Method for determining application-specific overall likelihood of a measured value of a measured variable

Abstract

The invention relates to a method of determining an application-specific overall likelihood of a measured value of a measured variable. A method of determining an application-specific likelihood of a measured value of a measured variable measured by a measuring device in a specific application comprises the steps of recording data comprising items of diagnostic information of the measuring device and comprising variable values of a specified variable of each measured variable, determining, for each measured variable, an overall likelihood of the current measured value of the measured variable based on a likelihood value, which is determined based on likelihood criteria comprising diagnostic criteria and application-specific threshold criteria and based on a reliability value indicating a statistical reliability of the current measured value of the measured variable, and providing the overall likelihood and / or an overall likelihood index determined based on the overall likelihood.

Description

Technical Field

[0001] The present invention relates to an application-specific total likelihood of a measurement of at least one measured variable measured by a measurement system installed in a particular application, wherein the measurement system includes at least one measuring device, and each measuring device measures at least one of at least one measured variable. Background Technology

[0002] Measurement systems, including those that measure the variables of interest in a specific application, are used in a wide variety of applications, including industrial and laboratory applications.

[0003] Measurements of analytes by measurement systems installed in specific applications are typically used to monitor, regulate, and / or control the operation of the analyte, the plant, or facility (e.g., a production facility), and / or at least one step of a process (e.g., a production process) performed at the application. For example, in chemical production processes, it is possible to monitor the concentrations of reactants used in the production process and / or the concentrations of analytes contained in precursors, intermediates, and / or precipitates produced by the process, and to schedule, regulate, and / or control a series of process steps based on the measured values ​​of the analyte. Other liquid analysis measurement systems that measure analytes such as pH, free chlorine concentration, and / or media turbidity are used, for example, in swimming pools, drinking water supply networks, and water purification plants to monitor, regulate, and / or control water quality.

[0004] Depending on the specific application, the efficiency and / or productivity of the production process, the quality of the products produced, the operational safety of facilities, plants and / or laboratories, and / or the quality of drinking water may depend on the measurement accuracy of the measuring equipment, the correct execution of the process at the specific application site, and / or the trouble-free operation of the plant or facility. Therefore, in most applications, the measurement accuracy of the measuring equipment and the conformity of the measured variable to the application-specific requirements are crucial.

[0005] As technical processes become increasingly complex and the number of measuring devices used to measure variables of interest in specific applications grows, assessing the reliability of measurements provided by a measurement system or whether they indicate a problem becomes increasingly difficult. Even when all measurements fall within the range specified for the application, there is a risk that the measurements may be compromised due to a problem associated with one of the measuring devices (e.g., due to measurement error), and / or that the measured variable may have been affected by a problem associated with the application (e.g., a failure or impairment of the process performed at the application). Furthermore, measurements may be temporarily compromised during cleaning or calibration of the measuring device, and during periods when the measuring device is temporarily removed from the measuring unit for visual inspection. Therefore, further use of measurements affected by one of these problems, such as for monitoring, conditioning, and / or control purposes, may pose a safety hazard, potentially reduce the quality of products manufactured at the application, and / or reduce the process efficiency of the process performed at the application.

[0006] By taking proactive measures, such as regularly maintaining and / or repeatedly verifying, calibrating, and / or adjusting measuring equipment, the probability of problems associated with measuring equipment can be significantly reduced. As another approach, the condition of the measuring equipment or its individual components can be monitored based on diagnostic parameters, and / or measuring equipment capable of performing self-diagnostics can be employed. For example, measuring equipment with a heartbeat function, provided by the Endress+Hauser Group, can be installed, enabling the measuring equipment to monitor its performance and / or provide status indicators indicating its condition.

[0007] Furthermore, EP 2 226 630 B1 discloses a method for determining the condition index of a water analysis apparatus based on its technical parameters. The technical parameters are measured by sensors, such as a humidity sensor measuring the humidity inside the apparatus housing and a sensor measuring the reagent level contained in a reservoir. For each technical parameter, the deviation between the measured parameter value and the corresponding reference value is determined. Subsequently, a deviation correlation value is determined for each deviation based on a parameter-specific deviation correlation function for the corresponding technical parameter. Finally, a condition index indicating the condition of the water analysis apparatus is calculated based on the deviation correlation value using an index function. The advantage of this method is that it considers the correlation of each deviation with respect to the overall condition of the entire apparatus. One disadvantage of this method is that, in addition to the analytical measuring equipment required to perform water analysis, the apparatus must also be equipped with a device for measuring the technical parameters, and damage to the analytical measuring equipment unrelated to the technical parameters indicating the condition of the apparatus, and damage to the water analysis measurement results determined by the device, may be overlooked.

[0008] Even if all measuring equipment is functioning correctly, problems can still arise in specific applications, such as due to malfunctions in the processes performed within the application and / or operational errors at the plant or facility. For example, a correctly measured high concentration of free chlorine in drinking water might be an indicator that excessive chlorine has been added.

[0009] In this context, DE 10 2008 042 969 A1 describes a method for detecting process states in alternating production processes executed using sensors that measure sensor values ​​representing process states. The sensor values ​​and / or process variables are evaluated simultaneously, and the diagnosis of process states and / or sensor states is performed based on pre-determined threshold ranges for sensor values ​​measured during individual process states. The advantage of this method is that the current process state can be determined based on simultaneous evaluation, and problems caused by defective sensors and / or erroneous performance of the process at the application can be detected based on deviations of measured values ​​from predefined threshold ranges for individual process states. However, the use of this method is limited to applications where sufficiently different predefined process states occur. Therefore, it is not suitable for applications where the measured variables vary in an unknown and / or continuous manner, making it impossible to identify sufficiently different process states. Summary of the Invention

[0010] The objective of this invention is to provide an application-specific likelihood of the measured value of a measured variable by a measurement system, which can be applied to a wider range of applications and provides likelihood results that realistically illustrate potential harms that may occur due to problems associated with the measuring device and problems associated with the application.

[0011] This objective is achieved by a method for determining the application-specific likelihood of measurements of at least one measured variable by a measurement system installed in a particular application; wherein the measurement system includes at least one measuring device, and each measuring device measures at least one of at least one measured variable, the method comprising the following steps:

[0012] For each measured variable, identify at least one piece of diagnostic information indicating the condition of the measuring device that measures the corresponding measured variable.

[0013] Specify the number of at least one variable to be measured or determined at the application, such that the specified variable is given by or includes each of the at least one measured variable;

[0014] For each measured variable, determine the likelihood criterion used to determine the likelihood value, including:

[0015] At least one diagnostic criterion is provided for determining a likelihood value indicating the likelihood of a measured value of a corresponding measured variable based on at least one item of diagnostic information indicating the condition of a measuring device measuring the corresponding measured variable, and

[0016] At least one threshold criterion is used to determine a likelihood value indicating the likelihood of a measurement of a corresponding measured variable based on the magnitude of the measurement relative to at least one application-specific threshold range applicable to the measurement of the corresponding measured variable.

[0017] Recorded data, which includes time series of values ​​for at least one item of diagnostic information and the variable values ​​for each specified variable, as well as the time of their determination or measurement; and

[0018] Based on the recorded data, perform the following steps at least once or repeatedly:

[0019] For each measured variable: a likelihood value is determined based on a likelihood criterion already established for the corresponding measured variable; given the measurements of at least one or all specified variables that have been measured prior to the current time interval and are given by or include the measurements of the corresponding measured variable, at least one likelihood measure is determined to indicate the reliability of one or more current measurements of the corresponding measured variable measured during the finite current time interval; and the total likelihood of the current measurements of the corresponding measured variable is determined based on the likelihood measure and likelihood value already established for the corresponding measured variable; and

[0020] Provide likelihood results, which include at least one of the following: total likelihood and total likelihood index determined based on total likelihood.

[0021] The advantage of this method is that, given the condition of the measuring equipment used to measure the variable being measured and the application-specific requirements described by the likelihood value and (multiple) likelihood measures, each total likelihood indicates the likelihood of the measured value of the corresponding variable being measured. Other advantages are that the method can be implemented based solely on data available for the specific application, and also in applications where sufficiently diverse process states are not available.

[0022] In some embodiments, for at least one or each measured variable, the likelihood measure includes at least one of the following: a first likelihood measure corresponding to the degree of agreement between the current measurement of the corresponding measured variable or the distribution of the current measurement and the distribution or one of a plurality of distributions of the measurement of the corresponding measured variable measured before the current time interval; and a second likelihood measure corresponding to the degree of agreement between the current measurement of the corresponding measured variable and the analysis result determined by multivariate analysis of the variable values ​​of at least two analytical variables determined before the current time interval; wherein the analytical variables are given by the corresponding measured variable and at least one variable or each of the other variables included in the selected variables.

[0023] In some embodiments, at least one or each total likelihood is determined by: a sum, weighted sum, product, or weighted product of likelihood measures and likelihood values ​​determined for the corresponding measured variable; or a minimum likelihood given by the smallest of the likelihood values ​​determined for the corresponding measured variable and a sum, weighted sum, product, or weighted product of likelihood measures determined for the corresponding measured variable; or by: subdividing the likelihood values ​​into operational likelihood values ​​including likelihood values ​​determined based on diagnostic criteria and application-specific likelihood values ​​including likelihood values ​​determined based on threshold criteria; and determining the total likelihood of the corresponding measured variable by: each of the likelihood measures and application-specific likelihood values ​​determined for the corresponding measured variable and a sum, weighted sum, product, or weighted product of the minimum likelihood given by the smallest of the operational likelihood values ​​determined for the corresponding measured variable.

[0024] In another embodiment, the method includes at least one of the following steps: a) displaying likelihood results on a dashboard-like display including icons visualizing the total likelihood index and / or a given number of icons, each icon visualizing one of the total likelihood values; wherein a portion of the icon area of ​​each icon corresponding to the size of the visualized total likelihood index or visualized total likelihood value is filled, and the icon or the filled portion thereof is displayed in a color selected according to the size; b) providing the likelihood results in the form of an email or message, the email or message being dispatched to at least one of predetermined recipients and / or predetermined devices, computers, mobile devices, mobile phones, tablets, and repair tools; and c) providing the likelihood results to a higher-level unit configured to regulate and / or control a process performed at the application, and configured to perform an action to stop or modify at least one process step of the process performed at the application and / or at least one other predetermined action when the likelihood results satisfy conditions specified for a corresponding action.

[0025] In another embodiment, the specified variable includes at least one of the following: at least one process parameter measured by one of the measuring devices of the measurement system; at least one process parameter measured by a measuring instrument installed at a particular application; and at least one diagnostic parameter determined by or for one of the measuring devices.

