thick dashed lines represents the retention time under normal conditions without leakage; the higher up the lines, the later the retention time. Figure 6It can be seen that if leakage occurs due to poor piping connections, the flow rate of the mobile phase towards the separation column decreases and the retention time tends to be longer compared to the normal state when no leakage occurs. Furthermore, Figure 6 The vertical axis represents the retention time, and the horizontal axis represents the data number.

[0058] Figure 7 This is a distribution map of chromatographic peak symmetry coefficients obtained under conditions where dead volume exists in the piping connections. The range between the two thick dashed lines represents the symmetry coefficients under normal conditions without dead volume; the higher the peaks move above the dashed lines, the worse the symmetry coefficients become. Figure 7 It can be seen that although no leakage occurred, if dead volume exists in the piping connection area due to poor piping connection, the peak will tail, and the peak's symmetry coefficient will tend to increase. Additionally, Figure 7 The vertical axis represents the symmetry coefficient, and the horizontal axis represents the data number.

[0059] By preparing a learning model that has learned the above trends, the anomaly estimation unit 12 can estimate whether there are anomalies such as leakage and dead volume in the analysis when acquiring the diagnostic object analysis data, based on the differences in parameters such as retention time and symmetry coefficient between the diagnostic object analysis data and the benchmark analysis data.

[0060] Next, refer to Figure 8 An example of an analytical anomaly diagnosis method using analytical anomaly diagnosis system 1 is illustrated.

[0061] First, a learning model is prepared to study the impact of anomalies (e.g., abnormal sample concentration, abnormal injection volume, poor piping connection, etc.) on the parameters of the analytical data, and this model is stored in the learning model holding unit 8 (steps 101 and 102). Next, baseline analysis data and diagnostic target analysis data are prepared and stored in the analysis data holding unit 10 (steps 103, 104, and 105). If baseline analysis data already exists, there is no need to prepare it again (step 103: Yes). That is, in the diagnosis of anomalies in multiple diagnostic target analysis data obtained by analyzing the same sample under the same analytical conditions using the same analytical apparatus 100, common baseline analysis data can be used.

[0062] Then, the anomaly estimation unit 12 applies the baseline analysis data and the diagnostic object analysis data to the learning model to estimate the anomalies that exist in the analysis when acquiring the diagnostic object analysis data (steps 106 and 107).

[0063] Next, an example of anomaly diagnosis will be explained, focusing on issues such as air bubble contamination and excessive sample in sample vials during continuous analysis.

[0064] In continuous analysis using a liquid chromatograph, if air bubbles are introduced into the flow path from the injection needle to the syringe pump, only a smaller sample volume than the injection volume set for the first analysis will be injected into the mobile phase. Therefore, if... Figure 9 As shown in the example, the peak area of ​​the chromatogram obtained in the first analysis is smaller than that in normal cases. On the other hand, in the second and subsequent analyses, bubbles are removed due to cleaning between analyses, etc., and the repeatability of peak area values ​​is improved, tending to obtain chromatograms with peak areas larger than those in the first chromatogram.

[0065] In addition, when the sample vial contains an excess of sample exceeding the prescribed amount, such as Figure 10 As shown, the peak area value of the chromatogram in the first analysis was larger, while the peak area value of the chromatogram in subsequent analyses was smaller than that in the first analysis, and the peak area value of the chromatogram in subsequent analyses tended to fluctuate.

[0066] By preparing a learning model that has learned the above trends, such as Figure 11 As shown, the anomaly estimation unit 12 can apply the analysis data of multiple diagnostic objects obtained from continuous analysis to the learning model, thereby estimating whether there are anomalies such as air bubbles or excessive sample in the sample bottle during continuous analysis.

[0067] That is, such as Figure 12 As shown in the flowchart, for anomalies diagnosed through multiple analyses (such as bubble contamination in continuous analysis, excessive sample in sample vials, etc.) and based on the changes in these analysis results, there is no need to prepare baseline analysis data. Anomaly diagnosis can be performed through the following steps: preparing a learning model (steps 201 and 202), preparing diagnostic object analysis data (step 203), and applying the diagnostic object analysis data to the learning model to infer anomalies (steps 204 and 205).

