Predictive Monitoring Method for Measuring Accuracy of Measuring Devices of Medium Variables and Measurement Variables

Abstract

The present disclosure relates to a method for predictive monitoring of the measurement accuracy of a measurement device for a medium variable and a measurement variable. The medium is located in a container of a facility that operates independently of the measured value and whose operation includes recurring specific operation phases during which the characteristics of the measured values measured are distinguishable from the measured values measured during other time periods, and when the facility is operating correctly and the measurement device meets the specified measurement accuracy, the characteristics conform to reference characteristics. The method includes: continuously recording data including the measured values during the operation of the facility; determining, based on training data included in the data, a classification method that identifies a data set measured during a specific operation phase; performing the classification method and determining, based on the identified data set, a time series of a conformity indicator indicating the degree of conformity of at least one property indicating the characteristics of the measured values with a reference property of the reference characteristics; determining, based on the time series at least once, the remaining time until the degree of conformity drops below a predetermined minimum degree of conformity; and providing an output notifying the remaining time.

Description

Field of the Invention

[0001] The present invention relates to a method for predicting the measurement accuracy of a measuring device for monitoring a variable of a medium in a container of a facility and measuring this variable during operation of the facility. Background Art

[0002] Measuring devices for measuring variables of a medium located inside a container (such as a tank or a pipe) are used in various types of facilities, such as facilities that repeatedly perform a predetermined intermittent process, facilities of an industrial plant (such as a production plant that performs a production process, a facility of a chemical plant, a facility of a biotech plant designed to perform a biotech process), and laboratory facilities. Measuring devices applied to these applications include various types of devices (such as a level measuring device for measuring the level of a medium in a container, a flow meter for measuring the flow rate of a medium flowing through a pipe, a temperature measuring device for measuring the temperature of a medium, a pressure measuring device for measuring the pressure of a medium), and measuring devices such as an amperometric measuring device, a potentiometric measuring device, a photometric measuring device, or a spectroscopic measuring device for measuring the pH value of a medium or the concentration of an analyte contained in the medium.

[0003] In many facilities, methods of process automation are applied to monitor, adjust, and / or control the operation of the facility based on measurement results provided by measuring instruments that measure parameters required for the corresponding purpose. These measuring instruments include measuring instruments that provide measurement results applied to adjust and / or control the operation of the facility, and measuring instruments that provide measurement values that are neither used for adjusting nor for controlling the operation of the facility. Hereinafter, the term measuring instrument is used as a general term for any instrument that provides a measurement result, and the subgroup of measuring instruments that provide measurement values (which are neither used for adjusting nor for controlling the operation of the facility) is referred to as a measuring device. Examples of measuring devices are, for example, measuring devices applied to monitor the variable they measure, and measuring devices applied to confirm that the variable of a medium (such as the medium given by an intermediate or final product produced on the facility) meets specific requirements. Thus, even though these measuring devices are neither applied to adjust nor to control the operation of the facility, they still play an important role, for example, in monitoring the operation of the facility and / or in ensuring and / or confirming compliance with predetermined quality standards. Therefore, non-compliance of a measuring device with the specified measurement accuracy may have serious consequences, ranging from impaired production processes, production and / or sale of non-conforming products to potential hazards to people and / or the environment.

[0004] The measurement properties (in particular the measurement accuracy) of measuring instruments (including measuring devices) vary over time, for example due to aging and / or due to the measuring instrument being exposed to harsh conditions. To ensure the proper operation of measuring instruments, they are calibrated regularly and subsequently repaired, adjusted or replaced in case it is found during calibration that they no longer meet the measurement accuracy specified for them. During calibration, for example, non - compliance is determined when the measurement error of the measuring instrument determined during calibration exceeds the maximum allowable error.

[0005] Calibration is not only time - consuming and costly, but also usually requires removing the measuring instrument from the process. For example, the calibration of a pH measuring instrument including a measurement chamber enclosed by an H+ ion - selective membrane usually requires immersing the measurement chamber in a reference solution (such as a buffer solution) so that the outer surface of the membrane is exposed to the reference solution. In some applications, removing the measuring instrument requires interrupting the operation of the facility. This is particularly disadvantageous for facilities (such as facilities performing chemical or biotechnological methods or processes) where high standards of hygiene or even sterility must be ensured. Even if calibration does not require interrupting the operation of the facility, it still causes the problem that the measuring instrument cannot perform measurements at the facility while it is being calibrated.

[0006] Therefore, it is desirable to reduce the number of calibrations to the minimum required to ensure the safe operation of the facility and full compliance with the quality standards regarding the tasks performed by the facility and regarding any intermediate or final products resulting from the performance of the tasks. Unfortunately, due to the multitude of influencing factors affecting the measurement properties of measuring instruments, it is usually not possible to reliably predict a future point in time at which the measurement properties of the measuring instrument will degrade so much that the measurement error of the measurement results provided by the measuring instrument exceeds the maximum allowable error. Therefore, for safety reasons, calibration is usually performed more frequently than may be necessary due to the true condition of the measuring instrument.

[0007] To improve this situation, the calibration time interval between successive calibrations can be optimized. In this context, EP 2602 680 B1 describes a method for determining an optimized next calibration time at which a particular measuring instrument needs to be recalibrated. This next calibration time is determined based on a Monte Carlo simulation that is performed based on the measurement errors of the measuring instrument determined during at least two previously performed calibrations and a probability density function for determining the measurement error in respect of the respective calibration due only to the uncertainty inherent in the respective calibration. However, this method does require two previously performed calibrations and the uncertainty inherent in calibration, which may not always be available.

[0008] As an alternative, methods can be carried out which enable the verification, calibration and / or adjustment of a measuring instrument for a measured variable while the measuring instrument remains at the measuring site. Examples are described, for example, in DE 102018109696 A1. These methods are based on a reference measurement of the variable carried out by an additional measuring instrument. As a result, they lead to additional costs and effort involved in providing an additional measuring instrument, carrying out the reference measurement and ensuring a sufficiently high measuring accuracy of the additional measuring instrument.

[0009] Furthermore, a method for operating a measuring instrument that measures a variable at an operating site and predicting and monitoring that at least one characteristic of the measuring instrument meets requirements specified for the measuring instrument is described in European Patent Application No. 20168733.2, filed on April 8, 2020. This method is carried out based on a continuously monitored deviation between a measured value determined by the measuring instrument and a corresponding reference value. Based on the deviation, a method based on time series prediction determines the remaining time until the deviation will exceed a deviation range defined for the deviation. Since this method is carried out based on the deviation between the monitored variable and the corresponding reference value, it can only be applied when the corresponding reference value is available. Summary of the Invention

[0010] The object of the present invention is to provide a method for predictive monitoring that is capable of monitoring both a variable measured by a measuring device and the measuring accuracy of the measuring device during the operation of a facility, and that does not require a reference measurement of the variable.

[0011] In this regard, the present invention includes a method for predictive monitoring of a variable of a medium located in a container of a facility and the measuring accuracy of a measuring device that measures this variable and provides a measured value of this variable during the operation of the facility, in particular a computer-implemented method, wherein the facility operates independently of the measured value, and wherein the operation of the facility includes recurring specific operating phases, wherein the measured values measured during the specific operating phases exhibit characteristics that distinguish these measured values from the measured values measured during other time periods, and wherein when the facility is operating correctly and the measuring device meets the specified measuring accuracy during the corresponding specific operating phase, this characteristic conforms to a reference characteristic. This method includes the following steps:

[0012] Install the measuring device at the facility;

[0013] During the operation of the facility, continuously record data, including the measured values measured by the measuring device and their measurement times;

[0014] Based on training data included in this data that has been recorded during a training time interval, a classification method is determined that can identify a data set in which each measurement included in the recorded data has been measured during a period in a specific operation phase. During the training time interval, the facility operates correctly and the measuring device meets the specified measurement accuracy during the training time interval;

[0015] Execute the classification method and, based on the data set identified by the classification method, determine a time series of a compliance indicator that indicates the degree of compliance of at least one property indicating the characteristics of the measured values included in the data set with a reference property corresponding to the reference characteristics;

[0016] Based on the time series, determine at least once the remaining time until the degree of compliance indicated by the compliance indicator to be determined based on the measured values to be measured during a future occurrence in a specific operation phase drops below a predetermined minimum degree of compliance; and

[0017] Provide an output notifying the remaining time.

[0018] This method provides the advantage that it can be executed during the operation of the facility, the measured values measured by the measuring device are available during the execution of the method, and it does not require reference measurements of variables.

[0019] Furthermore, on the premise that the operation of the facility includes a specific operation phase that repeats, where the measured values measured during the specific operation phase exhibit sufficiently distinguishable characteristics, this method provides the advantage that the identification of the specific operation phase, the determination of the characteristics and reference characteristics, and the determination of the classification method can be executed based only on the training data without any prior knowledge about the specific operation phase. Therefore, executing this method requires neither prior knowledge about the facility and the operation of the facility nor reference measurements of variables.

[0020] By requiring the facility to operate independently of the measured values, it is ensured that the measured values measured through measurement have no direct or indirect influence on the true value of the variable being measured. This provides the advantage that predictive monitoring not only detects damages caused by impaired measurement properties of the measuring device but also detects damages caused by impaired operation of the facility that affects the true value of the variable during a specific operation phase.

[0021] According to a first improvement,

[0022] The specific operation phase is predetermined based on available information about the operation of the facility or is identified based on the training data; and / or

[0023] The specific operation phase is:

[0024] a) An operating phase that occurs during each execution of a predefined intermittent process, where the intermittent process is repeatedly executed on or by a facility during the operation of the facility;

[0025] b) An operating phase in which a variable should be equal to a constant;

[0026] c) A cleaning phase, where the variable of the medium measured by a measuring device during each cleaning phase is the variable of the same cleaning agent applied to a cleaning container during each cleaning phase, or

[0027] d) An empty phase, where the variable of the medium measured by a measuring device is the variable of the gas or air contained in an empty container during each empty phase.