[0026] In another embodiment, for at least one of the measuring devices, at least one of the diagnostic information indicating the condition of the corresponding measuring device includes at least one of the following: the lifespan of the measuring device, the operating time of the measuring device, the maintenance time when repairing the measuring device, the verification time when verifying the measurement accuracy of the measuring device, the verification result obtained by verifying the measurement accuracy of the measuring device, the calibration time when calibrating the measuring device, the calibration result obtained by calibrating the measuring device, at least one diagnostic parameter determined by the measuring device, a status index determined based on the self-diagnosis performed by the measuring device, and an exposure index corresponding to the measuring device being exposed to adverse measurement conditions.

[0027] In another embodiment, the method is a method in which:

[0028] a) For at least one measured variable, the likelihood criterion determined for the corresponding measured variable includes at least one of the following: a threshold criterion for determining the likelihood value based on whether the current measurement of the corresponding measured variable falls within an application-specific threshold range that the measured variable will not exceed; a threshold criterion for determining the likelihood value indicating the likelihood of at least one current measurement of the corresponding measured variable based on the application-specific probability of the measurement of the measured variable falling within the application-specific threshold range; and a criterion for determining the likelihood value indicating the likelihood of at least one current measurement of the corresponding measured variable based on at least one current variable value of at least one or two other variables, wherein each other variable is given by one of the other measured variables or by a parameter included in the specified variable, and / or wherein:

[0029] (b) For at least one of the measured variables, the likelihood measure includes at least one of the following: a likelihood measure determined based on a method for detecting outliers included in the measurements of the corresponding measured variable; a likelihood measure determined based on a measurement of the corresponding measured variable taken before the measurement time of at least one current measurement of the corresponding measured variable; a likelihood measure determined based on a combination of the probability of occurrence of measurements of the magnitude of at least one current measurement of the corresponding measured variable at a particular application and the probability of occurrence of these measurements according to an empirical distribution determined based on measurements of the corresponding measured variable taken during a finite time interval prior to the measurement time of the current measurement; and a likelihood measure determined based on the deviation between at least one current measurement of the corresponding measured variable and a corresponding predicted value predicted based on measurements of the corresponding measured variable taken before the measurement time of the current measurement, wherein the predicted value is determined based on an autoregressive integral moving average model fitted to a time series of previously determined measurements, or based on a model of measurements of the corresponding measured variable already determined by a data-based machine learning method, or determined by another time series forecasting method.

[0030] In some embodiments, for each likelihood criterion, a corresponding likelihood value is determined based on a lookup table or likelihood function associated with the corresponding likelihood criterion, which assigns the likelihood value to the current measurement value of the corresponding measured variable based on or according to at least one attribute, which is given or may be determined based on at least one of the item values ​​of the diagnostic information items and / or at least one of the variable values ​​of at least one of the specified variables included in the data.

[0031] In another embodiment, for at least one of the measured variables, the determined likelihood measure includes a likelihood measure determined by the following steps: sorting the measured values ​​of the corresponding measured variable along the magnitude of a line according to the measured values ​​of the corresponding measured variable measured during a previous time interval; subdividing the line into four quartiles, each quartile comprising one-quarter of the measured value; and determining a likelihood measure based on a likelihood function that assigns the likelihood measure of the current measured value of the corresponding measured variable to all current measured values ​​appearing at positions within a predetermined first range of the line, according to the quartile in which the current measured value of the corresponding measured variable appears, and / or assigning the likelihood measure of the current measured value of the corresponding measured variable to all current measured values ​​appearing at positions within a predetermined second range of the line, according to the probability of occurrence of a measured value of magnitude of the current measured value; including that the probability of occurrence of a measured value of magnitude in the second range is determined based on training data included in the data or based on the measured values ​​measured during a previous time interval.

[0032] In some embodiments, the method is a method in which: for at least one of the measured variables, the determined likelihood measure includes a likelihood measure determined based on a likelihood function, which assigns the likelihood measure to at least one current measurement of the corresponding measured variable based on or as a sum, weighted sum, product, or weighted product of a first likelihood determined based on a first likelihood function and a second likelihood determined based on a second likelihood function; wherein: the first likelihood function assigns a first likelihood to the current measurement based on the probability of the current measurement being of a magnitude in a particular application; wherein the first likelihood function is determined based on an estimate of the probability of a measurement of a given magnitude occurring in a particular application, or based on the frequency of occurrence of measurement of different magnitudes, said different magnitudes being based on training... The second likelihood function is determined based on measurements already taken during a training interval that covers a sufficiently long time span to cover all operating modes and / or each process performed in a particular application; wherein the second likelihood function assigns the second likelihood to at least one current measurement based on the probability of occurrence of at least one current measurement according to an empirical distribution determined based on measurements taken during a finite time interval prior to the measurement time of the current measurement; and the second likelihood function is determined as a probability function based either on the frequency of occurrence of measurements of different sizes or based on kernel density estimation, the different sizes of measurements being determined based on measurements already taken during the finite time interval, the probability function representing the probability of occurrence of a measurement of the corresponding measured variable according to its size.

[0033] In some embodiments, for at least one of the measured variables, the method further includes the steps of: identifying at least one elimination criterion for the likelihood of a measurement of the corresponding measured variable; and performing a determination of the total likelihood of the measurement of the corresponding measured variable such that when the corresponding elimination criterion is met, the total likelihood is set to zero or reduced to a degree that takes into account the effect of meeting the corresponding elimination criterion on the likelihood of the measurement of the corresponding measured variable; wherein the elimination criterion identified for the corresponding measured variable includes at least one of the following: an elimination criterion associated with one of the diagnostic information identified for the measuring device measuring the measured variable; an elimination criterion requiring a status indicator determined for the measuring device measuring the measured variable to indicate that the measuring device is defective; an elimination criterion requiring that the measured value of the measured variable and / or the variable value of at least one other variable included in the specified variables exceed a maximum permissible range or exceed a given threshold or fall below a given threshold; and an elimination criterion associated with at least one parameter included in the specified variables, the at least one parameter being measured or determined by one of the measuring devices or by a measuring instrument installed at a particular application.

[0034] In some embodiments, the method is a method in which the measurement system is an analysis system, wherein the measuring device measures a measured variable of a medium flowing through a flow chamber, wherein the specified variable includes the flow rate of the medium flowing through the flow chamber, and wherein: the likelihood criteria include at least one criterion for determining the likelihood value of a measurement of one of the measured variables based on the measured flow rate, and / or when the flow rate through the flow chamber drops below the absolute minimum flow rate required to measure the corresponding measured variable, the total likelihood of at least one or each of the measurements of at least one measured variable is set to zero.

[0035] In some embodiments, the method further includes the following method steps: performing an iterative process to optimize the method by performing a machine learning method based on labeled training data obtained by an expert operator who evaluates and classifies the previously determined total likelihood, the machine learning method being configured to optimize the determination of the total likelihood and / or optimize at least one of the following: an application-specific threshold, an application-specific threshold range, a lookup table and function used to determine the likelihood value and the likelihood metric used to determine the total likelihood.

[0036] In some embodiments, the method is a computer-implemented method, wherein: each likelihood result is determined and provided by a computing device configured to determine and provide the likelihood result based on recorded data and based on a computer program implemented on the computing device, and the computing device performs the determination; and the computing device is included in a measurement system, in a transmitter connected to or connected to a measurement device, in a device located near the measurement system and connected to or communicating with the measurement device, an edge device or a higher-level unit, or embodied in the cloud.

[0037] This objective is also achieved by a computer program including instructions and a computer program product including the computer program and at least one computer-readable medium, wherein when the computer executes the program, the instructions cause the computer to implement the methods described herein based on data provided to the computer, wherein at least the computer program is stored on a computer-readable medium. Attached Figure Description

[0038] The invention and its further advantages are explained in more detail below based on the examples shown in the accompanying drawings, wherein:

[0039] Figure 1 A measurement system installed on a container in a specific application is shown.

[0040] Figure 2 An analytical measurement system installed on a pipeline in a specific application is shown.

[0041] Figure 3The analytical measurement system installed at the measurement location is shown.

[0042] Figure 4 The method steps of the method disclosed herein are shown.

[0043] Figure 5 The discrete likelihood function is shown.

[0044] Figure 6 The continuous likelihood function is shown.

[0045] Figure 7 The time series of the measured values ​​of the variable being measured is shown.

[0046] Figure 8 The diagram shows the line crossed by measurements classified according to their magnitude and a likelihood function, and...

[0047] Figure 9 The first likelihood function and the second likelihood function are shown. Detailed Implementation

[0048] This invention relates to a method for determining at least one measured variable m measured by a measurement system installed in a specific application. i The measured value mv i An application-specific likelihood method.

[0049] This method can be applied to any measurement system known in the art, which is installed in a particular application and configured to measure at least one measured variable m. i In this regard, the measurement system includes at least one measuring device Mi, and each measuring device Mi measures at least one measured variable m. i . Relative to at least one measured variable m known in the art and configured to measure the variable of interest in a particular application. i The measuring device Mi is capable of using process parameters, for example, those related to the properties of the process performed at the measuring location and / or the medium generated, processed, and / or monitored at the measuring location. The measured variable m i Examples include liquid level, pressure, temperature, density, conductivity, flow rate, pH value, turbidity, spectral absorption of the medium, concentration of analytes included in the medium, and / or at least one other analyte m. i As an example, oxygen content, ammonium content, and / or phosphorus content can be measured. Regardless of the measured variable m measured by the measuring device Mi... i How can each measuring device Mi be represented, or simply as a device that measures only a single variable m? i The sensor or probe, or the means of measuring at least two measured variables m i and / or at least one measured variable mi A more complex device with at least one parameter (e.g., diagnostic parameter and / or process parameter). Examples of applications include industrial applications such as manufacturing plants, chemical plants, water purification plants, and laboratory applications. Further examples include measurement systems that perform measurements in natural environments, and measurement systems used in medical diagnostics, such as systems performing in-situ, in vitro, or in vivo measurements. Regardless of the application, the measured value mv provided by the measurement system... i For example, for monitoring, regulating and / or controlling processes performed at the application site, such as at or by a plant or facility, to monitor, regulate and / or control at least one attribute or quality of a medium (e.g., a medium given by a semi-finished, intermediate or final product processed and / or produced at the application site), and / or for monitoring, regulating and / or controlling the efficiency of processes performed at the application site.

[0050] Figure 1 An example of a measurement system 100 installed on container 1 in a specific application is shown, such as based on the measurement value mv provided by the measurement system 100. i Applications include monitoring, regulating, and / or controlling processes (e.g., production processes performed in container 1). The exemplary measuring device Mi shown includes a level measuring device M1 for measuring the level L of the medium 3 contained in container 1, a conductivity sensor M2 for measuring the conductivity ρ of the medium 3, and two flow meters M3 and M4, each measuring the flow rate F1 and F2 of the additive flowing into container 1.