[0068] Furthermore, the embodiments described above are merely examples of implementations of the analytical anomaly diagnosis system and method of the present invention. The implementations of the analytical anomaly diagnosis system and method of the present invention are as follows.

[0069] In the first embodiment of the analytical anomaly diagnosis system of the present invention

[0070] have:

[0071] The learning model maintenance unit maintains the learning model, which is created using data for creating the learning model. The data for creating the learning model includes multiple parameters of analytical data obtained through normal analysis, and the multiple parameters of multiple analytical data obtained through analysis under the conditions of multiple anomalies existing alone or mixed. The learning model learns the influence of each of the multiple anomalies on the multiple parameters in a trend-based manner.

[0072] An analytical data holding unit holds baseline analytical data and analytical data intended for diagnosis. The baseline analytical data is obtained through normal analysis of a reference sample containing a specific component under specific analytical conditions. The analytical data intended for diagnosis is obtained through analysis of a sample containing at least the specific component under the same specific analytical conditions.

[0073] An anomaly estimation unit is configured to apply the respective plurality of parameters of the baseline analysis data and the analysis data as the diagnostic object held in the analysis data holding unit to the learning model held in the learning model holding unit, thereby estimating the types of anomalies present during the analysis performed to obtain the analysis data as the diagnostic object.

[0074] In the first embodiment of the above-described anomaly diagnosis system, in the first form of the analysis...

[0075] have:

[0076] A database that stores data used to create the learning model; and

[0077] The learning model creation department is configured to create the learning model using the learning model creation data stored in the database;

[0078] The learning model holding unit is configured to hold the learning model created by the learning model making unit.

[0079] In the second embodiment of the above-described first implementation of the analytical anomaly diagnostic system, the analytical data is a chromatogram.

[0080] In the second form of the first embodiment of the analysis anomaly diagnosis system described above,

[0081] The parameters may include the area values ​​of the peaks appearing in the chromatogram and the retention times of the peaks;

[0082] The various anomalies may include abnormal sample concentration and abnormal sample injection volume.

[0083] In the second form of the first embodiment of the analysis anomaly diagnosis system described above,

[0084] The parameters may include the retention time of the peaks appearing in the chromatogram and the symmetry coefficient of the peaks;

[0085] The various anomalies may include leakage and the occurrence of dead volume.

[0086] In the second embodiment of the analytical anomaly diagnostic system of the present invention

[0087] have:

[0088] The learning model maintenance unit maintains the learning model, which is created using learning model creation data. The learning model creation data includes parameters of multiple analysis data obtained by continuously executing multiple normal analyses, and parameters of multiple analysis data obtained by continuously executing multiple analyses under the condition that multiple anomalies exist individually or in a mixed manner. The learning model learns the influence of each of the multiple anomalies on the parameters of the multiple analysis data in a trend-based manner.

[0089] An analytical data holding unit holds multiple analytical data points used for diagnosis, said multiple analytical data points being obtained by performing multiple analyses on the same sample; and

[0090] An anomaly estimation unit is configured to apply the parameters of each of the multiple analytical data as diagnostic objects held in the analytical data holding unit to the learning model held in the learning model holding unit, thereby estimating the types of anomalies present during the multiple analyses performed to obtain the multiple analytical data as diagnostic objects.

[0091] In the first embodiment of the above-described second implementation of the anomaly diagnosis system,

[0092] have:

[0093] A database that stores data used to create the learning model; and

[0094] The learning model creation department is configured to create the learning model using the learning model creation data stored in the database;

[0095] The learning model holding unit is configured to hold the learning model created by the learning model making unit.

[0096] In the second embodiment of the above-described abnormality diagnosis system, in the second form...

[0097] The analytical data is a chromatogram;

[0098] The parameter includes the area values ​​of the peaks appearing in the chromatogram;

[0099] The various anomalies include air bubbles entering the sample during injection, and anomalies within the container holding the sample.

[0100] In the first embodiment of the analytical anomaly diagnosis method of the present invention

[0101] have:

[0102] The production steps involve using multiple parameters from analytical data obtained through normal analysis, as well as the multiple parameters from multiple analytical data obtained through analysis under various abnormalities, either individually or in a mixed state, to create a learning model that trend-learns the influence of each of the various abnormalities on the multiple parameters.