[0028] According to the second improvement:

[0029] This characteristic is determined based on at least one of training data and reference characteristics, and / or includes at least one property of the measured values, which includes at least one of the following: the value of the measured value, the slope of the measured value, at least one fitting coefficient that can be determined by fitting the measured value to a function of time, and a set of one or more coefficients that describe the measured values measured during a specific operating phase, the value range in which the measured values appear, the distribution of the measured values, the pattern described by the measured values, the model corresponding to the measured values used to measure during a specific operating phase, at least one property of the model properties of a deterministic model, a statistical model, or a hybrid model including deterministic model components and statistical model components, and at least one other property; and

[0030] The reference characteristic is determined based on training data and / or includes reference properties for each property of the characteristic, where the reference property represents the measured values measured during one of the specific operating phases while the facility is operating correctly and the measuring device meets the specified measurement accuracy, and the reference property includes at least one of the following: the reference value of the measured value, the reference slope, a set of one or more reference coefficients, the reference pattern, the reference distribution, the reference property for at least one model property, and at least one other reference property that would be expected for the measured values measured during a specific operating phase.

[0031] According to the third improvement:

[0032] The method for determining classification includes the following steps: identifying a specific operating phase, identifying a specific operating phase based on training data, or identifying a specific operating phase based on potential candidates for a specific operating phase determined based on training data and available information about the operation of the facility;

[0033] Execute a classification method based on classification criteria determined for a specific operation phase, the classification criteria including at least one of the following: at least one criterion of an expected value or value range regarding a measured value measured during the specific operation phase, at least one criterion of a pattern described by an expected measured value regarding the measured value measured during the specific operation phase, at least one criterion of a distribution of an expected measured value regarding the measured value measured during the specific operation phase, at least one criterion of a degree of compliance with at least one property included in a characteristic and a corresponding reference property included in a reference characteristic of the measured value, at least one criterion related to a model property of a model for the measured value measured during the specific operation phase, and at least one other criterion, and / or

[0034] Determine a data set included in the data and satisfying the classification criteria applied to identify the data set by performing at least one of the following: correlation analysis, pattern recognition methods, autocorrelation analysis, and at least one other data analysis method capable of identifying a data set that satisfies the classification criteria.

[0035] According to the fourth improvement, the method further includes the following steps:

[0036] Identify at least one group of subsets of training data, where each subset consists of data measured during a subset time interval, and where subsets belonging to the same group exhibit a similarity greater than or equal to a minimum similarity required for the subsets to be considered to belong to the same group, and where identifying at least one group of subsets is performed by performing at least one of the following: correlation analysis, pattern recognition methods, autocorrelation analysis, and at least one other data analysis method capable of identifying subsets representing the same operation phase;

[0037] For at least one of the groups, apply the subsets included in the corresponding group as a reference set representing the same operation phase;

[0038] Identify one for which a reference set has been determined in an operation phase as a specific operation phase,

[0039] Determine a characteristic and a reference characteristic based on the reference set representing the specific operation phase, and

[0040] Based on the reference set for the specific operation phase, determine at least one of a classification method and classification criteria for identifying a data set included in the recorded data.

[0041] According to the improvement of the fourth improvement, where reference sets have been determined for at least two different operation phases, this method includes at least one of the following steps:

[0042] Identify a particular operation phase as one of these different operation phases that has a longer duration and / or a higher occurrence frequency than at least one other operation phase, for which a reference set has been determined, and

[0043] Identify a particular operation phase such that the reference set for the particular operation phase has a higher similarity than the reference sets determined for at least one other operation phase.

[0044] According to the fifth improvement,

[0045] Determine a time series by determining, for each data set identified by a classification method, a quantitative measure that complies with the similarity of one of the indicators being equal to the overall characteristic and the overall reference characteristic exhibited by the measured values included in the corresponding data set;

[0046] Or: The degree of compliance of one of the properties indicated by the compliance indicator with the corresponding reference property, and each compliance indicator of the time series is given by this property of the measured values included in one of the data sets;

[0047] Or: a) The particular operation phase is an operation phase in which, during each occurrence of this operation phase, the variable should be equal to the same constant, b) The compliance indicator is given by the measured values included in the data set; and c) When the compliance indicator given by one of the measured values exceeds the range of indicator values for the target value for the constant or the reference constant for the constant included in the reference characteristic, the compliance indicator drops below the minimum degree of compliance;

[0048] Or: a) The particular operation phase is an operation phase in which, during each occurrence of this operation phase, the variable should be equal to the same constant, b) The compliance indicator is given by the deviation between the measured values included in the data set and the target value for the constant or the reference constant for the constant included in the reference characteristic; and c) When the compliance indicator given by one of the deviations exceeds the corresponding deviation range, the compliance indicator drops below the minimum degree of compliance.

[0049] According to the sixth improvement, the training data is labeled as training data including measured values and the corresponding operation phases during which they were measured, and at least one of the following steps is performed by executing a supervised learning method: identifying a particular operation phase, determining a characteristic, determining a reference characteristic, and determining a classification method.

[0050] According to the seventh improvement, the method additionally includes the step of determining and discarding at least one of the following before determining the time series: potentially contaminated measured values and potentially contaminated measured values given by the edge values measured at the start and end of the particular operation phase included in the identified data set.

[0051] According to the eighth improvement, the method includes the step of providing continuously recorded data to a computing unit, where the computing unit: is implemented to execute, is trained to execute, and / or is designed to learn and execute at least one of the following: identify a specific operation phase and determine a classification method based on the data provided to it; determine and execute a classification method based on the data provided to it; determine a time series; and / or determine the remaining time.

[0052] According to the ninth improvement, the facility is implemented to perform a predetermined task or a predetermined process and / or repeatedly perform a predetermined intermittent process; and / or the measuring device is an electrochemical measuring device for measuring the concentration of an analyte contained in a medium or a pH sensor for measuring the pH value of the medium.

[0053] According to the tenth improvement, the measuring device measures at least one parameter; the continuously recorded data includes the measured parameter values of the (multiple) parameters measured and provided by the measuring device and their measurement times; and at least one of the following is performed based on the measured values and the measured parameter values included in the training data: identify a specific operation phase, determine characteristics, determine reference characteristics, determine a classification method, execute a classification method, and determine a time series.

[0054] According to the improvement of the tenth improvement, the at least one parameter includes at least one of the following:

[0055] at least one parameter measured by a sensor of the measuring device,

[0056] at least one parameter applied by the measuring device to determine a measured value of a variable;

[0057] at least one parameter applied by the measuring device to compensate for a measurement error dependent on a parameter;

[0058] the temperature measured by a temperature sensor of the measuring device; and

[0059] the electrode potential of a measuring electrode of the measuring device and / or the impedance of an ion-selective membrane of the measuring device, where the measuring device is an electrochemical measuring device for measuring the concentration of an analyte contained in a medium or a pH sensor for measuring the pH value of the medium, and includes: a measuring chamber enclosed by an ion-selective membrane, the ion-selective membrane having an inner surface exposed to an electrolyte located inside the measuring chamber and an outer surface exposed to the medium, and the measuring electrode being immersed in the electrolyte.

[0060] According to the eleventh improvement,

[0061] each remaining time is determined by performing a method of time series prediction or by performing a method of time series prediction, the method including the following steps:

[0062] For each compliance indicator included in the time series, determine the deviation between the corresponding compliance indicator and a target value of the degree of compliance or a target value of 100% degree of compliance.

[0063] Filter the deviation.

[0064] Based on the deviation and the filtered deviation, determine the noise superimposed on the filtered deviation, and

[0065] At the end of at least one monitoring time interval (during which three or more compliance indicators included in the time series have been determined and none of the compliance indicators is below the minimum degree of compliance), determine the remaining time as the remaining time until the deviation will exceed the deviation range.

[0066] wherein the deviation range is determined based on the minimum degree of compliance such that the deviation exceeds the deviation range when the degree of compliance indicated by the compliance indicator drops below the minimum degree of compliance, and

[0067] wherein the remaining time is determined by:

[0068] For at least two different pairs of deviations each including a first deviation and a second deviation determined based on the filtered deviation included in the monitoring time interval, determine a simulated value of the remaining time by performing a Monte Carlo simulation based on the noise and the corresponding pair of deviations, and

[0069] Determine the remaining time based on the average or weighted average of the simulated values determined for each pair of deviations.

[0070] According to the twelfth improvement, the method includes the following steps:

[0071] Calibrate the measuring device at or before the time point when the degree of compliance indicated by the compliance indicator will drop below the minimum degree of compliance according to the previously determined remaining time;

[0072] During calibration, determine the measurement error of the measuring device;

[0073] In the case where the measurement error is less than a predetermined threshold, perform at least one of the following: determine the impaired operation of the facility as the root cause leading to the degree of compliance dropping below the minimum degree of compliance, and determine the fault causing the impaired operation and apply a remedial measure to solve the fault; and

[0074] In the case where the measurement error is greater than a predetermined threshold, perform at least one of the following: determine the impaired measurement property of the measuring device as the root cause leading to the degree of compliance dropping below the minimum degree of compliance, and

[0075] Adjust, repair or replace the measuring device and restart the method from the beginning by installing a measuring device that meets the measuring accuracy specified for it.

[0076] According to the thirteenth improvement, the operation of the facility includes additional specific operation phases that occur repeatedly, where the measured values measured during the additional specific operation phases exhibit characteristics that distinguish these measured values from the measured values measured during other time periods, and where when the facility is operating correctly and the measuring device meets the specified measuring accuracy during the corresponding additional specific operation phases, this characteristic conforms to a reference characteristic. The method includes the following steps:

[0077] Based on the training data, determine an additional classification method that can identify an additional data set of measured values included in the recorded data, each of which has been measured during one of the additional specific operation phases;

[0078] Execute the additional classification method and, based on the measured values included in the additional data set identified by the additional classification method, determine an additional time series of compliance indicators that indicate the degree of compliance of at least one property indicating the characteristics of the measured values included in the additional data set with the corresponding reference property of the reference characteristic;

[0079] Based on the additional time series, determine at least once the remaining time until the degree of compliance indicated by the compliance indicator determined based on the measured values to be measured during future occurrences of the additional specific operation phase will drop below a predetermined additional minimum degree of compliance; and

[0080] Provide an output notifying the additional remaining time. Description of the Drawings

[0081] The present invention and additional advantages are explained in more detail using the figures in the drawings.