[0051] Figure 2 Another example is shown, in which the measurement system 200 is an analytical system, such as a liquid analysis system, including the measured variable m of the measurement medium 5. i The measuring device Mi. To provide an example of a specific application, Figure 2 The measurement system shown is employed, for example, as a water quality measurement system installed at a measurement point along a water pipe or bypass supplying water, such as fresh water supplied to or drawn from a swimming pool. In the example shown, the measuring device Mi includes a pH sensor M5 for measuring the pH value of the medium 5 and a chlorine sensor M6 for measuring the concentration of free chlorine Cl2 contained in the medium 5.

[0052] Figure 3Another example is shown, in which the measurement system 300 is embodied, for example, as a water quality measurement system for monitoring the quality of drinking water in a specific application, such as at a measurement point in a drinking water supply network where the measurement system 300 is installed. In this case, the measurement device Mi includes, for example, a turbidity sensor M7 for measuring turbidity TB, a pH sensor M8 for measuring pH value pH, a redox potential sensor M9 for measuring oxidation-reduction potential ORP, a conductivity sensor M10 for measuring conductivity ρ, a spectral absorption sensor M11 for measuring spectral absorption SAK, and / or a temperature sensor M12 for measuring the temperature T of the medium 5.

[0053] Various methods of installing measurement systems 100, 200, 300 and / or individual measuring devices Mi, known in the art, can be applied. Figure 1 In this system, each measuring device M1, M2, M3, and M4 is installed in a different location, such as on inlet pipes 7 and 9 connected to the inlet of container 1, above container 1, or on the container wall of container 1. Figure 2 In this design, measuring devices M5 and M6 are embodied as immersion probes mounted on a flow chamber 11, such that they are immersed in a medium 5 flowing through the flow chamber 11. This medium is supplied to the flow chamber 11 via an inlet pipe 13 and exits the flow chamber 11 via an outlet pipe 15. Figure 3 In this process, each measuring device M7, M8, M9, M10, M11, and M12 is individually installed at different locations along the vessel 17, which is, for example, an open channel, pipe, or container containing the medium 5.

[0054] like Figure 4 As shown, the measured value mv is determined. i Application-specific likelihood methods include method steps a): for each measured variable m i Identify and indicate the measurement of the corresponding measured variable m i Diagnostic information Iij of the condition of the measuring device Mi.

[0055] For example, the available, measurable, or determined diagnostic information Iij for the corresponding measuring device Mi includes: the lifespan of the measuring device Mi, the operating time of the measuring device Mi, the maintenance time when repairing the measuring device Mi, the verification time when verifying the measurement accuracy of the measuring device Mi (e.g., based on reference measurements), the verification results obtained by verifying the measurement accuracy of the measuring device Mi, the calibration time when calibrating the measuring device Mi, and / or the calibration results obtained by calibrating the measuring device Mi.

[0056] The diagnostic information Iij includes, for example, the diagnostic parameters determined and provided by the measuring device Mi, relative to the measuring device Mi which is configured to determine at least one diagnostic parameter indicating the condition of the corresponding measuring device Mi. As an example, at least one of the pH sensors M5 and M8 can be configured to determine and provide diagnostic parameters given by the impedance of its ion-selective glass membrane, which indicates the condition of the membrane.

[0057] Relative to the measuring device Mi configured to perform self-diagnosis, the diagnostic information Iij includes, for example, a status indicator representing the condition of the measuring device Mi, which is determined and provided by the measuring device Mi based on the self-diagnosis performed by the measuring device Mi.

[0058] As another example, the diagnostic information Iij may include an exposure index corresponding to the measurement device Mi being exposed to adverse measurement conditions. The exposure index is, for example, based on the exposure time and / or based on the measured variable m. i The exposure time is determined by the exposure time beyond the measurement range of the measuring device Mi, during which the measuring device Mi is exposed to adverse conditions, such as being exposed to temperatures beyond the temperature range specified for the measuring device Mi and / or being exposed to pressures beyond the pressure range specified for the measuring device Mi.

[0059] The method further includes method step b): specifying at least one variable v that is measured at the application or determined for the application. n The quantity of , such that the specified variable v n The measured variable m is measured by measurement systems 100, 200, and 300. i Each given or included in the table is a variable m being measured. i In the first case, specify the variable v. n The quantity equals the measured variable m i The quantity. In the latter case, specify the variable v. n Including each measured variable m i and at least one parameter measured at the application or determined for the application. Depending on the application, this may include the specified variable v. n The parameters include, for example, at least one process parameter measured by one of the measuring devices Mi of the measuring systems 100, 200, and 300 and / or at least one process parameter measured by another measuring instrument Sm, other than the measuring systems 100, 200, and 300, installed at a particular application. Examples of process parameters are shown in... Figure 2 As shown, pH sensor M5 and chlorine sensor M6 are each configured to measure process parameters, such as temperature Tph measured by temperature sensor 45 included in pH sensor M5 and temperature Tcl2 measured by temperature sensor 45 included in chlorine sensor M6.

[0060] As an alternative, at least one or each of the process parameters measured by (multiple) measuring devices Mi is included, for example, in the specified variable v. n In this context, m is the measured variable, which is measured by measurement systems 100, 200, and 300. i one of the.

[0061] In many applications, in addition to measurement systems 100, 200, and 300, at least one measuring instrument Sm is installed for measuring process parameters, such as the properties of medium 5 and / or process parameters related to the process performed at the application. Figure 2 An exemplary measuring instrument Sm is shown, which includes a flow meter S1 for measuring the flow rate F of the medium 5 flowing through the flow chamber 11, a pressure sensor S2 for measuring the pressure p in the supply pipe 13, and a temperature sensor S3 for measuring the ambient temperature Ta.

[0062] exist Figure 2 In the example shown, the variable v is specified. n The measured variable m is given by the pH value (pH) and chlorine content (Cl) of medium 5. i It may include at least one or all of the available parameters, such as temperature Tph, temperature Tcl2, flow rate F, pressure p inside supply pipe 13 and / or ambient temperature Ta.

[0063] Alternatively, or as an alternative to process parameters measured by measuring device Mi and / or measuring instrument Sm, it may include the specified variable v. n The parameters may include at least one diagnostic parameter measured or determined by one of the measuring devices Mi.

[0064] The method also includes method step c): for each measured variable m i To determine the likelihood criterion Cik, which is used to determine the indicator for the corresponding measured variable m. i The measured value mv i The corresponding likelihood value PCik is the likelihood of the likelihood.

[0065] For each measured variable m i The likelihood criterion Cik includes at least one diagnostic criterion C(Iij), which is used to measure the corresponding measured variable m based on the indication. i The diagnostic information Iij of the measuring device Mi is used to determine the value of at least one item in the corresponding measured variable m, thereby indicating the condition of the device. i The measured value mv i The likelihood value P(Iij) of the likelihood.

[0066] For each measured variable m iThe likelihood criterion Cik also includes at least one threshold criterion Cj(m) i ), which is used based on the corresponding measured variable mv i The measured value mv i Relative to the corresponding measured variable mv i The measured value mv i The size of at least one application-specific threshold range is used to determine the indication of the corresponding measured variable m. i The measured value mv i The likelihood value Pj(m) of the likelihood i As an example, the threshold criterion Cj(m) i For example, it includes at least one standard C1(m) i ), used to determine the corresponding measured variable m based on the limit that should not be exceeded in a specific application. i The likelihood value P1(m) is determined based on the application-specific threshold range. i The likelihood value P1(m) i Indicates the measured variable m i At least one of the current measurements mv i The likelihood of (tr). As an example, the pH of drinking water may always be required to be greater than or equal to 6 and less than or equal to 8. Alternatively, or as an alternative, the threshold criterion Cj(m i For example, including standard C2(m) i ), used to determine the measured variable m based on the corresponding threshold range within the application-specific range. i The measured value mv i The likelihood value P2(m) is determined based on the probability of occurrence of the application. i The likelihood value P2(m) i Indicates the measured variable m i At least one of the current measurements mv i Likelihood of (tr). As an example, in a particular application, the probability of measured pH values ​​of medium 5 occurring in the central range of 6 to 8 may be significantly higher than their probability of occurring in the lateral ranges of 4 to 6 and 8 to 10, and the application-specific probability of measured pH values ​​below 4 and above 10 may be negligible or zero.

[0067] The method further includes: a method step d) of recording data D, wherein the data D includes at least one item value for each item of diagnostic information Iij and each specified variable v n The variable value mv n The time series and their determination or measurement time t; and the method steps of performing method step e) at least once or repeatedly based on the recorded data D, wherein method step e) includes for each measured variable m iDetermine the corresponding measured variable m i The current measured value mv i Total likelihood of (tr) Ptot(m) i Method steps f), and method steps g) to provide the corresponding likelihood result PR.

[0068] The method disclosed herein is preferably implemented as a computer-based method. In this case, the method steps, particularly each determination of the likelihood result PR, are executed by the computing device 19 by means of a computer program SW. Therefore, the invention is also implemented in the form of a computer program SW comprising instructions that, when executed by a computer, cause the computer to implement the method disclosed herein. Furthermore, the invention also includes a computer program product comprising the aforementioned computer program SW and at least one computer-readable medium, wherein at least the computer program SW is stored on the computer-readable medium.

[0069] When this method is executed as a computer-implemented method, data D is, for example, transferred to and at least temporarily stored in memory 21 associated with computing device 19. Computing device 19 is embodied, for example, as a unit including hardware (e.g., a computer or computing system), included in or near measurement system 200, such as in an edge device or upper-level unit. As an alternative, cloud computing can be applied. Cloud computing refers to a method in which IT infrastructure, such as hardware, computing power, memory, network capacity, and / or software, is provided via a network (e.g., via the Internet). In this case, computing device 19 is embodied in the cloud. In either case, the measured value mv i The memory 21 is provided directly or indirectly to the computing device 19 or associated with the computing device 19. In this regard, hardwired or wireless connectivity and / or communication protocols known in the art, such as LAN, W-LAN, Fieldbus, Profibus, HART, Bluetooth, NFC, TCP / IP, etc., can be applied. Alternatively, an application programming interface (REST-API) conforming to an architectural style known in the art as Representational State Transition (REST) ​​can be employed.

[0070] When computing device 19 is embodied in the cloud, for example, Figure 1 As shown, the software including the algorithm, the recorded data D, and administration are hosted, for example, in a web-based database. This provides the advantage that the measurement system 100, the upper-level unit 23, and / or the edge device 25 that can connect to and / or communicate with the measurement device Mi do not need to access the recorded data D. Figure 1In this configuration, each measuring device Mi is connected directly to and / or communicates with the computing device 19, as indicated by arrow A, via a higher-level unit 23 (e.g., a programmable logic controller), as indicated by arrows B1 and B2, and / or via an edge device 25 located near the measuring device Mi, as indicated by arrows C1 and C2. As an example, at least one or each of the measuring device Mi, the edge device 25, and / or the higher-level unit 23 may be connected directly or indirectly to the computing device 19 via the Internet, for example via a communication network (e.g., TCP / IP).