[0103] The preparation steps include preparing baseline analysis data and diagnostic analysis data. The baseline analysis data is obtained through normal analysis of a reference sample containing a specific component under specific analytical conditions. The diagnostic analysis data is obtained through analysis of a sample containing at least the specific component under the same specific analytical conditions.

[0104] The estimation step involves applying the respective plurality of parameters of the baseline analysis data and the analysis data as the diagnostic object to the learning model after the production step and the preparation step, thereby estimating the types of anomalies present during the analysis performed to obtain the analysis data as the diagnostic object.

[0105] In the first embodiment of the above-described method for analyzing abnormal diagnostics, the analytical data is a chromatogram.

[0106] In the first embodiment of the above-described first form of the analysis anomaly diagnosis method,

[0107] The parameters may include the area values ​​of the peaks appearing in the chromatogram and the retention times of the peaks;

[0108] The various anomalies may include abnormal sample concentration and abnormal sample injection volume.

[0109] Furthermore, in the first embodiment of the above-described first form of the analysis anomaly diagnosis method,

[0110] The parameters may include the retention time of the peaks appearing in the chromatogram and the symmetry coefficient of the peaks;

[0111] The various anomalies may include leakage and the occurrence of dead volume.

[0112] In the second embodiment of the analytical anomaly diagnosis method of the present invention

[0113] have:

[0114] The production steps involve using the parameters of multiple analytical data obtained by continuously executing multiple normal analyses, as well as the parameters of multiple analytical data obtained by continuously executing multiple analyses under the conditions of multiple anomalies existing alone or mixed, to create a learning model that trend-learns the influence of each of the multiple anomalies on the parameters of the multiple analytical data.

[0115] Preparation steps include preparing multiple analytical data sets for the diagnostic object, wherein the multiple analytical data sets are obtained by performing multiple analyses on the same sample; and

[0116] The estimation step involves applying the parameters of each of the multiple analytical data points used as diagnostic objects to the learning model, thereby estimating the types of anomalies present during the multiple analyses performed to obtain the multiple analytical data points used as diagnostic objects.

[0117] In the second embodiment of the above-described analysis of the abnormality diagnosis method,

[0118] The analytical data may be a chromatogram;

[0119] The parameter may include the area values ​​of the peaks appearing in the chromatogram;

[0120] The various anomalies may include air bubbles entering the sample during injection, as well as anomalies within the container holding the sample.

[0121] Symbol Explanation

[0122] 1. Analysis of the anomaly diagnosis system

[0123] 2. Processing Unit

[0124] 4. Database

[0125] 6. Learning Model Making Department

[0126] 8. Learning Model Retention Section

[0127] 10. Data Preservation Department

[0128] 12. Anomaly Prediction Department

[0129] 100 Analytical apparatus

Claims

1. An analysis and diagnosis system for anomalies, characterized in that... have: The learning model maintenance unit maintains the learning model, which is created using data for creating the learning model. The data for creating the learning model includes multiple parameters of analytical data obtained through normal analysis, and the multiple parameters of multiple analytical data obtained through analysis under the conditions of multiple anomalies existing alone or mixed. The learning model learns the influence of each of the multiple anomalies on the multiple parameters in a trend-based manner. The analytical data holding unit holds baseline analytical data and analytical data as a diagnostic target. The baseline analytical data is analytical data obtained by performing normal analysis on a reference sample containing a specific component under specific analytical conditions. The analytical data as a diagnostic target is obtained by analyzing a sample containing at least the specific component under the specific analytical conditions. as well as An anomaly estimation unit is configured to apply the respective plurality of parameters of the baseline analysis data and the analysis data as the diagnostic object held in the analysis data holding unit to the learning model held in the learning model holding unit, thereby estimating the types of anomalies present during the analysis performed to obtain the analysis data as the diagnostic object.

2. The analytical anomaly diagnostic system according to claim 1, characterized in that... have: A database that stores data used to create the learning model; and The learning model creation department is configured to create the learning model using the learning model creation data stored in the database; The learning model holding unit is configured to hold the learning model created by the learning model making unit.

3. The analytical anomaly diagnostic system according to claim 1, characterized in that, The analytical data are chromatograms.