[0082] Figure 1 Shows: A facility including a measuring device;

[0083] Figure 2 Shows: An example of a measuring device;

[0084] Figure 3 Shows: By Figure 1 The measured values measured by the measuring device;

[0085] Figure 4 Shows: A set of measured values measured during the first filling phase;

[0086] Figure 5 Shows: A time series of compliance indicators;

[0087] Figure 6 Shows: An excerpt of the time series of measured values;

[0088] Figure 7 shows: a method for time series prediction; and

[0089] Figure 8 shows: determination of simulated remaining time. Detailed implementation mode

[0090] The present invention relates to a method for predicting and monitoring the accuracy of a measurement device 3 that measures a variable of a medium 5 in a container 1 of a facility and provides a measured value m(ti) of this variable during operation of the facility, in particular a computer-implemented method, wherein the facility operates independently of the measured value m(ti), and wherein the operation of the facility includes a specific operation phase Ps that recurs, wherein the measured values m(ti) measured during the specific operation phase Ps exhibit a characteristic C that distinguishes these measured values m(ti) from the measured values m(ti) measured during other time periods, and wherein when the facility operates correctly and the measurement device 3 meets a specific measurement accuracy during the corresponding specific operation phase Ps, this characteristic C conforms to a reference characteristic Cr.

[0091] Figure 1 An example of a facility designed to repeatedly perform a predetermined intermittent process is shown. In this example, the container 1 is a tank, such as a bioreactor, and the measuring device 3 is mounted on the container 1 and measures a variable of the medium 5 contained in the container 1. The present invention is not limited to facilities that perform intermittent processes. As an alternative, the facility can be another type of facility, such as a facility designed to perform at least one given task, such as a facility of a production plant that performs a production process, a facility of a chemical plant, a facility of a biotechnology plant, or a facility of a laboratory (such as a facility that performs laboratory analysis). In addition, the container 1 does not have to be a tank. Regarding a measuring device mounted on another type of container having an interior containing a medium, the method can be applied in the same way. Examples are open or closed vessels, and pipes including a medium located inside and / or flowing through the pipe. For example, the method can be applied in the same way, for example, to a measuring device 3' indicated by the dashed line in Figure 1 one of the pipes 7 connected to the tank.

[0092] The measuring device 3 is, for example, an electrochemical sensor, such as a potentiometric sensor that measures the activity or concentration of an analyte included in the medium 5 or the pH value of the medium 5. Examples of electrochemical sensors (such as pH sensors) are shown in Figure 2 This sensor includes an ion-selective membrane 11 (such as H +A measuring chamber 9 enclosed by an ion-selective membrane), the ion-selective membrane having an inner surface exposed to an electrolyte 13 (e.g., a pH buffer solution) located inside the measuring chamber 9 and an outer surface exposed to the surrounding medium 5. Due to the ion-selective interaction between the membrane 11 and the medium 5, a potential Uel corresponding to the variable to be measured, e.g., the pH value of the medium 5, can be derived by a measuring electrode 15 extending into the electrolyte 13. The measuring electrode 15 is connected to measuring electronics 17, which is implemented to quantitatively determine the variable based on the electrode potential Uel provided by the measuring electrode 15 or based on the difference between the electrode potential Uel and a reference potential Uref. As an option, the reference potential Uref is, for example, the potential provided by a reference chamber 19. In the example shown, the reference chamber 19 includes an electrolyte 21 located inside the reference chamber 19, a reference electrode 23 extending into this electrolyte 21, and a diaphragm 25 permeable to charge carriers. The diaphragm 25 is inserted into a wall section of the outer wall of the reference chamber 19 such that the inner surface of the diaphragm 25 is exposed to the electrolyte 21 located inside the reference chamber 19 and the outer surface of the diaphragm 25 is exposed to the surrounding medium 5. The present invention is not limited to pH sensors. As an alternative, the measuring device 3 can, for example, be another type of measuring device and / or a measuring device for another variable (e.g., pressure or turbidity) of the medium 5 included in the measuring vessel 1.

[0093] For example, Figure 1 The operation of the facility shown, for example, includes the repeated execution of an intermittent process (e.g., an intermittent process for producing a batch of products (e.g., proteins or lemonade)) that includes a sequence of operating phases. For example, the sequence, for example, includes:

[0094] a) An empty phase Pe, during which the container 1 is empty,

[0095] b) A first filling phase Pf1, during which a preform is supplied to the container 1 through one of the pipes 7,

[0096] c) A second filling phase Pf2, during which reactants are supplied to the container 1 through one of the pipes 7,

[0097] d) A reaction phase Pr, during which a reaction takes place inside the container 1, and

[0098] e) A discharge phase Pd, during which a batch of products obtained by the reaction is discharged through one of the pipes 7.

[0099] As an option, the operation of the facility can additionally include a repeated cleaning phase Pc, during which the container 1 is cleaned using a cleaning agent. The cleaning phase Pc can, for example, be performed between successive executions of the intermittent process, e.g., after each discharge phase Pd.

[0100] As mentioned above, the operation of the facility is performed independently of the measured value m(ti) measured by the measuring device 3. This means that the operation of the facility is neither adjusted nor controlled based on the measured value m(ti) measured by the measuring device 3, and the measured value m(ti) is neither applied for adjusting nor for controlling the operation of the facility. For example, this is the case where the measuring device 3 is only applied for monitoring the operation of the facility and / or for verifying that the variables of the medium 5 comply with the requirements specified for it. By operating the facility independently of the measured value m(ti), it is ensured that the measured value m(ti) has no direct or indirect influence on the true value of the variables measured by the measuring device 3.

[0101] As an option, the operation of the facility is monitored, adjusted and / or controlled, for example, by the hypercoordination unit 27. For example, the hypercoordination unit 27 is a unit that includes or consists of a system, such as a programmable logic controller, which adjusts and / or controls the operation of the facility based on the measurement results provided by measuring instruments for the parameters required for measuring, adjusting and / or controlling the operation of the facility. For example, Figure 1 shown in which the operation of the facility is adjusted and / or controlled, for example, based on the measurement results f1, f2, f3 provided by flow meters F1, F2, F3 installed on the pipeline 7 for measuring the flow rate flowing into or out of the container 1 through the respective pipeline 7 and / or the measurement result h of the liquid level measuring instrument L for measuring the liquid level of the medium 5 included in the container 1. The adjustment and / or control is performed, for example, by adjusting the valve settings of the valves V1, V2, V3 inserted into the pipeline 7 based on these measurement results f1, f2, f3, h. As an alternative, other means and / or methods for operating the facility independently of the measured value m(ti) can be applied.

[0102] The method includes the steps of installing the measuring device 3 at the facility and operating the facility. During the operation of the facility, the variables of the medium 5 are measured by the measuring device 3, and the data D including the (multiple) measured values m(ti) measured and provided by the measuring device 3 and their measurement times ti are continuously recorded.

[0103] Figure 3 An example of the measured value m(ti) measured by the measuring device 3 during a single execution of the above intermittent process and during the subsequent cleaning phase Pc is shown. In this example, the measured value m(ti) is approximately constant during the empty phase Pe, during the reaction phase Pr, and during the cleaning phase Pc. Further, they decrease during the first filling phase Pf1 and the second filling phase Pf2, and increase during the discharge phase Pd.

[0104] During operation of the facility, predictive monitoring is performed based on continuously recorded data D. As a prerequisite, the monitoring method requires that the operation of the facility includes a specific operation phase Ps that recurs, where the measured values m(ti) measured during the specific operation phase Ps exhibit a characteristic C that distinguishes the measured values m(ti) measured during the specific operation phase Ps from the measured values m(ti) measured during other time periods, and where the characteristic C conforms to a reference characteristic Cr when the facility is operating correctly and the measuring device 3 conforms to the specified measurement accuracy during the respective specific operation phase Ps.

[0105] Depending on the type of facility, the task performed by the facility, and the variable measured by the measuring device 3, the operation phases that recur during the operation of the facility produce measured values m(ti) that exhibit a sufficiently distinguishable characteristic C in a sufficiently reproducible manner and can thus be used as the specific operation phase Ps in the method described herein. Examples are:

[0106] a) An operation phase that occurs during each execution of a repeatedly executed intermittent process, such as one of the operation phases of the above-mentioned intermittent process;

[0107] b) An operation phase during which the variable should be equal to a constant K, such as the above-mentioned empty phase Pe, reaction phase Pr, or cleaning phase Pc;

[0108] c) A cleaning phase Pc, where the measured variable is the variable of the same cleaning agent applied to clean the container 1 during each cleaning phase Pc, and

[0109] d) An empty phase Pe, where the measured variable is the variable of the same gas or air contained in the empty container 1 during each empty phase Pe.

[0110] Assuming correct operation of the facility, each of the recurring operation phases will be performed in the same way each time it occurs. Thus, under ideal conditions, the variable should be equal to the same constant K, or be given by the same deterministic time function, or be given by the same statistical pattern during each recurrence of the same operation phase. Thus, additionally assuming correct operation of the measuring device 3, the measured values m(ti) measured during consecutive occurrences of the same operation phase should be the same within a tolerance that takes into account the variations due to the operation of the facility, the associated reproducibility of the constant, function, or pattern of the variable's reality, and the variations in the nature of the measurement and the limited measurement accuracy of the measuring device 3.

[0111] It is assumed that the operation of the facility includes specific operation phases Ps that occur repeatedly, during which the measured values m(ti) measured satisfy the requirements mentioned above, which automatically causes each characteristic C of the measured values m(ti) measured during one of the specific operation phases Ps to conform to the reference characteristic Cr while the facility is operating correctly and the measuring device 3 meets the specified measurement accuracy.