[0071] exist Figure 2 In this embodiment, computing device 19 is included in device 27 located near measurement system 200. Device 27 is provided, for example, by an edge device or upper-level unit, or by another device directly or indirectly connected to and / or communicating with each measurement device Mi. In this embodiment, software including algorithms and recorded data D is hosted in device 27, and administration and total likelihood Ptot(m) are also managed. i The computing device 19 is hosted on device 27 or in a web-based database. This embodiment offers the advantage of being implemented at a relatively low cost because it requires less data traffic than cloud-based implementations and minimizes data usage in the cloud. Implementing the computing device 19 near the measurement system 200 is particularly suitable for applications where connectivity interruptions to the web-based database may occur.

[0072] Regardless of whether the computing device 19 is physically present in the cloud or near the measurement systems 100 and 200, the measurement device Mi is configured, for example, to provide the measured value mv. i and their measurement time t, for example, Figure 1 As shown. Figure 2 In the alternative shown, the measuring device Mi is connected, for example, to the transmitter 29, which is configured to determine and provide the measured value mvi and its measurement time t based on the measurement signal provided by the individual measuring device Mi. In this case, the transmitter 29 is, for example, as shown in the diagram. Figure 2 The device shown is connected to and / or communicates with the computing device 19 directly or via a higher-level unit and / or via an edge device located near the transmitter 29.

[0073] exist Figure 3 In this system, computing device 19 and memory 21 are included in component 31 of the measurement system 300. In this case, component 31 may, for example, be connected to a device that provides the measurement value mv. i And their measurement time t, and the measuring device Mi or communicating with said measuring device. Alternatively, such as Figure 3As shown, component 31 is included, for example, in or connected to transmitter 33, which is connected to measuring device Mi and configured to determine and provide the measured value mv based on the measurement signal provided by the individual measuring device Mi. i And their measurement time t. When the computing device 19 is implemented in the measurement system 300, the software, data D, and administration are also implemented in the measurement system 300. This requires the measurement system 300 to be equipped with advanced data analysis capabilities and sufficient computing power. Therefore, implementing the computing device 19 in the measurement system 300 is particularly suitable for highly critical applications or remote applications where internet connectivity is not available.

[0074] Regardless of where the computing device 19 is implemented, the item values ​​of each item of the diagnostic information Iij can be provided to the computing device 19 and / or the memory 21, for example, by one or more sources. As an example, at least one item value or a time series of item values ​​of at least one item of the diagnostic information Iij may be determined and provided, for example, by one of the measuring devices Mi, transmitters 29 and 33, edge device 25, or a higher-level unit 23 directly or indirectly connected to or communicating with the computing device 19. Alternatively, at least one item value or a time series of item values ​​of at least one item of the diagnostic information Iij may be provided to the computing device 19, for example, via an interface 35, which may include, for example, a user interface enabling an operator to provide item values ​​as indicated by arrow OP and / or an interface connecting to or communicating with an external data source 37 (e.g., a database) that provides the item values.

[0075] The specified variable v is given by the parameters determined by the measuring device Mi. n The measured value mv n For example, the measured variable m is compared with the measured variable m by the corresponding measuring device Mi. i The measured value mvi is provided to the computing device 19 in the same manner. The specified variable v is given by the parameter measured by the measuring instrument Sm. n The measured value mv n For example, it is provided to the computing device 19 via a corresponding measuring instrument Sm that is directly connected to and / or communicates with the computing device, via a higher-level unit 23, and / or via an edge device 25 located near the corresponding measuring instrument Sm.

[0076] As mentioned above, each determination of the likelihood result PR involves considering each measured variable m. i Determine the total likelihood Ptot(m) i Step f). These total likelihoods Ptot(m) i Each of these is based on the corresponding measured variable m. iThe likelihood value Pcik is determined by defining the likelihood criterion Cik, and is determined by performing method step f1), which takes into account the values ​​measured prior to the current time interval and determined by the corresponding measured variable m. i The measured value mv i Give or include the corresponding measured variable m i The measured value mv i At least one or all of the specified variables v n The measured value mv n Determine the corresponding measured variable m that has been measured during the finite current time interval. i One or more current measurements mv i At least one likelihood measure Rin of the reliability of (tr). Alternatively, each likelihood measure Rin is determined, for example, as a value, preferably a normalized value greater than or equal to 0 and less than or equal to 1, or in the form of a corresponding percentage greater than or equal to 0% and less than or equal to 100%.

[0077] For each measured variable m i The likelihood measure Rin preferably includes a first likelihood measure Ri1 and / or a second likelihood measure Ri2. This is achieved by analyzing the corresponding measured variable m. i The measured value mvi is analyzed and each first likelihood measure Ri1 is determined by identifying a first likelihood measure Ri1, which corresponds to the corresponding measured variable m that has been measured during the finite current time interval. i The current measured value mv i (tr) or current measurement value mv i The distribution of (tr) and the corresponding measured variable m that has been measured before the current time interval. i The measured value mv i The degree to which the presented distribution or one of the distributions conforms to the given distribution. As an example, the first likelihood measure Ri1 is, for instance, determined by the current measurement mv. i The statistical probability of (tr) is given to form the value mv that has been measured before the current time interval. i The presented distribution or a sample of one of the distributions.

[0078] Each second likelihood measure Ri2 is determined by performing a multivariate analysis of the predefined analysis variables. These analysis variables are either derived from all specified variables v. n Give, or through including the corresponding measured variable m i The specified variable v nThe choice is given. Based on multivariate analysis, the second likelihood measure Ri2 is then determined as the likelihood measure Ri2, which corresponds to the corresponding measured variable m that has been measured during the finite current time interval. i (Multiple) current measurements mv i (tr) The degree of agreement between the analysis results and those determined by multivariate analysis based on the values ​​of the analysis variables that have been measured prior to the current time interval.

[0079] Regarding multivariate analysis, known mathematical methods for analyzing datasets based on multivariate statistical principles, such as time series analysis of a given number of analytical variables, can be employed. These methods support the determination of relationships between datasets, such as structural relationships, correlations, and / or interdependencies between analytical variables, and / or structural relationships, correlations, and / or interdependencies between patterns and / or distributions of analytical variable values. In the context of the methods disclosed herein, for example, multivariate data analysis is used to determine application-specific relationships between analyzed variables, and based on the corresponding measured variable m... i The current measured value mv i (tr) determines each second likelihood measure Ri2 based on the degree of conformity with the cross-relationship, which is determined based on the variable values ​​of the analytical variables included in the data D, and is based on the current measurement value mv. i The measurement time tr of (tr) is determined beforehand.

[0080] The analytical variables include the corresponding measured variable m. i And preferably also includes all other measured variables m measured by measurement systems 100, 200, and 300. i≠i In most applications, at least one of the parameters, for example, is determined by... Figure 2 The pH sensor M5 and / or chlorine sensor M6 shown in the diagram measure the ambient temperature Ta, pressure p, flow rate F, and / or temperature Tcl and TpH, respectively, for the measured variable m. i At least one and / or the corresponding measured variable m i The measured value mv i It has an impact. Therefore, by additionally including the specified variable v n The inclusion of at least one of the parameters in the analysis variable further enhances the ability of multivariate analysis to distinguish the measured variable m determined using correctly operated measuring equipment Mi during the correct execution of the process performed at a specific application. i The current measured value mv i (tr) and the current measurement value mv affected by at least one problem associated with the measuring device Mi and the application. i (tr).

[0081] Total likelihood Ptot(m) i The determination of ) also includes method step f2), which is for each measured variable m i Based on the corresponding measured variable m i Given the established likelihood measure Rin and likelihood value PCik, the corresponding measured variable m is determined. i The current measured value mv i Total likelihood of (tr) Ptot(m) i ).

[0082] For each likelihood criterion Cik, the corresponding likelihood value PCik is determined, for example, based on a lookup table or likelihood function f(Cik) associated with the corresponding likelihood criterion Cik, which assigns the likelihood value PCik to the corresponding measured variable m based on or according to a function of at least one attribute. i The current measured value mv i (tr), this attribute is determined by at least one of the item values ​​of the diagnostic information Iij and / or the specified variable v included in the data D. n The value of at least one of the variables mv n The likelihood value PCik is given or determined based on it. Like the likelihood measure Rin, the likelihood value PCik is determined, for example, as a normalized likelihood value greater than or equal to 0 and less than or equal to 1, or as a corresponding percentage greater than or equal to 0% and less than or equal to 100%.

[0083] As an example, for at least one or each diagnostic criterion C(Iij), the corresponding likelihood value PIij is determined, for example, based on a lookup table or a likelihood function f(Iij), which represents the measurement value mv according to at least one attribute given or determined by the item values ​​of the corresponding items based on the diagnostic information Iij. iThe likelihood of the measurement equipment Mi can be assessed using a lookup table that includes predetermined individual likelihood values ​​for different lifespan groups, such as a high value applicable when the equipment is new, an intermediate value applicable when the equipment is in its middle lifespan, and a low value applicable when the equipment reaches the end of its service life. Alternatively, a likelihood function can be used, which represents the reliability of the measurement values ​​determined by measuring equipment of the corresponding type of equipment Mi based on their lifespan. In this case, the likelihood value is given by the reliability function for the reliability value provided by the measuring equipment for the lifespan of the corresponding equipment Mi. For each item of the diagnostic information Iij given by the calibration time, the corresponding likelihood value is determined, for example, based on a continuous likelihood function of the time elapsed since the last calibration of the equipment Mi. This likelihood function is defined, for example, such that the likelihood value provided by the function decreases as the time elapsed since the last calibration of the equipment Mi increases.

[0084] Regarding the corresponding measured variable m i The threshold standard C1(m) is defined as the application-specific threshold range that should not be exceeded in a particular application. i The corresponding likelihood value P1(m) i For example, it is determined based on a lookup table, when the corresponding measured variable m i The current measured value mv i (tr) When the value exceeds an application-specific threshold range, the lookup table provides a 0% likelihood value, and when the current measurement value mv i When (tr) appears within this threshold range, the lookup table provides a 100% likelihood value.

[0085] Regarding the corresponding measured variable m i The measured value mv i The threshold criterion C2(m) is defined by the application-specific probability of occurrence within different threshold ranges. i The corresponding likelihood value P2(m) i The value is determined, for example, based on a lookup table or a likelihood function. Figure 5 An example of a lookup table is shown, which assigns a likelihood value f1(pH) to the measured pH value pH based on one of the threshold ranges listed in the left column, where the likelihood value f1(pH) appears within that threshold range. Figure 6 An example of the likelihood function f2 is shown, which represents the likelihood value f2(pH) based on the magnitude of the measured pH value.