4. The analytical anomaly diagnostic system according to claim 3, characterized in that, The parameters include the area values ​​of the peaks appearing in the chromatogram and the retention times of the peaks. The various anomalies include abnormal sample concentration and abnormal sample injection volume.

5. The analytical anomaly diagnostic system according to claim 3, characterized in that, The parameters include the retention time of the peaks appearing in the chromatogram and the symmetry coefficient of the peaks. The various anomalies include leakage and the occurrence of dead volume.

6. An analysis and diagnosis system for anomalies, characterized in that... have: The learning model maintenance unit maintains the learning model, which is created using learning model creation data. The learning model creation data includes parameters of multiple analysis data obtained by continuously executing multiple normal analyses, and parameters of multiple analysis data obtained by continuously executing multiple analyses under the condition that multiple anomalies exist individually or in a mixed manner. The learning model learns the influence of each of the multiple anomalies on the parameters of the multiple analysis data in a trend-based manner. The analytical data holding unit holds multiple analytical data points as diagnostic targets, which are obtained by performing multiple analyses on the same sample. as well as An anomaly estimation unit is configured to apply the parameters of each of the multiple analytical data as diagnostic objects held in the analytical data holding unit to the learning model held in the learning model holding unit, thereby estimating the types of anomalies present during the multiple analyses performed to obtain the multiple analytical data as diagnostic objects.

7. The analytical anomaly diagnostic system according to claim 6, characterized in that... have: A database that stores data used to create the learning model; and The learning model creation department is configured to create the learning model using the learning model creation data stored in the database; The learning model holding unit is configured to hold the learning model created by the learning model making unit.

8. The analytical anomaly diagnostic system according to claim 6, characterized in that, The analytical data are chromatograms. The parameter includes the area values ​​of the peaks appearing in the chromatogram. The various anomalies include air bubbles entering the sample during injection, and anomalies within the container holding the sample.

9. A method for analyzing and diagnosing abnormalities, characterized in that... have: The production steps involve using multiple parameters from analytical data obtained through normal analysis, as well as the multiple parameters from multiple analytical data obtained through analysis under various abnormalities, either individually or in a mixed state, to create a learning model that trend-learns the influence of each of the various abnormalities on the multiple parameters. The preparation steps include preparing baseline analysis data and analytical data for diagnosis. The baseline analysis data is obtained by performing normal analysis on a baseline sample containing a specific component under specific analytical conditions. The analytical data for diagnosis is obtained by analyzing a sample containing at least the specific component under the specific analytical conditions. as well as The estimation step involves applying the respective plurality of parameters of the baseline analysis data and the analysis data as the diagnostic object to the learning model after the production step and the preparation step, thereby estimating the types of anomalies present during the analysis performed to obtain the analysis data as the diagnostic object.

10. The analytical anomaly diagnosis method according to claim 9, characterized in that, The analytical data are chromatograms.

11. The analytical anomaly diagnosis method according to claim 10, characterized in that, The parameters include the area values ​​of the peaks appearing in the chromatogram and the retention times of the peaks. The various anomalies include abnormal sample concentration and abnormal sample injection volume.

12. The analytical anomaly diagnosis method according to claim 10, characterized in that, The parameters include the retention time of the peaks appearing in the chromatogram and the symmetry coefficient of the peaks. The various anomalies include leakage and the occurrence of dead volume.

13. A method for analyzing and diagnosing abnormalities, characterized in that... have: The production steps involve using the parameters of multiple analytical data obtained by continuously executing multiple normal analyses, as well as the parameters of multiple analytical data obtained by continuously executing multiple analyses under the conditions of multiple anomalies existing alone or mixed, to create a learning model that trend-learns the influence of each of the multiple anomalies on the parameters of the multiple analytical data. Preparation steps include preparing multiple analytical data points for the diagnostic object, which are obtained by performing multiple analyses on the same sample; as well as The estimation step involves applying the parameters of each of the multiple analytical data points used as diagnostic objects to the learning model, thereby estimating the types of anomalies present during the multiple analyses performed to obtain the multiple analytical data points used as diagnostic objects.

14. The analytical anomaly diagnosis method according to claim 13, characterized in that, The analytical data are chromatograms. The parameter includes the area values ​​of the peaks appearing in the chromatogram. The various anomalies include air bubbles entering the sample during injection, and anomalies within the container holding the sample.

Images