[0112] For example, the characteristic C is, for example, a characteristic including at least one property of the measured values m(ti), such as their values, their slopes, the value ranges in which they occur, their distributions, and / or the patterns described by them during the specific operation phase Ps. At least one property of the measured values m(ti) includes, for example, at least one property corresponding to the model properties of the model used for the measured values m(ti) measured during the specific operation phase Ps. Suitable models include deterministic models (such as physical models) for the deterministic behavior of the measured values m(ti), statistical models (such as models for the statistical behavior of the measured values m(ti)), and hybrid models that combine deterministic model components and statistical model components. Accordingly, the reference characteristic Cr includes, for each property included in the characteristic C, reference properties representing the measured values m(ti) measured during one of the specific operation phases Ps while the device is operating correctly and the measuring device 3 meets the specified measurement accuracy, such as reference values, reference slopes, reference value ranges, reference patterns, reference distributions, and / or reference properties for each of the applied model properties. Thus, the measured values m(ti) measured during one of the specific operation phases Ps can be distinguished from the measured values m(ti) measured during other time periods based on the degree of conformity of their characteristic C with the reference characteristic Cr.

[0113] To this extent, the method includes a preliminary step of determining a classification method capable of identifying a data set S included in the data D that has been measured during one of the specific operation phases Ps. The classification method is determined based on training data included in the data D that has been measured during a training time interval TI during which the facility is operating correctly and the measuring device 3 meets the measurement accuracy specified for it during the training time interval. For example, the training time interval TI is, for example, the interval after the measuring device 3 is installed. For example, compliance with the specified measurement accuracy is ensured by requiring that the measuring device 3 to be installed is a new device or a newly calibrated device that meets the specified measurement accuracy.

[0114] As an option, the determination of the classification method is, for example, performed by a computing unit 29, which is implemented to determine the classification method based on the continuously recorded data D provided to it. The computing unit 29 is, for example, trained to perform or is designed to learn the determination of the classification method and subsequently perform the classification method. In this case, the computing unit 29 is, for example, implemented (e.g., trained or designed to learn so as to) determine the classification criteria for identifying the data set S based on training data, the data set including the measured values m(ti) measured during one of the specific operation phases Ps in the recorded data D.

[0115] The provision of the data D to the computing unit 29 is, for example, performed by transferring the data D to a memory 31 associated with the computing unit 29 and at least temporarily storing the data D in this memory 31. For this purpose, the measuring device 3 providing the measured value m(ti) and its measurement time ti can, for example, be directly connected to and / or communicate with the computing unit 29 via a hypercoordination unit 27 (e.g., as shown by arrows B1 and B2) and / or via an edge device 33 located near the measuring device 3 (e.g., as shown by arrows C1, C2). To this extent, hardwired or wireless connections and / or communication protocols known in the art can be applied, such as LAN, W-LAN, fieldbus, Profibus, Hart, Bluetooth, near field communication, etc. For example, the measuring device 3, the edge device 33, and / or the hypercoordination unit 27 can be directly or indirectly connected to the computing unit 29 via the Internet (e.g., via a communication network (e.g., TCP / IP)).

[0116] The computing unit 29 is, for example, implemented as a unit including hardware located near or at a remote location from the measuring device 3, such as a computer or a computing system. As an alternative option, cloud computing can be applied. Cloud computing names a method in which IT infrastructure (such as hardware, computing power, memory, network capacity, and / or software) is provided via a network (e.g., the Internet). In this case, the computing unit 29 is implemented in the cloud.

[0117] The classification method can be any method capable of identifying a data set S in the continuously recorded data D. Since during the correct operation of the facility and the measuring device 3, the characteristic C exhibited by the measured value m(ti) measured during a specific operation phase Ps conforms to the reference characteristic Cr, it is possible to determine the classification method based on the training data. Therefore, assuming that the reference characteristic Cr is sufficiently discriminative, the reference characteristic Cr can be determined based on the training data and / or is determined based on the training data. As an option, at least one of the models of the above-mentioned measured values m(ti) and at least one corresponding model property can be determined based on the training data and / or are determined based on the training data. Therefore, the characteristic C can be determined based on the reference characteristic Cr and / or is determined based on the reference characteristic Cr, and the classification criterion enabling the identification of the data set S can be determined based on the training data and / or is determined based on the training data, without any prior knowledge of the characteristic C and the reference characteristic Cr.

[0118] For example, the classification criterion includes, for example, at least one criterion regarding the value, slope, or expected value range of the measured value m(ti) measured during a specific operation phase Ps, at least one criterion regarding the distribution of the expected measured value m(ti) of the measured value m(ti) measured during a specific operation phase Ps, at least one criterion regarding the pattern described by the expected measured value m(ti) of the measured value m(ti) measured during a specific operation phase Ps, and / or at least one criterion regarding one of the expected model properties of the measured value m(ti) measured during a specific operation phase Ps. As an additional or alternative option, the classification criterion includes, for example, at least one criterion related to the degree of conformity between at least one property of the measured value m(ti) included in the characteristic C and the corresponding reference property included in the reference characteristic Cr.

[0119] As an option, the determination of the characteristic C, the determination of the reference characteristic Cr, the determination of the degree of conformity between at least one or all properties of the characteristic C of the measured value m(ti) and the corresponding reference properties of the reference characteristic Cr, the determination of the classification criterion, and / or the identification of the data set S included in the data D are determined, for example, by applying a classification algorithm in dynamic time warping or a neural network (such as the algorithm applied in a support vector machine).

[0120] The specific operation phase Ps is, for example, a recurring operation phase pre-determined based on the available information regarding the operation of the facility. As an alternative option, the specific operation phase Ps is identified based on the training data. In the latter case, as an option, the determination of the classification method includes, for example, the step of identifying the specific operation phase Ps.

[0121] When information about the operation of a facility is available, this information can be used to pre-determine a specific operation phase Ps or to identify one or more potential candidates that may be suitable for use as a specific operation phase Ps, and to identify the specific operation phase Ps based on these candidates. For example, an operation phase in which a variable should be equal to the same constant K during each occurrence of the respective operation phase can be pre-determined as the specific operation phase Ps, or can be identified as one of at least one potential candidate for the specific operation phase Ps. In Figure 1 and Figure 3 In the example shown in

[0122] - An empty phase Pe, in which the variable should be equal to a constant K(Pe) given by the variable value of the medium 5 (i.e., the air or gas included in the empty container 1) during each empty phase Pe,

[0123] - A reaction phase Pr, in which the variable should be equal to a constant K(Pr) given by the variable value of the medium 5 (i.e., the product reacting inside the container 1) during each reaction phase Pr, and

[0124] - A cleaning phase Pc, in which the variable should be equal to a constant K(Pc) given by the variable value of the medium 5 (i.e., the applied cleaning agent) during each cleaning phase Pc.

[0125] In the case where the variable should be constant during a specific operation phase Ps, classification criteria include, for example, a criterion requiring that the measured value m(ti) is static.

[0126] However, the present invention is not limited to a specific operation phase Ps during which the variable should be constant.

[0127] As an additional or alternative option, when labeled training data including the measured values m(ti) and the corresponding operation phases during which they are measured is available, one of at least one recurring operation phase can be identified as the specific operation phase Ps and the characteristic C, with reference to the characteristic Cr and a classification method, for example, by performing a supervised learning method.

[0128] As an alternative option, the identification of the specific operation phase Ps is performed, for example, without any prior knowledge about the recurring operation phases and / or about the potentially suitable candidates for the specific operation phase Ps. Thus, the specific operation phase Ps can even be an operation phase unknown to the owner or user of the facility. Whether or not the candidates have been identified, both the identification of the specific operation phase Ps and the determination of the classification method are performed, for example, based on the training data included in the data D. This identification can be performed without any prior knowledge about the characteristic C and the reference characteristic Cr associated with the specific operation phase Ps.

[0129] For example, the identification of a specific operation phase Ps is performed, for example, by identifying at least one group Gj of subsets of training data, where each subset contains data measured during a subset time interval, and where the subsets belonging to the same group Gj exhibit a similarity greater than or equal to a minimum similarity required for the subsets to be considered to belong to the same group Gj. Thus, the subsets included in the same group Gj can be considered to have been measured during the same operation phase and thus represent the corresponding operation phase. Determining the subsets belonging to the same group Gj is performed, for example, by performing at least one of the following: analysis of the values of the measured values m(ti), correlation analysis, pattern recognition methods, autocorrelation analysis, and another data analysis method capable of identifying subsets of high enough similarity.

[0130] For example, based on training data measured during several consecutive executions of an intermittent process (each followed by a cleaning phase Pc based on Figure 1 and Figure 3 the above-described cleaning phase Pc), up to six different sub-groupings Gj (each corresponding to one of six operation phases Pe, Pf1, Pf2, Pr, Pd, Pc) can be determined.

[0131] Next, for at least one of the groups Gj, the subsets included in the corresponding group Gj that represent the same operation phase are identified as a reference set for the corresponding operation phase.

[0132] When a reference set can be and / or is determined only for a single operation phase, this operation phase is determined as the specific operation phase Ps, and the characteristics C of the measured values m(ti) measured during this specific operation phase Ps, the reference characteristics Cr, and the classification criteria for identifying the data set S (e.g., a classification criteria including at least one of the above-mentioned criteria) are determined based on these reference sets.

[0133] When reference sets have been determined for two or more operating phases, one of these operating phases is determined as the specific operating phase Ps, and based on the reference set for this specific operating phase Ps, the characteristic C, the reference characteristic Cr, and the classification criteria for identifying the data set S measured during the specific operating phase Ps are determined as described above. As an option, one of the operating phases for which a reference set has been determined can be arbitrarily selected as the specific operating phase Ps. As an alternative, the selection is performed, for example, based on the occurrence frequency and / or duration of these operating phases. Additionally or alternatively, preferably the selection is performed based on the similarity of the reference sets included in the same group Gj. Selecting an operating phase that has a higher occurrence frequency than at least one of the other operating phases, and selecting an operating phase that has a longer duration than at least one of the other operating phases provides the advantage that it increases the number and availability of the measured values m(ti) measured during the specific operating phase Ps. Selecting an operating phase for which a reference set has been determined that exhibits a higher similarity improves the accuracy of the classification criteria and thus the ability of the classification method to identify the data set S.