[0086] For each measured variable m i Total likelihood Ptot(m)i ) is based on the corresponding measured variable m i The likelihood measure Rin and the likelihood value PCik have already been determined. According to the first embodiment, each total likelihood Ptot(m) is determined... i ) is based on the corresponding measured variable m i The total likelihood is determined by all likelihood measures Rin and all likelihood values ​​PCik. As an example, the total likelihood Ptot(m) is... i At least one or each of them is determined based on or as a sum or weighted sum of the likelihood measure Ril and the likelihood value PCik; for example, as:

[0087]

[0088] Or it may be determined as or based on the product or weighted product of the likelihood metric Rin and the likelihood value PCik, for example as

[0089]

[0090] When a weighted summation or weighted product is used, each of the likelihood measure Rin and the likelihood value PCik is multiplied by a weighting factor w. i,k ;w i,n This demonstrates that the corresponding value relates to the corresponding measured variable m. i The measured value mv i Total likelihood Ptot(m i The impact of ) . Using the correct weighting factor w i,n w i,k Provides the total likelihood Ptot(m) i The advantage of very high accuracy. In this regard, the weighting factor w for each likelihood measure Rin... i,n For example, the accuracy corresponding to the corresponding likelihood measure Rin is determined. This accuracy is determined, for example, by performing a mathematical method that determines the reliability of the corresponding likelihood measure Rin based on the method used to determine Rin and the statistical properties of the data D used to determine Rin. Although the weighting factor w of the likelihood measure Rin... i,n The determination of the likelihood value PCik can be performed entirely by data-driven methods, but the weighting factor w i,k Determining the likelihood value PCik typically requires expertise regarding the measuring equipment Mi and the application. Therefore, determining the weighting factor w for the likelihood value PCik is crucial. i,k It could be a lengthy and time-consuming process, and these weighting factors w i,k Its accuracy is only as good as available expert knowledge. Alternatively, the same weighting factor w i,kThis can be used in each likelihood value PCik. In this case, the individual likelihood value PCik relative to the corresponding measured variable m is no longer considered. i The measured value mv i Total likelihood Ptot(m i The difference in the correlation between )

[0091] According to the second embodiment, at least one or each total likelihood Ptot(m) i ) is based on the corresponding measured variable m i Each determined likelihood measure Rin and the result of the calculation for the corresponding measured variable m i The minimum likelihood Pmin1 is determined by the smallest of the given likelihood values ​​PCik. In this case, the corresponding total likelihood Ptot(m) is... i :=Ptot(PCik;Rin) :=Ptot(Pmin1;Rin) For example, determined according to the minimum likelihood Pmin1 and for the corresponding measured variable m i Each likelihood measure Rin is determined. This constitutes a more conservative approach that increases safety by eliminating the risk that a potential critical value in one of the likelihood values ​​PCik will not lead to a correspondingly lower total likelihood Ptot(m). i ), because its impact is incorrectly underestimated. Just as in the first embodiment, the corresponding total likelihood Ptot(m) is... i For example, it is determined by the sum, weighted sum, product, or weighted product of the minimum likelihood Pmin1 and the likelihood measure Rin. Alternatively, even more conservative methods can be applied, such as by using the corresponding total likelihood Ptot(m i The product of Pmin1 and Rtot1 is determined, for example, by Ptot(m i ): = Pmin1 * Rtot1, where the term Rtot1 is based on, or as, or on the corresponding measured variable m. i The sum, weighted sum, product, or weighted product of all known likelihood measures Rin is determined.

[0092] According to the third embodiment, the likelihood value PCik can be subdivided into an operational likelihood value including a likelihood value P(Iij) determined based on diagnostic criteria C(Iij) and an operational likelihood value including a likelihood value Cj(m) determined based on threshold criteria Cj(m). i The determined likelihood value Pj(m) i The application-specific likelihood value. In this case, for the measured variable m i One of the determined at least one or each total likelihood Ptot(m) i() is, for example, based on, according to, or based on the corresponding measured variable m i For each determined likelihood measure Rin, for the corresponding measured variable m i Each determined application-specific likelihood value is derived from the probability of the corresponding measured variable m. i The minimum operational likelihood value is determined by the sum, weighted sum, product, or weighted product of the given minimum likelihood Pmin2. As an example, the corresponding total likelihood Ptot(m) is... i For example, it can be determined as or based on the product of minimum likelihood Pmin2 and the term Rtot2, for example, by Ptot(m i ): = Pmin2 * Rtot2, where the term Rtot2 is based on, or as, or on the corresponding measured variable m. i It is determined by the sum, weighted sum, product, or weighted product of all application-specific likelihood values ​​and all likelihood measures Rin that have already been determined.

[0093] In method step g), determine the parameters for each measured variable m. i The measured value mv i Total likelihood Ptot(m i Following this, the corresponding likelihood result PR is provided, which includes the total likelihood Ptot(m) i ) and / or based on all measured variables m i The measured value mv i Total likelihood Ptot(m i The total likelihood index (TPI) is determined by this.

[0094] In some embodiments, providing the likelihood result PR is performed by displaying the likelihood result PR on a display 39 that is accessible to the operator of the particular application. One approach is to utilize visualization tools such as dashboards to provide information in an intuitive and readable graphical form. As an example, it is possible to apply an indicator of the total likelihood Ptot(m i The size of the total likelihood index (TPI) and / or its graphical, chart, or geometric form can be used. As an additional option, the displayed information can be color-coded. In this regard, intuitive color schemes can be employed, such as using red to visualize low likelihood and green to visualize high likelihood. The advantage of this is that it supports rapid identification and intuitive understanding of all displayed information. Figure 2 An example is shown, in which for Figure 2 The likelihood result PR determined by the measurement system 200 shown in the figure is displayed on the display 39 in the form of a dashboard. The dashboard shows an icon 41 and several icons 43 that display the total likelihood index TPI. Each icon 43 shows the result for each measured variable m. iThe determined total likelihood Ptot(m) i Each icon 41, 43 has a given shape, such as a circle or a ring. Furthermore, a portion of the icon area for each icon 41, 43 is filled, and this portion corresponds to the total likelihood index TPI or total likelihood Ptot(m) visualized by the corresponding icon 41, 43. i The size of each icon 41, 43 or its filling portion is preferably based on the total likelihood index TPI or total likelihood Ptot(m) displayed by the corresponding icon 41, 43. i The size of the symbol is selected to display the color. For example, when the total likelihood index (TPI) is 100%, the entire ring icon 41 displaying the total likelihood index (TPI) can be filled, and the icon 41 or its filled portion can be colored green. When the total likelihood index (TPI) is only 50%, only half of the ring can be filled and the icon 41 or its filled portion can be displayed in red.

[0095] Alternatively, or as an alternative, the likelihood result PR may be provided, for example, in the form of an email or message automatically generated by the computing device 19 and dispatched to a predetermined recipient and / or a predetermined device, such as a computer or mobile device, such as a cellular phone, tablet, or repair tool.

[0096] Alternatively, or as an alternative, the likelihood result PR is provided to the higher-level unit, for example, Figure 1 The parent unit 23 shown is configured to regulate and / or control the process executed at the application in a manner that describes the likelihood result PR. As an example, the parent unit 23 can be configured to execute at least one predefined action when the likelihood result PR satisfies a condition specified for a corresponding action. The predefined action may include at least one process step that changes or stops the process executed at the application. As an example, when the measured variable m specified in the condition... i Or each measured variable m i Determined total likelihood index TPI and / or total likelihood Ptot(m) i When the value drops below the corresponding predetermined threshold, the parent unit 23 can stop the entire process. Furthermore, when the measured variable m specified in the conditions... i Or each measured variable m i Determined total likelihood index TPI and / or total likelihood Ptot(m) i When the value drops below the corresponding predetermined threshold, the execution of at least one process step can be modified.

[0097] The present invention provides the advantages mentioned above. Individual steps of the method can be implemented in different ways without departing from the scope of the invention. Several alternative embodiments are described in more detail below.

[0098] In some embodiments, the likelihood criterion Cik includes at least one likelihood criterion C3(m). i ), which is used based on at least one other variable v j And determine the corresponding measured variable m i The measured value mv i The corresponding likelihood value P3(m) of the likelihood. i ), where each of the other variables v j Other measured variables m j≠i One of or included in the specified variable v n One of the parameters is given. In this case, as described above, it is based on the corresponding standard C3(m) i Each other variable v specified in ) j At least one current measurement value mv j (tr) and determine the corresponding likelihood value P3(m) i ), and then adopt the corresponding likelihood value P3(m) in the same manner as the aforementioned likelihood value PCik. i To determine the corresponding measured variable m i Total likelihood Ptot(m i When determining the total likelihood Ptot(m) according to the third embodiment. i When the operation likelihood value is used, it additionally includes each likelihood value P3(m) i The likelihood value P3(m) i The instruction is based on the corresponding measured variable m. i At least one other variable v that has been determined j The corresponding measured variable m i The measured value mv i Likelihood.

[0099] As an example, the threshold criterion related to the temperature TpH measured by the pH sensor M5 is, for example, based on a temperature range specified for that pH sensor M5. Figure 2 The pH value is determined by the pH value measured by the pH sensor M5 shown. In this case, the corresponding likelihood value is determined, for example, based on a lookup table or likelihood function, which indicates high likelihood when the measured temperature Tph falls within a specified temperature range, and low likelihood when the measured temperature Tph exceeds that temperature range.

[0100] As another example, a threshold criterion could be based on the temperature difference between the temperature TpH measured by pH sensor M5 and the temperature Tcl measured by chlorine sensor M6, and for... Figure 2 The likelihood values ​​are determined by the pH value measured by pH sensor M5 and / or the chlorine concentration Cl2 measured by chlorine sensor M6. Each of these likelihood values ​​is determined, for example, based on a lookup table or likelihood function that provides the corresponding measured variable m when the temperature difference falls within a predetermined application threshold range for that temperature difference. i The measured value mv i The lookup table or likelihood function indicates a high likelihood value, while when the temperature difference exceeds the application-specific threshold range, it indicates a low likelihood.

[0101] Furthermore, for example, a threshold standard related to the flow rate F of the medium 5 flowing through the flow chamber 11 can be used for... Figure 2 Additional likelihood values ​​are determined by the pH value measured by pH sensor M5 and / or the chlorine concentration Cl2 measured by chlorine sensor M6, as shown in the diagram. Each of these likelihood values ​​is determined, for example, based on a lookup table or likelihood function that provides a high likelihood value when the flow rate F exceeds the application-specific flow rate, and a low likelihood value when the flow rate F falls below the application-specific flow rate. Here, the application-specific flow rate is, for example, the minimum flow rate required to ensure that measurements performed by measurement system 200 are always performed on fresh samples of medium 5 routed through flow cell 11, and thus indicates the current condition of the monitored medium 5, such as the current water quality in a swimming pool, based on measurements performed on samples routed through flow cell 11.