[0134] Due to the fact that the training data was recorded during the correct operation of the facility and the measuring device 3, each characteristic C that can be determined based on the measured values m(ti) included in one of the reference sets conforms to the reference characteristic Cr. Thus, the reference characteristic Cr can be determined based on and / or is determined based on the reference set. Additionally, the classification criteria corresponding to the reference characteristic Cr can be determined based on and / or are determined based on the reference set.

[0135] As an alternative, other ways of identifying the specific operating phase Ps, another type of classification method, and / or another method of determining the classification method and / or the classification criteria can be used instead. For example, the computing unit 29 is implemented, for example, to perform a process of machine learning. In this case, the computing unit 29 learns the classification method based on the training data and then identifies the data set S included in the data D by executing the learned method.

[0136] After determining the classification method, the data set S included in the data D is identified by executing this method. For example, the classification method is performed, for example, by determining finite time intervals (where the measured values m(ti) measured during each time interval satisfy the classification criteria characteristics for the specific operating phase Ps) and by identifying the portions of the data D measured during these finite time intervals as the data set S. Determining the measured values m(ti) measured during one of the finite time intervals and satisfying the classification criteria is performed, for example, by executing at least one of the following: correlation analysis, pattern recognition methods, autocorrelation analysis, and / or another data analysis method capable of identifying the data set S that satisfies the classification criteria.

[0137] Based on the data set S identified by a classification method, a time series ts of a compliance indicator I is determined that indicates the degree of compliance of at least one property of the characteristic C indicating the measured values m(ti) included in each data set S with the corresponding reference characteristic of the reference characteristic Cr. This time series ts is continuously extended based on the continuously recorded data D during the operation of the facility. Further, it is applied at least once to determine the remaining time RT remaining until the degree of compliance indicated by the compliance indicator I determined by the measured values m(ti) measured during a future occurrence based on a specific operation phase Ps drops below a predetermined minimum compliance degree Imin. The determination of each remaining time RT is performed, for example, by performing a method of time series prediction based on the compliance indicator I included in the time series ts.

[0138] Each remaining time RT is, for example, at a future time point t RT (at which time point, the compliance indicator I(t RT ) will drop below the minimum compliance degree Imin) and / or in the form of the remaining remaining time interval RTI until this time point t RT is determined.

[0139] The determination of the time series ts is performed, for example, by a computing unit 29, which is implemented to determine the time series ts based on the continuously recorded data D provided to it or based on the data set S provided to it and at least one property of the reference characteristic Cr determined or provided to it. The computing unit 29 is, for example, trained to perform or is designed to learn and perform the determination of the time series ts. In addition, the computing unit 29 is, for example, implemented to determine the remaining time RT.

[0140] After determining the remaining time RT, an output notifying the corresponding remaining time RT is provided. For example, the output is provided, for example, in the form of an email or message automatically generated by the computing unit 29 and is dispatched to a predetermined recipient or a predetermined device, such as a hypercoordination unit 27, a computer, or a mobile device (such as a cellular phone, a tablet computer, or a service tool).

[0141] The method according to the invention provides the above-mentioned advantages. The individual steps of the method can be implemented in different ways without departing from the scope of the invention. Several alternative embodiments will be described in more detail below.

[0142] Regarding the determination of a time series ts performed based on a data set S identified by a classification method, different methods can be applied. As an option, the time series ts is determined, for example, by determining one of the compliance indicators I of the time series ts for each data set S identified by the classification method. In this case, the time associated with each of the compliance indicators I is, for example, a timestamp determined based on the time period during which the measured values m(ti) included in the respective data sets S are measured, such as a timestamp given by a start time, an end time, or a time point in the middle of the time period.

[0143] For example, these compliance indicators I are each determined to be equal to a quantitative measure of the similarity of the overall characteristic C exhibited by the measured values m(ti) included in the corresponding data set S and the overall reference characteristic Cr. This type of compliance indicator I is particularly suitable when the reference characteristic Cr only includes the reference patterns and / or reference distributions described above. In this case, the compliance indicators I are each determined based on the similarity of the pattern described by the measured values m(ti) included in the respective data sets S to the reference pattern and / or the similarity of the distribution of the measured values m(ti) included in the corresponding data set S to the reference distribution.

[0144] When the compliance indicators I are determined based on the degree of similarity between the corresponding overall characteristic C and the overall reference characteristic Cr, they can be determined quantitatively, for example, in the form of a percentage, where in the ideal case where the characteristic C and the reference characteristic are the same, a target value of 100% compliance is achieved.

[0145] Figure 4 An example of a set of measured values m(ti) measured during a specific operation phase Ps is shown. In this example, the first filling phase Pf1 that recurs in the examples shown in Figure 1 and Figure 3 is applied as the specific operation phase Ps. Each set includes the measured values m(ti) included in one of the data sets S that have been measured during one occurrence of the first filling phase Pf1. The shown excerpt includes three examples of sets of measured values m(ti) that have been measured during the training time interval TI and three examples of sets of measured values m(ti) that have been measured after the training time interval TI. Figure 5 An example of the corresponding time series ts := ts(I(t)) given by the degree of compliance of the characteristic C of the measured values m(ti) included in one of the data sets S with the reference characteristic Cr is shown. In this example, the compliance indicator I determined based on the data set S including the measured values m(ti) measured after the training time interval TI decreases over time. As Figure 5As shown by the arrow in, based on this time series ts, for example, a time series prediction method is applied to predict the remaining time RT until the compliance indicator I will drop below the minimum compliance level Imin.

[0146] As an alternative option, the compliance indicator I is determined, for example, based on the degree of compliance of one or at least two of the properties of characteristic C with respect to the corresponding reference properties. This option is particularly suitable when the measured value m(ti) measured during a specific operating phase Ps can be described by a deterministic function f(t) of time t and a set of coefficients. The function f(t) of time t can be determined, for example, based on and / or determined based on training data. In this case, characteristic C includes, for example, a set of fitting coefficients determined by fitting the measured value m(ti) to the function f(t) of time, and the reference characteristic Cr includes the corresponding reference coefficients. In this case, the degree of compliance is quantitatively determined, for example, based on the deviation between the fitting coefficients and the reference coefficients. For example, when this type of compliance indicator I is applied in the example shown in Figure 4 each compliance indicator I determined for one in the data set S is determined, for example, as the deviation between the slope of the straight line G fitted to the measured value m(ti) included in the data set S and the corresponding reference slope included in the reference characteristic Cr.

[0147] When only one property (e.g., slope) is applied, the compliance indicators I are each given, for example, by this property determined based on the corresponding data set S. Thus, the time series ts is the time series of this property, for example, the time series of the slope. In this case, when the deviation between the property and the reference property exceeds the corresponding deviation range, when the property is the same as the reference property and the compliance indicator I drops below the minimum compliance level, the target value of 100% compliance is reached.

[0148] When two or more properties are applied, the compliance indicators I are each given, for example, by a quantitative measure of the deviation between the properties determined based on the corresponding data set S and the reference properties. In this case, when the compliance indicator I given by these deviations exceeds the corresponding deviation range, in the ideal case where the deviation is zero (because all the properties determined based on the corresponding data set S are the same as the reference properties and the compliance indicator I drops below the minimum compliance level), the target value of 100% compliance is achieved.

[0149] As another option (available when a specific operating phase Ps is the operating phase), in which the variable should be equal to the same constant K during each occurrence of this operating phase, including the measured values m(ti) in the data set S or the deviation d(ti):=m(ti)-Kr between these measured values m(ti) and the reference value Kr of the constant K can be used as a compliance indicator I. In the first case, the time series ts is the time series ts(m(ti)) of the measured values m(ti) measured during the specific operating phase Ps. Here, when the measured value m(ti) exceeds the indicator value range ΔK (for example, the indicator value range ΔK in the case of a given ΔK:=[Kr-ΔKr;Kr+ΔKr] including the reference constant Kr), the compliance indicator I (each given by one of these measured values m(ti)) drops below the minimum compliance level Imin. In the second case, the time series ts is the time series ts(d(ti)) of the deviations d(ti), and when the deviation d(ti) exceeds the corresponding deviation range DR (for example, DR=[-ΔKr;+ΔKr]), the compliance indicator I given by each of these deviations d(ti) drops below the minimum compliance level Imin.

[0150] Figure 6 An example of an excerpt of the time series ts of the measured values m(ti) is shown, where each measured value m(ti) is included in one of the data sets S identified by the classification method and measured during one of the specific operating phases Ps. Since the variable measured during the specific operating phase Ps should be equal to the constant K, the reference constant Kr of the measured values m(ti) included in the reference characteristic Cr corresponds to this constant K. For example, the time series ts is determined, for example, by discarding all data elements including data D except for the measured values m(ti) included in the identified data set S. Figure 6 The excerpt shown includes three examples of sets of measured values m(ti) measured during the training time interval TI and two examples of sets of measured values m(ti) measured after the training time interval TI. Based on this time series ts, a method of time series prediction can be applied to predict the remaining time RT until the compliance indicator I given by the measured values m(ti) will drop below the minimum compliance level Imin. As Figure 6 shown, this remaining time RT is given, for example, by the remaining time RT until the measured value m(ti) to be measured during a future occurrence of the specific operating phase Ps will exceed the indicator value range ΔK. In the example shown, the measured values m(ti) measured after the training time interval TI increase over time and will therefore exceed the upper limit Kr+ΔKr of the value range ΔK.

[0151] If the target value of the constant K is known, the reference constant Kr can be determined to be equal to the target value. As an alternative, which is available whether the target value is known or unknown, the reference constant Kr is determined, for example, based on the measured values m(ti) that have been measured during the training time interval TI and are included in the time series ts. In this case, the reference constant Kr is determined, for example, to be equal to the average or mean value of the measured values m(ti) that have been measured during the training time interval TI and are included in the time series ts.