[0102] Alternatively, in some embodiments, the method includes targeting the measured variable m. i At least one of them, based on the measured variable m that has been measured before the current time interval. i The measured value mv i And determine the corresponding measured variable m that has been measured during the current time interval. i At least one current measurement value mv i At least one additional likelihood measure Rin for the reliability of (tr). As an example, these likelihood measures Rin include, for instance, a likelihood measure determined based on an outlier detection method, or a likelihood measure based on the corresponding measured variable m. i The measured value mv i A likelihood measure determined by at least one time series prediction and / or based on the corresponding measured variable m i The measured value mv i Likelihood measures are determined by describing the probability, distribution, or pattern of occurrence.

[0103] This is Figure 7 The diagram shows the measured variable m measured by a measuring device Mi. i One of the measured values ​​mv i The time series, and the device-specific threshold RMi given by the measurement range of the measuring device Mi, and the measured variable m in this application. i The application-specific threshold Rapp will not be exceeded. As shown in the figure, there is no measurement value mv. i The value mv exceeds the device-specific threshold RMi, and is measured only at time t1. i (t1) appears unreliable because it exceeds the application-specific threshold Rapp. Even the measured value mv at time t2... i (T2) appears within the application-specific threshold Rapp, but it may still be unreliable, for example, because it is identified as being based on a previously determined measurement mv. i The outlier detected by the outlier detection method is, and / or because it deviates significantly from the corresponding predicted value.

[0104] As an example, it is possible to use a measurement value mv based on measurements taken during the previous time interval. i The analysis evaluates (multiple) current measurements mv i The outlier detection method (tr) is used to determine the corresponding likelihood metric Ri3. In this case, the length of the previous time interval is preferably predetermined, such that the measured value mv during the previous time interval is... i This represents the expected measurement value mv in a specific application. i The statistical distribution. For example, such as... Figure 8 As shown, the analysis may include based on the measurement value mv measured during the previous time interval. i Along Figure 8 The double-headed arrows in the figure indicate the magnitude of the line relative to the measured value mv. i Sort the data. The line thus determined is then subdivided into four quartiles: Q1:=[q0;q1], Q2:=[q1;q2], Q3:=[q2;q3], and Q4:=[q3;q4], each comprising one-quarter of the measured value mv. i Therefore, the lower limit q0 of the interquartile range [q0; q1] of the first quartile Q1 corresponds to the minimum measured value mv. i The size of the fourth quartile Q4, the upper limit of the quartile range [q3; q4], q4 corresponds to the maximum measured value mv. i The size of the quartile, and the limit q2 separating the two middle quartiles Q2 and Q3, corresponds to the measured value mv measured during the previous time interval. i The median.

[0105] Based on this outlier detection method, the corresponding likelihood measure is, for example, based on at least one current measurement value mv. i (tr) is determined along the position q of the line based on the likelihood function f(q), which is based on the current measurement mv. i (tr) Assign the likelihood measure to (multiple) current measurements mv along the position q of the line. i (tr). As an example, the likelihood function is based on the current measurement value mv. i (tr) The quartiles Q1, Q2, Q3, and Q4 that appear in it, the current measured value mv i The likelihood measure (tr) is assigned to all current measurements mv that appear within a predetermined first range of the line. i (tr), and / or based on the current measurement value mv i The measured value of (tr) mv i The probability of occurrence is assigned to all current measurements mv that appear within a predetermined second range on the line. i (tr). Includes the measured value mv within the second range. i The probability of occurrence is based, for example, on the training data included in the data D or on the measurement value mv measured during the previous time interval. i And it is certain.

[0106] As an example, the likelihood function can, for instance, assign the same likelihood measure to all current measurements mvi(tr) that appear in the same quartiles Q1, Q2, Q3, Q4. In this case, the first range includes the quartile ranges [q0; q1], [q1; q2], [q2; q3], [q3; q4] of all four quartiles Q1, Q2, Q3, Q4. Alternatively, the likelihood function can, for instance, be based on the current measurement mv i The measured value of (tr) mv i The probability of occurrence is assigned to all current measurements mv that occur along the line. i (tr). In this case, the second range includes the quartile ranges [q0; q1], [q1; q2], [q2; q3], and [q3; q4] of all four quartiles Q1, Q2, Q3, and Q4.

[0107] As another example, a combination of the first and second ranges can be applied. Figure 8 The measured value mv i An example of the corresponding likelihood function Ri3(q) is shown above the line. This likelihood function Ri3(q) assigns a 100% likelihood measure to the current measurement mv that appears in the two middle quartiles Q2 and Q3.i (tr), and will correspond to the current measurement value mv i The measured value of (tr) mv i The likelihood measure of the probability of occurrence is assigned to all current measurements mv that occur in the first intermediate range R1:=[q3;q3+Δq]. i (tr), the position q in the first intermediate range R1:=[q3;q3+Δq] exceeds the upper limit of the third quartile Q3, q3 is less than the predetermined addend +Δq, and will correspond to the current measured value mv i The measured value of (tr) mv i The likelihood measure of the probability of occurrence is assigned to all current measurements mv that occur in the second intermediate range R2:=[q1-Δq;q1]. i (tr), the position q in the second intermediate range R2:=[q1-Δq;q1] drops to the lower limit q1 of the second quartile Q2, which is less than the predetermined subtrahend -Δq. Figure 8 The likelihood function Ri3(q) shown also assigns a 0% likelihood measure to all current measurements mv that exceed the upper limit q3+Δq of the first intermediate range R1. i (tr) and all current measurements mv occurring at position q, which is below the lower limit q1-Δq of the second intermediate range R2. i (tr). Therefore, in this example, the first range includes the quartile ranges [q1; q2] and [q2; q3] of the two middle quartiles Q2 and Q3, which exceeds the upper limit q3+Δq of the first middle range R1 and does not exceed the lower limit q1-Δq of the second middle range R2, and the second range includes the two middle ranges R1 and R2.

[0108] In the application, at least one of the measured variables m i The measured value mv i It is normally distributed, and the corresponding m can be determined. i The measured value mv i The mean and standard deviation. In this case, the likelihood measure Rin includes, for example, at least one likelihood measure Ri4, which is based on the corresponding measured variable m. i At least one current measurement value mv i The likelihood measure Ri4 is determined by the deviation between (tr) and the mean. These likelihood measures are determined, for example, based on a likelihood function that assigns the likelihood measure to (multiple) current measurements mvi(tr) according to their position in the normal distribution. Similar to the previous example, in this case, the 100% likelihood measure Ri4 is, for example, assigned to (multiple) current measurements mv that occur within one standard deviation of the mean.i (tr), and corresponds to the current measurement mv according to a normal distribution. i The measured value of (tr) mv i The likelihood measure Ri4, for example, is assigned to (multiple) current measurements mv that deviate from the mean by more than one standard deviation. i (tr).

[0109] As another example, the determination of the likelihood measure Rin includes, for example, determining the measured variable m. i At least one of them, based on the magnitude of(multiple) current measurements mvi(tr) and(multiple) current measurements mv i The measured value of (tr) mv i The application-specific probability of occurrence and their probability based on the measured value mv i The likelihood measure Ri5 is determined by the combination of the occurrence probabilities of a given empirical distribution, and this measure is mv. i It is (multiple) current measurement values ​​mv i The measurement (tr) is performed during a finite time interval preceding the measurement time tr. Here, the finite time interval is implemented, for example, in the form of a sliding window of a given width, which extends to (multiple) current measurement values ​​mv. i The measurement time tr is (tr). In this case, the corresponding likelihood metric Ri5 is determined, for example, based on a likelihood function, which is based on or as a first likelihood fa(mv) i (tr)) and second likelihood fb(mv i The sum, weighted sum, product, or weighted product of (tr) will assign the likelihood measure to (multiple) current measurements mv. i (tr), the first likelihood is determined based on the first likelihood function fa, and the second likelihood is determined based on the second likelihood function fb.

[0110] The first likelihood function fa is based on the current measurement value mv at a specific application location. i The measured value of (tr) mv i The probability of occurrence of , the first likelihood fa(mv) i (tr) is assigned to (multiple) current measurement values ​​mv i (tr). A quick and simple method is based on a measurement of mv at a given size for a specific application. i The first likelihood function fa is determined by estimating the probability of occurrence. Alternatively, the first likelihood function fa can be, for example, based on measurements mv of different magnitudes. i The different sizes of the measured values ​​mv are determined by the frequency of their occurrence. iIt is based on the measurement value mv that has been measured during the training interval. i The training time interval is determined to cover a sufficiently long time span to cover all operating modes and / or each process performed in a particular application.

[0111] The second likelihood function fb is based on the current measured value mv. i The probability of occurrence of (tr) according to the empirical distribution will be used to determine the second likelihood fb(mv) i (tr) is assigned to (multiple) current measurement values ​​mv i (tr). The second likelihood function fb is determined, for example, by or based on measurements mv of different magnitudes. i The frequency of occurrence, the measured values ​​mv of different sizes. i It is based on the measured value mv that has been measured during a finite time interval. i And it is certain.

[0112] Alternatively, the second likelihood function fb can be determined in a significantly more accurate manner by performing kernel density estimation (KDE). In this regard, kernel density estimation methods developed in statistics can be used to determine the probability distribution of a random variable based on a sample of a statistical population. In this case, the probability function, representing the measured value mv, is determined based on kernel density estimation (KDE). i The probability of its occurrence is determined based on the probability of its size, and the second likelihood function fb is determined as or based on this probability function.

[0113] Figure 9 Examples of the first likelihood function fa and the second likelihood function fb are shown. (Like...) Figure 6 Similar to the likelihood function f2(pH) shown, the first likelihood function fa represents the current measured value mv. i The likelihood of (tr) is based on the magnitude of their occurrence probability within a threshold range specific to the application, and the second likelihood function fb is based on their occurrence probability according to an empirical distribution. i (tr) is assigned to (multiple) current measurement values ​​mv i (tr). For example, Figure 9 The points da and db shown indicate the current measurement value mv. i (tr), for example, a pH value of 8, the first likelihood function fa can render only 50% of the first likelihood, while the second likelihood function fb renders 100% of the second likelihood. Therefore, by basing the first likelihood on fa(mv)... i (tr)) and second likelihood fb(mv i (tr)) These two factors determine the likelihood metric Ri5, which describes the current measurement value mvi The higher confidence level in (tr) is demonstrated by the degree of conformity with the empirical distribution.

[0114] As another example, in some embodiments, the determination of the likelihood measure Rin includes based on the corresponding measured variable m. i The current measured value mv i (tr) and based on (multiple) current measurements mv i (tr) represents the measurement value mv measured before the measurement time tr. i At least one likelihood measure Ri6 is determined by the deviation between the predicted values. This predicted value is determined, for example, based on a time series forecasting method. In this case, the corresponding likelihood measure Ri6 is determined, for example, based on a likelihood function that assigns the likelihood measure to (multiple) current measurements mv according to the magnitude of the deviation. i (tr). As an example, time series forecasting is performed, for instance, based on an autoregressive integrated moving average (ARIMA) model, which is fitted to a previously determined measurement mv. i The time series data is then used to predict future points in the time series. Alternatively, or as an alternative, a Kalman filter or machine learning method can be used to determine the corresponding measured variable m based on the recorded data D. i The measured value mv i The model. In this case, the predicted value is determined based on such a determined model.

[0115] As an optional feature, the method includes targeting the measured variable m i At least one of them, identify the corresponding measured variable m i The measured value mv i At least one elimination criterion (KO) is used for the likelihood. In this case, the corresponding measured variable m is determined. i The measured value mv i Total likelihood Ptot(m i ), such that when the corresponding elimination criterion KO is met, the total likelihood Ptot(m) is... i The value is either set to zero or reduced to take into account the corresponding elimination criterion (KO) for the measured variable m. i The measured value mv i The extent to which likelihood is affected.

[0116] In some embodiments, the elimination criterion KO includes, for example, the measurement of the corresponding measured variable m. i The measurement device Mi identifies at least one relevant elimination criterion in the diagnostic information Iij, and the measured variable m. iThe measured value mv i The relevant elimination criteria, and / or those included in the specified variable v n At least one variable v in j The variable value mv j The relevant elimination criteria (KO) are as follows. For example, when the requirement is met by measuring the measured variable m... i When the measuring device Mi provides status indicators to indicate the criteria for elimination due to defects in the measuring device Mi, the measured variable m... i One of the measured values ​​mv i Total likelihood Ptot(m i For example, it is determined to be zero. Furthermore, when the requirements are met, the measured variable m... i The measured value mv i When the elimination standard exceeds the maximum allowable range, the measured variable m i One of the measured values ​​mv i Total likelihood Ptot(m i For example, it may be determined to be zero. Regarding the maximum permissible range, both application-specific and device-specific maximum permissible ranges can be adopted. As an example, when the measured chlorine concentration Cl2 exceeds the application-specific maximum permissible range to which a swimmer can be safely exposed and / or when the measured chlorine concentration Cl2 exceeds the measurement range of the chlorine sensor M6, [the maximum permissible range is determined by the device]. Figure 2 The chlorine sensor M6 shown in the diagram measures the chlorine concentration Cl2 in swimming pool water. The total likelihood Ptot(Cl2) of the measured chlorine concentration Cl2 is set to zero, for example. As another example, Figure 2 The pH sensor M5 shown can be specified, for example, to operate within a device-specific pressure range, for example because pressures exceeding this range may impair the permeability of its ion-selective membrane. In this case, when the elimination criterion requiring the pressure p measured by the pressure sensor S2 to exceed the device-specific pressure range is met, the total likelihood Ptot(pH) of the pH(multiple) pH values ​​measured by the pH sensor M5 is determined to be zero, for example. Furthermore, if the elimination criterion requiring the flow rate F through the flow cell 11, measured by the flow meter S1, to drop below the absolute minimum flow rate required to perform the measurement of the pH value and / or the chlorine concentration Cl2 is met, the pH sensor M5 will be used to measure the pH value and / or the chlorine concentration Cl2. Figure 2 The total likelihood Ptot(Cl2) of the chlorine concentration Cl2 measured by the chlorine sensor M6 and / or the total likelihood Ptot(pH) of the pH(multiple) pH values ​​measured by the pH sensor M5 are, for example, determined to be zero.

[0117] In some embodiments, this can be achieved by performing an optimization of the total likelihood Ptot(m) iThe ability to further improve the method disclosed in this paper is based on a deterministic iterative process, which involves evaluating and classifying the previously determined total likelihood Ptot(m) by expert operators. i The machine learning method is executed on the labeled training data obtained, and the machine learning method is configured to optimize the total likelihood Ptot(m) i The determination of ) and / or at least one of the following: an application-specific threshold and an application-specific threshold range, used to determine the lookup table and function for the likelihood value PCik, and used to determine the total likelihood Ptot(m i The lookup table and function of the likelihood measure Rin used in determining the likelihood of ) are used.

[0118] In this regard, the previously determined total likelihood Ptot(m) i For example, the method is evaluated by an expert operator, and is classified as "false negative" when it incorrectly determines low likelihood, "false positive" when it incorrectly determines high likelihood, "true negative" when it correctly determines low likelihood, and "true positive" when it correctly determines high likelihood. The labeled training data determined based on these classifications is then fed into a machine learning algorithm configured to optimize the total likelihood Ptot(m) i The determination of the likelihood value PCik and the likelihood measure Rin is achieved, for example, by adjusting at least one of the application-specific thresholds and threshold ranges, and / or by replacing the lookup table and function used to determine the likelihood value PCik with a more detailed correction and / or a more accurate version thereof.

[0119] The method disclosed in this article can be based on a specified variable v. n It begins with a rough estimate of the threshold range specific to the application. As an example, by... Figure 3 The application-specific temperature range of the temperature T of medium 5 measured by the temperature sensor M12 shown may have initially been given as an estimate of 10°C to 15°C. However, the measured temperature T may actually be below 10°C. This will result in a corresponding decrease in the total likelihood Ptot(T) of the temperature T measured by the temperature sensor M12, and may also affect other measured variables m. i Total likelihood Ptot(m i If we evaluate the total likelihood Ptot(m) i If the operator determines that there are no problems associated with either the measuring device Mi or the application, they will classify them as "false negatives." Labeled training data subsequently determined based on these classifications is used to optimize the method, for example, by implementing a machine learning method configured to determine the total likelihood Ptot(m). iThe root cause of false positive and / or false negative determinations, and accordingly modify the total likelihood Ptot(m) i The determination of ). In the example given here, this will result in an adjustment of the temperature range to cover temperatures below 10°C.

[0120] List of reference numerals

[0121] 1 container 25 edge devices

[0122] 3 media 27 equipment

[0123] 5-medium 29 transmitter

[0124] 7 Inlet pipe 31 components

[0125] 9 Inlet Pipe 33 Transmitter

[0126] 11 flow chamber 35 interface

[0127] 13 Inlet pipes 37 Data sources

[0128] 15 outlet pipes, 39 monitors

[0129] 17 Vessels 41 Icons

[0130] 19 Calculation Units 43 Icons

[0131] 21 Memory 45 Temperature Sensor

[0132] 23 Upper-level unit

Claims

1. A method for determining at least one measured variable (m) measured by a measurement system (100, 200, 300) installed in a specific application. i ) measurement value (mv i Application-specific likelihood methods; where, The measurement system (100, 200, 300) includes at least one measuring device (Mi), and each measuring device (Mi) measures the at least one measured variable (m). i The method comprises at least one of the following steps: For each measured variable (m) i ), identify and indicate the measurement of the corresponding measured variable (m) i At least one of the diagnostic information (Iij) of the condition of the measuring device (Mi); Specify at least one variable (v) that is measured at or determined for the specific application. n The number of ) makes the specified variable (v) n ) by the at least one measured variable (m i Each measured variable in the ) gives or includes the at least one measured variable (m) i Each measured variable in ); For each measured variable (m) i ), determine the likelihood criterion (Cik) used to determine the likelihood value (PCik), including: At least one diagnostic criterion (CIij) is used to measure the corresponding measured variable (m) based on the indication. i The value of at least one item in the diagnostic information (Iij) of the condition of the measuring device (Mi) is used to determine the value of the corresponding measured variable (m). i ) measurement value (mv i The likelihood value of the likelihood of ), and At least one threshold criterion (Cj(m) i ), used based on the corresponding measured variable (m) i The measured value (mv) i ) relative to the corresponding measured variable (m) i The measured value (mv) i The magnitude of at least one application-specific threshold range is used to determine the magnitude of the corresponding measured variable (m). i The measured value (mv) i The likelihood value of the likelihood of ) Recorded data (D), which includes at least one item value for each item of diagnostic information (Iij) and each specified variable (v) n The variable value (mv) n The time series of ) and their determination or measurement time (t); and Based on the recorded data (D), perform the following steps at least once: For each measured variable (m) i ): Based on the corresponding measured variable (m) i The likelihood value (PCik) is determined based on the already determined likelihood criterion (Cik). Based on measurements taken before the current time interval and by the corresponding measured variable (m) i The measured value (mv) i ) gives or includes the corresponding measured variable (m) i The measured value (mv) i At least one or all of the specified variables (v) n The variable value (mv) n ), determine the corresponding measured variable (m) indicating the measurement during the finite current time interval. i One or more current measurements (mv) i (tr)) at least one likelihood measure (Rin) of reliability; as well as Based on the corresponding measured variable (m) i The corresponding measured variable (m) is determined based on the already determined likelihood measure (Rin) and the likelihood value (PCik). i The current measurement value (mv) i Total likelihood of (tr) (Ptot(m)) i )),as well as Provide likelihood results (PR), which include at least one of the following: the total likelihood (Ptot(m)). i )) and based on the total likelihood (Ptot(m) i The total likelihood index (TPI) is determined by this.

2. The method according to claim 1, wherein, For at least one or each measured variable (m) i The likelihood measure (Rin) includes at least one of the following: The first likelihood measure (Ri1), which corresponds to the corresponding measured variable (m) i The current measurement value (mv) i (tr) or the current measurement value (mv) i The distribution of (tr) and the corresponding measured variable (m) measured before the current time interval. i ) measurement value (mv i The degree of conformity between the distribution presented or one of multiple distributions; as well as The second likelihood measure (Ri2), which corresponds to the corresponding measured variable (m) i The current measurement value (mv) i (tr) is the degree of agreement between the results of the analysis and the results determined by multivariate analysis of the values ​​of at least two analytical variables determined prior to the current time interval; wherein the analytical variables are determined by the corresponding measured variable (m) i ) and included in the specified variable (v n At least one other variable or each other variable is given in ).