[0152] Regardless of the type of the applied indicator I, at least one of the remaining times RT is preferably determined by performing a method of time series prediction. As an option, at least one of the remaining times RT is determined, for example, by performing the method of time series prediction described in European Patent Application No. 20168733.2 filed on April 8, 2020, which is incorporated herein by reference. In this case, the execution of this method includes the following method steps:

[0153] a) For each indicator I compliance included in the time series ts, determine the deviation d(ti) between the corresponding indicator I compliance and the target value of the compliance degree (e.g., the target value 100%),

[0154] b) Filter the deviation d(ti),

[0155] c) Based on the deviation d(ti) and the filtered deviation FD(ti), determine the noise superimposed on the filtered deviation FD(ti), and

[0156] d) At the end of at least one monitoring time interval MTI (during which three or more indicator I compliances have been determined and no indicator I compliance is below the minimum compliance degree Imin), determine the remaining time RT until the deviation d(ti) will exceed the deviation range DR corresponding to the minimum compliance degree Imin.

[0157] The deviation range DR is determined based on the minimum compliance degree Imin such that when the compliance degree indicated by the indicator I drops below the minimum compliance degree Imin, the deviation d(ti) exceeds the deviation range DR.

[0158] When the indicator I indicates the compliance degree of a single property of the characteristic C with the corresponding reference property, and the time series ts is the time series ts of this property, the deviation d(ti) determined in step a) is given by the deviation between the corresponding property and the reference property.

[0159] When the variable should be constant during each specific operating phase Ps and the time series ts is the time series ts(m(ti)) of measured values m(ti) measured during the specific operating phase Ps, as Figure 6 shown, the deviation d(ti) determined in step a) is given by the deviation d(ti) between the measured value m(ti) and the reference constant Kr. Accordingly, when the measured value m(ti) is equal to the reference constant Kr and the compliance indicator I given by the measured value m(ti) drops below the minimum compliance level Imin, when the deviation d(ti) exceeds the deviation range DR (e.g., DR := [-ΔKr; +ΔKr]) described above in the context of the time series ts(d(ti)) of the deviation d(ti)), the target value of 100% compliance is reached.

[0160] This method is shown in Figure 7 where Figure 6 the deviation d(ti) between the measured value m(ti) of the time series ts shown therein and the reference constant Kr is indicated by cross marks, and where the filtered deviation FD(ti) is shown in the form of a time function FD(t). As described in European Patent Application No. 20168733.2 filed on April 8, 2020, the remaining time RT is determined as follows: for at least two different deviation pairs k (each including a first deviation d1 k ) determined based on the filtered deviations FD(t1 k ) and FD(t2 k ), FD(t2 k ) and a second deviation d2 k ), the simulated value SRTk of the remaining time RT is determined by performing a Monte Carlo simulation based on the noise and the corresponding deviation pair k. As shown by the double-arrow dotted line in k ), the deviation pair k is determined, for example, such that the time elapsed between the first and second deviations d1 Figure 7 of consecutive deviation pairs k k ), d2 k ), d2 k ), forms a sliding time window of fixed or variable length sliding along the monitoring time interval MTI. As an alternative (not shown), the deviation pair k is determined, for example, such that the second deviation d2k(t2k) of each pair k is given by one of the last filtered deviations FD(tn) or the last filtered deviation FD(ti) included in the monitoring time interval MTI. k ) forms a sliding time window of fixed or variable length sliding along the monitoring time interval MTI. As an alternative (not shown), the deviation pair k is determined, for example, such that the second deviation d2k(t2k) of each pair k is given by one of the last filtered deviations FD(tn) or the last filtered deviation FD(ti) included in the monitoring time interval MTI.

[0161] Regarding the Monte Carlo simulation performed to determine the simulated time SRT k , simulation methods known in the art can be applied. Figure 8Shows an example of a simulation method performed on one of the biases k. It includes a first step, namely, based on the noise indicated by N, the first and second biases d1 k (t1 k ), d2 k (t2 k ), and the corresponding times t1 k , t2 k of this bias pair k, generate a statistically representative quantity of the first and second random bias pairs [E1(t1 k ):=d1 k (t1 k ) + e1; E2(t2 k ):=d2 k (t2 k ) + e2], where each random bias E1, E2 is equal to the corresponding bias d1 k (t1 k ), d2 k (t2 k ) and the sum of the random additives e1, e2 considering the noise. In this regard, the random additives e1, e2 are preferably generated according to a probability distribution reflecting the nature of the noise. For each pair of random biases [E1(t1 k ); E2(t2 k )], the crossing time t x is determined as the time when the straight line that crosses the first random bias E1(t1 k ) at the first time t1 k and crosses the second random bias E2(t2 k ) at the second time t2 k will exceed the bias range DR. Some examples of such determined straight lines are shown as dashed lines in Figure 8 , and the bias range DR (represented by the bias upper and lower limits -ΔKr, +ΔKr shown in Figure 8 ) is shown. Based on the crossing time t x , determine the probability density function PDF(t x ) of the crossing time t x , and based on the probability density function PDF(t x ) of the crossing time t x determine the simulated value SRT k . To this extent, a confidence level γ can be set, and the simulated value SRT S can be determined to be equal to the time T k predicted based on the probability density function PDF(t x ) by solving the following equation for T S(γ), at which time the deviation d(ti) will exceed the deviation range DR with a given confidence level γ:

[0162]

[0163] Thereafter, based on the simulated value SRT determined for each deviation pair k (e.g., as the average or weighted average of the simulated values SRT k ), the remaining time RT is determined. In the case of a weighted average, preferably based on the first and second deviations d1 k corresponding to the respective deviation pair k k (t1 k ), d2 k (t2 k ), the times t1 k and t2 k are used to determine, for example, by the following formula, the weighting factor applied to the simulated value SRT k :

[0164]

[0165] This method of time series prediction has the advantage that each simulated value SRTk takes into account the average rate of change of the deviation d(ti) that occurs in the time interval elapsed between the first and second deviations d1 k (t1 k ), d2 k (t2 k ) in the respective deviation pair k. Thus, in a combined manner, they truly take into account the temporal correlation of the deviation d(ti) in all time intervals covered by the deviation pair k, even if the temporal correlation changes during the monitoring time interval MTI. Another advantage is that this method does not require the measured values m(ti) included in the time series ts to be available at a fixed rate, nor does it require measurements to be made at consecutive previously known or predetermined time points. In addition, ensuring that each simulation is based on noise takes into account the uncertainty inherent in determining the deviation d(ti). Since the noise is determined based on the deviation d(ti) and the filtered deviation FD(ti), no additional knowledge about the uncertainty inherent in determining the measured value m(ti) and the reference constant Kr is required.

[0166] As an alternative, another method of time series prediction can be applied alternatively or in combination.

[0167] Regardless of the method applied to determine the remaining time RT, as an option, this method can be followed at the time point t at which the degree of compliance will drop below the minimum degree of compliance according to the previously determined remaining time RT RTThe determination of at least one remaining time RT, as described above before, is further improved. In this case, the output of the notified remaining time RT is preferably updated accordingly each time a new remaining time RT has been determined.

[0168] As an additional or alternative option, the method can be further improved by calibrating the measuring device 3 at or before the time point t at which the compliance indicator I drops below the minimum compliance level Imin according to the previously determined remaining time RT. During calibration, the measurement error of the measuring device 3 is determined. This measurement error is then compared with a predetermined threshold. In the case where the measurement error is greater than the threshold, the impaired measurement property of the measuring device 3 is determined as the root cause for the compliance indicator I dropping below the minimum compliance level Imin. In this case, the measuring device 3 is preferably adjusted, repaired or replaced, and the method starts over from the beginning with the measuring device 3 obtained in such a way that it meets the specified measurement accuracy. In the case where the measurement error is less than the threshold, the impaired operation of the facility is determined as the root cause for the compliance indicator I dropping below the minimum compliance level Imin. In this case, the fault causing the impaired operation is preferably determined, and the corresponding remedial measures to solve the fault are preferably applied. RT As an option, the determination of the time series ts can be further refined by identifying datasets S that include and discard potentially contaminated measurement values m(ti). For example, potentially contaminated measurement values m(ti) include, for example, boundary values measured at the start and / or end of a corresponding specific operation phase Ps. These boundary values may be contaminated due to side effects that occur during the transition from the previous operation phase to the corresponding specific operation phase Ps and during the transition from the corresponding specific operation phase Ps to the next operation phase. The identification and discarding of potentially contaminated measurement values m(ti) is, for example, performed by a computing unit 29 that is implemented as, for example, trained or designed to learn to identify and discard potentially contaminated measurement values m(ti). To this extent, machine learning methods can be applied, such as machine learning methods for anomaly detection or novelty detection, such as Isolation Forest, Local Outlier Factor, Elliptic Envelope or One-Class Support Vector Machine. Regardless of how the potentially contaminated measurement values m(ti) are identified, the identification and discarding of potentially contaminated measurement values m(ti) is preferably also applied during the determination of the classification method, for example by identifying and discarding potentially contaminated measurement values included within a subset of the group Gj determined based on the training data.

[0169]

[0170] ​As an additional or alternative option, the identification of a specific operating phase Ps, the determination of characteristics C and / or reference characteristics Cr, the determination of a classification method, the identification of a data set S, the execution of a classification method, and / or the determination of a time series ts can be further improved by considering at least one parameter measured by the measuring device 3 during the operation of the facility. As an option, the parameters include, for example, at least one parameter measured by a sensor of the measuring device 3, at least one parameter measured and applied by the measuring device 3 to determine a measured value m(ti) and / or to compensate for a parameter-dependent measurement error of the measuring device 3. For example, the parameters include, for example, the temperature measured by the temperature sensor 35 of the measuring device 3. When the measuring device 3 is an electrochemical measuring device for measuring the pH value of a medium or the concentration of an analyte contained in the medium, such as Figure 2 the pH sensor shown in Figure 2 , the parameters include, for example, a parameter given by the electrode potential Uel of the measuring electrode 15 extending into the measuring chamber 9 and / or a parameter given by the impedance Z of the membrane 11 enclosing the measuring chamber 9. In this case, the measuring electronics 17 of the measuring device 3 is, for example, implemented to determine the electrode potential Uel of the electrode 19 connected to the measuring electronics 17 and / or is implemented to determine the membrane impedance Z. For example, the membrane impedance Z is determined, for example, by applying an alternating voltage to the reference electrode 23 such that the electrode potential Uel changes according to the alternating voltage and the membrane impedance Z. In this case, the membrane impedance Z is determined, for example, based on the dependence of the electrode potential Uel on the alternating voltage and the membrane impedance Z, which occurs when the alternating voltage is applied to the reference electrode 23.

[0171] When considering at least one parameter, the continuously recorded data D additionally includes the parameter values p(ti) of the (multiple) parameters measured by the measuring device 3 during the operation of the facility and their measurement times ti. In this case, at least one of the following is performed based on the measured values m(ti) and the measured parameter values p(ti) included in this data D as described above: the identification of a specific operating phase Ps, the determination of a classification method, the identification of a data set S, the execution of a classification method, and the determination of a time series ts. When the measured parameter values p(ti) are applied to determine and / or execute a classification method, the classification criteria include at least one criterion regarding the measured values m(ti) and the measured parameter values p(ti) measured during a specific operating phase Pc. These classification criteria are determined, for example, based on a reference set determined as described above that respectively includes the measured values m(ti) and the measured parameter values p(ti). These classification criteria include, for example, criteria regarding values, value ranges, distributions, and / or patterns described by the measured values m(ti) and / or the measured parameter values p(ti) measured during a specific operating phase Pc. Additionally, determining the time series ts based on the measured parameter values p(ti) has the advantage of providing a more accurate and reliable identification of the data set S measured during a specific operating phase Ps.

[0172] As an option, a measured parameter value p(ti) is applied, for example, relative to a characteristic C, a reference characteristic Cr, and a compliance indicator I. In this case, a specific operating phase Ps is considered to be a specific operating phase Ps during which the measured value m(ti) and the measured parameter value p(ti) measured during the specific operating phase Ps exhibit a characteristic C that distinguishes them from the measured values m(ti) and the measured parameter values p(ti) measured during other time periods. Just like the characteristic C exhibited by the measured value m(ti), when the facility is operating correctly and the measuring device 3 complies with the specified measurement accuracy, the characteristic C exhibited by the measured value m(ti) and the measured parameter value p(ti) complies with the corresponding reference characteristic Cr. Therefore, the compliance indicator I indicating the degree of compliance of at least one property of the characteristic C of the measured value m(ti) and the measured parameter value p(ti) included in the set S with the corresponding reference property of the reference characteristic Cr can be applied in the same way as described above for the compliance indicator I indicating the degree of compliance of at least one property of the characteristic C of the measured value m(ti).

[0173] In some applications, the operation of the facility includes two or more recurring different operating phases that are suitable for application as the specific operating phase Ps. As an option (available in these applications), at least one additional remaining time RT’ can be determined as described above based on at least one additional specific operating phase Ps’, during which the measured value m(ti) measured during the additional specific operating phase Ps’ exhibits a characteristic C’ that distinguishes these measured values m(ti) from the measured values m(ti) measured during other time periods, and during which, when the facility is operating correctly and the measuring device 3 complies with the specified measurement accuracy, the characteristic C’ complies with the reference characteristic Cr’. In this case, the method includes the following additional steps:

[0174] Based on the training data, determine an additional classification method capable of identifying an additional data set S’ included in the recorded data D that has been measured during one of the additional specific operating phases Ps’;

[0175] Execute the classification method and, based on the additional data set S’ identified by the additional classification method, determine an additional time series ts’ of the compliance indicator I’ indicating the degree of compliance of at least one property of the characteristic C’ of the measured value m(ti) included in the additional data set S’ with the corresponding reference property of the reference characteristic Cr’;

[0176] Based on the additional time series ts’ determine at least once the remaining time RT’ until the degree of compliance indicated by the compliance indicator I’ determined based on the measured value m(ti) to be measured during a future occurrence during the additional specific operating phase Ps’ drops below a predetermined additional minimum degree of compliance Imin’; and

[0177] Provide an output notifying the additional remaining time RT'.

[0178] Performing the method based on a specific operation phase Ps and at least one additional specific operation phase Ps' increases the number and availability of measurement values m(ti) based on which at least one of the remaining time RT and the additional remaining time RT' can be determined. This is particularly advantageous in applications where longer time intervals may occur between successive occurrences of the specific operation phase Ps.

[0179] List of reference signs

[0180] 1 Container 19 Reference chamber

[0181] 3 Measuring device 21 Electrolyte

[0182] 5 Medium 23 Reference electrode

[0183] 7 Pipe 25 Diaphragm

[0184] 9 Measuring chamber 27 Hypercoordination unit

[0185] 11 Ion-selective membrane 29 Computing unit

[0186] 13 Electrolyte 31 Memory

[0187] 15 Measuring electrode 33 Edge device

[0188] 17 Measuring electronic device 35 Temperature sensor

Claims

1. A method for predictive monitoring of a variable of a medium (5) and the measurement accuracy of a measuring device (3), the medium (5) being located in a container (1) of a facility, the measuring device (3) measuring the variable during operation of the facility and providing a measured value (m(ti)) of the variable, wherein the facility operates independently of the measured value (m(ti)) by neither adjusting nor controlling the operation of the facility based on the measured value (m(ti)), and wherein the operation of the facility includes recurring specific operation phases (Ps), during which the measured values (m(ti)) measured during the specific operation phases (Ps) exhibit a characteristic (C) that differentiates these measured values (m(ti)) from the measured values (m(ti)) measured during other time periods, and wherein when the facility is operating correctly and the measuring device (3) meets a specified measurement accuracy during the corresponding specific operation phase (Ps), the characteristic (C) conforms to a reference characteristic (Cr), the method comprising the steps of: Installing the measuring device (3) at the facility; During operation of the facility, continuously recording data (D), the data (D) including the measured values (m(ti)) measured by the measuring device (3) and their measurement times (ti); Based on training data included in the data (D), determining a classification method capable of identifying a data set (S) included in the recorded data (D), the training data having been recorded during a training time interval (TI), during which the facility was operating correctly and during which the measuring device (3) met the specified measurement accuracy, the data sets (S) having each been measured during one of the specific operation phases (Ps); Executing the classification method and determining, by executing the classification method, data sets (S), each data set including measured values (m(ti)) that have been measured during one of consecutive specific operation phases; Determining a time series (ts) of compliance indicators (I) by determining at least one compliance indicator (I) for each data set (S), wherein each compliance indicator (I) corresponds to the degree of compliance of at least one property of the characteristic (C) of the measured values (m(ti)) included in the corresponding data set (S) with a corresponding reference property of the reference characteristic (Cr); Based on the time series (ts) of the compliance indicators (I), determining, by performing a time series prediction method, at least once the remaining time (RT) remaining until the degree of compliance indicated by the compliance indicator (I) determined based on the measured values (m(ti)) to be measured during a future occurrence of the specific operation phase (Ps) drops below a predetermined minimum degree of compliance (Imin); and Providing an output notifying of the remaining time (RT).

2. The method according to claim 1, wherein, The method is a computer-implemented method.

3. The method according to claim 1, wherein, The specific operation phase (Ps) is pre-determined based on available information regarding the operation of the facility or is identified based on the training data.

4. The method according to claim 1, wherein The specific operation phase (Ps) is: a) an operation phase that occurs during each execution of a predetermined intermittent process, where the intermittent process is repeatedly executed on or by the facility during the operation of the facility; b) an operation phase in which the variable should be equal to a constant (K); c) a cleaning phase (Pc), where the variable of the medium (5) measured by the measuring device (3) during each cleaning phase (Pc) is the variable of the same cleaning agent applied to clean the container (1) during each cleaning phase (Pc), or d) an empty phase (Pe), where the variable of the medium (5) measured by the measuring device (3) is the variable of the gas or air contained in the empty container (1) during each empty phase (Pe).

5. The method according to any one of claims 1 to 3, wherein: The characteristic (C) is determined based on at least one of the training data and the reference characteristic (Cr), and / or includes at least one property of the measured values (m(ti)), and the property includes at least one of the following: the value of the measured value (m(ti)); the slope of the measured value (m(ti)); at least one fitting coefficient that can be determined by fitting the measured value (m(ti)) to a function of time (f(t)), and a set of one or more coefficients that describe the measured values (m(ti)) measured during the specific operation phase (Ps); the value range in which the measured value (m(ti)) appears; the distribution of the measured value (m(ti)); the pattern described by the measured value (m(ti)); at least one property of the model property corresponding to a model, a deterministic model, a statistical model, or a hybrid model including a deterministic model component and a statistical model component for the measured values (m(ti)) measured during the specific operation phase (Ps), and at least one other property; and The reference characteristic (Cr) is determined based on the training data and / or includes reference properties for each property of the characteristic (C), where the reference property represents the measured value (m(ti)) measured during one of the specific operation phases (Ps), while the facility is operating correctly and the measuring device (3) meets the specified measurement accuracy. The reference property includes at least one of the following: a reference value (Kr) for the measured value (m(ti)), a reference slope, a set of one or more reference coefficients, a reference pattern, a reference distribution, a reference property for at least one model property, and at least one other reference property that would be expected for the measured value (m(ti)) measured during the specific operation phase (Ps).

6. The method according to any one of claims 1 to 3, wherein: Determining that the classification method includes the following steps: identifying the specific operation phase (Ps), identifying the specific operation phase (Ps) based on the training data, or identifying the specific operation phase (Ps) based on potential candidates for the specific operation phase (Ps) determined based on the training data and available information regarding the operation of the facility; Performing the classification method based on classification criteria determined for the specific operation phase (Ps), the classification criteria including at least one of the following: at least one criterion for an expected value or value range of measurement values (m(ti)) measured during the specific operation phase (Ps), at least one criterion for a pattern described by the measurement values (m(ti)) expected to be measured during the specific operation phase (Ps), at least one criterion for the distribution of the measurement values (m(ti)) expected to be measured during the specific operation phase (Ps), at least one criterion related to the degree of compliance with at least one property included in the characteristic (C) and corresponding reference properties included in the reference characteristic (Cr) of the measurement values (m(ti)), at least one criterion related to the model properties of the model for the measurement values (m(ti)) measured during the specific operation phase (Ps), and at least one other criterion, and / or Determining the data set (S) included in the data (D) and satisfying the classification criteria applied to identify the data set (S) by performing at least one of the following: correlation analysis, pattern recognition methods, autocorrelation analysis, and at least one other data analysis method capable of identifying the data set (S) that satisfies the classification criteria.

7. The method according to any one of claims 1 to 3, comprising the following steps: Identifying at least one group (Gj) of subsets of the training data, where each subset consists of data measured during a subset time interval, and where subsets belonging to the same group (Gj) exhibit a similarity greater than or equal to the minimum similarity required for the subsets to be considered to belong to the same group (Gj), where identifying at least one group (Gj) of subsets is performed by performing at least one of the following: correlation analysis, pattern recognition methods, autocorrelation analysis, and at least one other data analysis method capable of identifying subsets representing the same operation phase; For at least one of the groups (Gj), applying the subsets included in the corresponding group (Gj) as a reference set representing the same operation phase; Determining an operation phase for which a reference set has been determined as the specific operation phase (Ps), Determining the characteristic (C) and the reference characteristic (Cr) based on the reference set representing the specific operation phase (Ps), and Determining at least one of a classification method and classification criteria for identifying the data set (S) included in the recorded data (D) based on the reference set for the specific operation phase (Ps).

8. The method according to claim 7, wherein reference sets have been determined for at least two different operating phases, and the method comprises at least one of the following steps: determining the specific operating phase (Ps) as one of these different operating phases for which a reference set has been determined and which has a longer duration and / or a higher occurrence frequency than at least one other operating phase, and determining the specific operating phase (Ps) such that the reference set for the specific operating phase (Ps) has a higher similarity than the reference sets determined for at least one other operating phase.

9. The method according to any one of claims 1 to 3, wherein: the time series (ts) is determined by determining, for each data set (S) identified by the classification method, a quantitative measure of the similarity of one of the compliance indicators (I) to the overall characteristic (C) and the overall reference characteristic (Cr) exhibited by the measured values (m(ti)) included in the respective data set (S); or wherein: the compliance indicator (I) indicates the degree of compliance of one of the properties of the characteristic (C) with the corresponding reference property, and each compliance indicator (I) of the time series (ts) is given by the property of the measured value (m(ti)) included in one of the data sets (S); or wherein: a) the specific operating phase (Ps) is an operating phase during each occurrence of which the variable should be equal to the same constant (K), b) the compliance indicator (I) is given by the measured value (m(ti)) included in the data set (S); and c) when the compliance indicator (I) given by one of the measured values (m(ti)) exceeds the indicator value range (ΔK) for the target value for the constant (K) or the reference constant (Kr) for the constant (K) included in the reference characteristic (Cr), the compliance indicator drops below the minimum degree of compliance (Imin); or wherein: a) the specific operating phase (Ps) is an operating phase during each occurrence of which the variable should be equal to the same constant (K), b) the compliance indicator (I) is given by the deviation (d(ti)) between the measured value (m(ti)) included in the data set (S) and the target value for the constant (K) or the reference constant (Kr) for the constant (K) included in the reference characteristic (Cr); and c) when the compliance indicator (I) given by one of the deviations (d(ti)) exceeds the corresponding deviation range (DR), the compliance indicator (I) drops below the minimum degree of compliance (Imin).

10. The method according to any one of claims 1 to 3, wherein: the training data is labeled as training data including the measured values (m(ti)) and the respective operating phases during which they were measured, and Perform at least one of the following steps by performing a supervised learning method: identify the specific operation phase (Ps), determine the characteristic (C), determine the reference characteristic (Cr), and determine the classification method.

11. The method according to any one of claims 1 to 3, additionally comprising the step of determining and discarding at least one of the following before determining the time series (ts): measurements of potential contamination (m(ti)) and measurements of potential contamination (m(ti)) given by marginal values measured at the start and end of a specific operation phase (Ps) included in the identified data set (S).

12. The method according to any one of claims 1 to 3, comprising the step of providing continuously recorded data (D) to a computing unit (29), wherein the computing unit (29): is implemented to perform, trained to perform, and / or designed to learn and perform at least one of the following: identify the specific operation phase (Ps) and determine the classification method based on the data (D) provided to it; determine and perform the classification method based on the data (D) provided to it; determine the time series (ts); and / or determine the remaining time (RT).

13. The method according to any one of claims 1 to 3, wherein: the facility is implemented to perform a predetermined task or process and / or repeatedly perform a predetermined intermittent process; and / or the measuring device (3) is an electrochemical measuring device for measuring the concentration of an analyte contained in the medium (5) or a pH sensor for measuring the pH value of the medium (5).

14. The method according to any one of claims 1 to 3, wherein: the measuring device (3) measures at least one parameter; the continuously recorded data (D) includes measurement parameter values (p(ti)) of one or more parameters measured and provided by the measuring device (3) and their measurement times (ti); and perform at least one of the following based on the measured values (m(ti)) and measurement parameter values (p(ti)) included in the training data: identify the specific operation phase (Ps), determine the characteristic (C), determine the reference characteristic (Cr), determine the classification method, perform the classification method, and determine the time series (ts).

15. The method according to claim 14, wherein the at least one parameter includes at least one of the following: at least one parameter measured by a sensor of the measuring device (3); at least one parameter applied by the measuring device (3) to determine the measured value (m(ti)) of the variable; at least one parameter applied by the measuring device (3) to compensate for a parameter-dependent measurement error; the temperature measured by a temperature sensor (35) of the measuring device (3); and The electrode potential (Uel) of the measuring electrode (15) of the measuring device (3) and / or the electrical impedance (Z) of the ion-selective membrane (11) of the measuring device (3), wherein the measuring device (3) is an electrochemical measuring device for measuring the concentration of an analyte contained in the medium (5) or a pH sensor for measuring the pH value of the medium (5), and comprises: a measuring chamber (9) enclosed by the ion-selective membrane (11), the ion-selective membrane having an inner surface exposed to an electrolyte (13) located inside the measuring chamber (9) and an outer surface exposed to the medium (5), and the measuring electrode (15) being immersed in the electrolyte (13).

16. The method according to any one of claims 1 to 3, wherein each remaining time (RT) is determined by performing a method of time series prediction, the method comprising the steps of: for each compliance indicator (I) included in the time series (ts), determining the deviation (d(ti)) between the corresponding compliance indicator (I) and a target value of the degree of compliance or a target value of 100% degree of compliance, filtering the deviation (d(ti)), determining noise superimposed on the filtered deviation (FD(ti)) based on the deviation (d(ti)) and the filtered deviation (FD(ti)), and at the end of at least one monitoring time interval (MTI), determining the remaining time (RT) as the remaining time (RT) until the deviation (d(ti)) will exceed a deviation range (DR), during which at least three compliance indicators (I) included in the time series (ts) have been determined and none of the compliance indicators (I) is below the minimum degree of compliance (Imin), wherein the deviation range (DR) is determined based on the minimum degree of compliance (Imin) such that the deviation (d(ti)) exceeds the deviation range (DR) when the degree of compliance indicated by the compliance indicator (I) drops below the minimum degree of compliance (Imin), and wherein the remaining time (RT) is determined by the following method: For each of at least two different pairs of deviations (k) each including a first deviation (d1 k )(t1 k )) and a second deviation (d2k(t2k)) determined based on a filtered deviation (FD(t1 k ), FD(t2 k )) included in the monitored time interval (MTI), a simulated value (SRT k ) of the remaining time (RT) is determined by performing a Monte Carlo simulation based on the noise and the corresponding pair of deviations (k), and Based on the average or weighted average of the simulated values (SRT k ) determined for each deviation pair (k), determine the remaining time (RT).

17. The method according to any one of claims 1 to 3, comprising the steps of: Calibrate the measuring device (3) at or before the time point (t RT ) at which the degree of compliance indicated by the compliance indicator (I) will drop below the minimum degree of compliance (Imin) according to the previously determined remaining time (RT); during calibration, determining the measurement error of the measuring device (3); in the case where the measurement error is less than a predetermined threshold, performing at least one of the following: determining the impaired operation of the facility as the root cause leading to the degree of compliance dropping below the minimum degree of compliance (Imin), and determining the fault causing the impaired operation and applying a remedial measure to solve the fault; and In the case where the measurement error is greater than a predetermined threshold, at least one of the following steps is performed: determining that the impaired measurement property of the measuring device (3) is the root cause of the compliance level dropping below the minimum compliance level (Imin), and adjusting, repairing or replacing the measuring device (3), and restarting the method from the beginning by installing a measuring device (3) that complies with the specified measurement accuracy for it.

18. The method according to any one of claims 1 to 3, wherein the operation of the facility includes additional specific operation phases (Ps') that occur repeatedly, wherein the measured values (m(ti)) measured during the additional specific operation phases (Ps') exhibit a characteristic (C') that distinguishes these measured values (m(ti)) from the measured values (m(ti)) measured during other time periods, and wherein when the facility is operating correctly and the measuring device (3) complies with the specified measurement accuracy during the corresponding additional specific operation phase (Ps'), the characteristic (C') complies with a reference characteristic (Cr'), and the method includes the following steps: Based on the training data, determining an additional classification method capable of identifying an additional data set (S') of measured values (m(ti)) each of which has been measured during one of the additional specific operation phases (Ps') included in the recorded data (D); Performing the additional classification method, and based on the measured values (m(ti)) included in the additional data set (S') identified by the additional classification method, determining an additional time series (ts') of a compliance indicator (I') indicating the degree of compliance of at least one property indicating the characteristic (C') of the measured values (m(ti)) included in the additional data set (S') with the reference property corresponding to the reference characteristic (Cr'); Based on the additional time series (ts'), determining at least once the additional remaining time (RT') remaining until the degree of compliance indicated by the compliance indicator (I') to be determined based on the measured values (m(ti)) to be measured during future occurrences of the additional specific operation phase (Ps') drops below a predetermined additional minimum compliance level (Imin'); and Providing an output notifying of the additional remaining time (RT').

Images