3. The method according to claim 1, wherein, Determine at least one or each total likelihood (Ptot(m)) using the following method. i ): Based on or as a basis for the corresponding measured variable (m) i The sum or product of the likelihood measure (Rin) and the likelihood value (PCik) determined therefrom, or Based on or as a result of the corresponding measured variable (m) i The smallest of the determined likelihood values ​​(PCik) gives the minimum likelihood (Pmin1) and the likelihood for the corresponding measured variable (m). i The sum or product of the likelihood measures (Rin) determined by ) or Determined in the following ways: The likelihood value (PCik) is further subdivided into an operational likelihood value, which includes the likelihood value determined based on the diagnostic criteria (CIij), and an operational likelihood value, which includes the likelihood value determined based on the threshold criteria (Cj(m)). i The application-specific likelihood value determined by the likelihood value, and As or based on the corresponding measured variable (m) i The likelihood measure (Rin) and the application-specific likelihood value determined by the determination of the corresponding measured variable (m) are each of the following: i The sum or product of the minimum likelihoods (Pmin2) given by the smallest of the determined operational likelihood values ​​is used to determine the likelihood for the corresponding measured variable (m). i The total likelihood (Ptot(m)) i )).

4. The method according to claim 3, wherein, The sum is a weighted sum.

5. The method according to claim 3, wherein, The product is a weighted product.

6. The method according to any one of claims 1 to 5, comprising at least one of the following steps: a) Displaying the likelihood results (PR) on a dashboard-style display (39), the dashboard including an icon (41) visualizing the total likelihood index (TPI) and / or a given number of icons (43), each icon visualizing one of the total likelihood values; wherein, A portion of the icon area of ​​each icon is filled with a color corresponding to the size of the visualized total likelihood index (TPI) or the visualized total likelihood value, and the icon is displayed in a color selected according to the size of the visualized total likelihood index (TPI) or the visualized total likelihood value. b) Provide the likelihood result (PR) in the form of a message, which is assigned to a predetermined recipient and / or at least one of predetermined equipment and maintenance tools; as well as c) Providing the likelihood result (PR) to a higher-level unit (23), which is configured to regulate and / or control the process performed at the specific application, and is configured to perform an action to stop or modify at least one process step of the process performed at the specific application and / or at least one other predetermined action when the likelihood result (PR) satisfies the conditions specified for the corresponding action.

7. The method according to claim 6, wherein, A portion of the icon is filled and displayed in a color selected based on the size of the visualized total likelihood index (TPI) or the visualized total likelihood value.

8. The method according to claim 6, wherein, The likelihood results (PR) will be provided via email.

9. The method according to any one of claims 1 to 5, wherein, The specified variable (v) n The measurement system (100, 200, 300) includes at least one of the following: at least one process parameter measured by one of the measuring devices (Mi) of the measurement system (100, 200, 300); at least one process parameter measured by a measuring instrument (Sm) installed at the specific application location; and at least one diagnostic parameter determined by or for one of the measuring devices (Mi).

10. The method according to any one of claims 1 to 5, wherein: For at least one of the measuring devices (Mi), the at least one item in the diagnostic information (Iij) indicating the condition of the corresponding measuring device (Mi) includes at least one of the following: the lifespan of the measuring device (Mi), the operating time of the measuring device (Mi), the maintenance time when repairing the measuring device (Mi), the verification time when verifying the measurement accuracy of the measuring device (Mi), the verification result obtained by verifying the measurement accuracy of the measuring device (Mi), the calibration time when calibrating the measuring device (Mi), the calibration result obtained by calibrating the measuring device (Mi), at least one diagnostic parameter determined by the measuring device (Mi), a status index determined based on self-diagnosis performed by the measuring device (Mi), and an exposure index corresponding to the measuring device (Mi) being exposed to adverse measurement conditions.

11. The method according to any one of claims 1 to 5, wherein: a) For at least one measured variable (m) i ), for the corresponding measured variable (m) i The likelihood criterion (Cik) determined includes at least one of the following: Used based on the corresponding measured variable (m) i The current measurement value (mv) i (tr) Whether it appears in the measured variable (m) i The threshold criterion for determining the likelihood value is the range that will not exceed the application-specific threshold. Used based on the measured variable (m) i ) measurement value (mv i The corresponding measured variable (m) is determined by the application-specific probability of occurrence within an application-specific threshold range. i At least one current measurement value (mv) i The threshold criterion for the likelihood value of the likelihood (tr) and the likelihood of the likelihood. Used based on at least one other variable (v) j At least one current variable value (mv) n (tr)) and determine the corresponding measured variable (m) i At least one current measurement value (mv) i The standard for the likelihood value of the likelihood (tr) is given, where each other variable (v) j ) by other measured variables (m j≠i One of or included in the specified variable (v) n The parameters are given in ), and / or where: b) Regarding the measured variable (m) i At least one of the following, wherein the likelihood measure (Rin) includes at least one of the following: Based on the detection included in the corresponding measured variable (m) i The measured value (mv) i The likelihood measure is determined by the method of identifying outliers in ( ); Based on the corresponding measured variable (m) i The at least one current measurement value (mv) i The corresponding measured variable (m) measured before the measurement time (tr) of (tr) i ) measurement value (mv i And a definite likelihood measure; Based on the corresponding measured variable (m) in the specific application. i The at least one current measurement value (mv) i The measured value of the size of (tr) (mv) i The probability of occurrence of () and a likelihood measure determined by a combination of their occurrence probabilities based on an empirical distribution, which is based on the current measurement (mv). i The corresponding measured variable (m) is measured during a finite time interval prior to the measurement time (tr). i ) measurement value (mv i And determined by; and Based on the corresponding measured variable (m) i The at least one current measurement value (mv) i (tr) and based on the current measurement value (mv) i The corresponding measured variable (m) measured before the measurement time (tr) of (tr) i ) measurement value (mv i The likelihood measure is determined by the deviation between the predicted and corresponding predicted values, wherein the predicted values ​​are based on fitting to previously determined measurements (mv). i The time series of the data (D) is determined by an autoregressive integral moving average model, or based on the corresponding measured variable (m) already determined by a machine learning method based on the data (D). i The measured value (mv) i It can be determined by the model, or by another time series forecasting method.

12. The method according to any one of claims 1 to 5, wherein: For each likelihood criterion (Cik), the corresponding likelihood value (PCik) is determined based on a lookup table or likelihood function (f(Cik)) associated with the corresponding likelihood criterion (Cik), which assigns the likelihood value (PCik) to the corresponding measured variable (m) based on at least one attribute. i The current measurement value (mv) i (tr)), the at least one attribute is comprised of at least one of the item values ​​of the items in the diagnostic information (Iij) and / or the specified variable (v) included in the data (D). n The value of at least one of the variables (mv) n At least one of them is given or can be determined based on it.

13. The method according to any one of claims 1 to 5, wherein, For the measured variable (m) i At least one of the following, the determined likelihood measure (Rin) includes a likelihood measure determined by the following steps: Based on the corresponding measured variable (m) measured during the previous time interval i ) measurement value (mv i The magnitude of the measured value (mv) along the line is relative to the line. i Sort them. The line is subdivided into four quartiles (Q1, Q2, Q3, Q4), each quartile including the measured value (mv). i A quarter of ) and The likelihood measure is determined based on a likelihood function, which is based on the corresponding measured variable (m). i The current measurement value (mv) i (tr)) The quartiles (Q1, Q2, Q3, Q4) appearing therein will be the corresponding measured variables (m) i The current measurement value (mv) i The likelihood measure (tr) is assigned to all current measurements (mv) that appear at positions (q) within a predetermined first range of the line. i (tr)), and / or based on the current measurement (mv) i The measured value of the size of (tr) (mv) i The probability of occurrence of ) will be used to determine the corresponding measured variable (m). i The current measurement value (mv) i The likelihood measure (tr) is assigned to all current measurements (mv) that appear at positions (q) within a predetermined second range of the line. i (tr)); wherein, the measured value of the size included in the second range (mv) i The probability of occurrence of ) is based on training data included in the data (D) or on the measurement value (mv) measured during the previous time interval. i It is determined by ( ).

14. The method according to any one of claims 1 to 5, wherein, For the measured variable (m) i At least one of the following, the determined likelihood measure (Rin) includes a likelihood measure determined based on a likelihood function, which is determined according to, as, or based on a first likelihood (fa(mv)) determined based on a first likelihood function (fa). i (tr))) and the second likelihood (fb(mv) determined based on the second likelihood function (fb) i The sum or product of (tr) assigns the likelihood measure to the corresponding measured variable (m). i The at least one current measurement value (mv) i (tr)), where: The first likelihood function (fa) is based on the current measurement value (mv) at the specific application. i The measured value of the size of (tr) (mv) i The probability of occurrence of ), and the first likelihood (fa(mv)) i (tr))) is assigned to the current measurement value (mv) i (tr)); The first likelihood function (fa) is based on a measurement (mv) of a given magnitude at the specific application. i It is determined by an estimate of the probability of occurrence of ), or based on measurements of different magnitudes (mv). i The different sizes of the measurements are determined by the frequency of occurrence of the measurements (mv) during the training time interval. i The training time interval is determined to cover a sufficiently long time span to cover all operating modes and / or each process performed at the specific application; The second likelihood function (fb) is based on at least one current measurement (mv). i (tr)) The second likelihood (fb(mv) is determined based on the probability of occurrence of the empirical distribution. i (tr))) is assigned to the at least one current measurement value (mv) i (tr)), the empirical distribution is based on the current measurement value (mv) i The measured value (mv) is measured during a finite time interval prior to the measurement time (tr). i And determined by; and The second likelihood function (fb) is determined as or based on measurements of different magnitudes (mv). i The frequency of occurrence of ) or is determined as a probability function based on kernel density estimation (KDE), wherein the different sizes of the measurements are based on the measurements (mv) already measured during the finite time interval. i The probability function is determined by the fact that the corresponding measured variable (m) represents the probability function. i ) measurement value (mv i The probability of occurrence is based on its size.

15. The method according to claim 14, wherein, The sum is a weighted sum.

16. The method of claim 14, wherein, The product is a weighted product.

17. The method according to any one of claims 1 to 5, for the measured variable (m) i At least one of the following, further comprising the following steps: Identify the corresponding measured variable (m) i The measured value (mv) i At least one elimination criterion (KO) for the likelihood of the above. as well as Execute the corresponding measured variable (m) i The measured value (mv) i The total likelihood (Ptot(m)) i The determination of )) ensures that when the corresponding elimination criterion (KO) is met, the total likelihood (Ptot(m)) is determined such that the total likelihood (Ptot(m)) is [value]. i The value is set to zero or reduced to take into account the impact of meeting the corresponding elimination criterion (KO) on the corresponding measured variable (m). i The measured value (mv) i The extent to which the likelihood is affected; Among them, for the corresponding measured variable (m) i The identified elimination criteria (KO) include at least one of the following: Compared with the measurement of the measured variable (m) i One of the relevant elimination criteria for each item of the diagnostic information identified by the measuring device (Mi). The requirement is to measure the measured variable (m) i The state indicators determined by the measuring device (Mi) are used to indicate the criteria for rejection of the measuring device (Mi) due to defects. Requirements for the measured variable (m) i The measured value (mv) i ) and / or included in the specified variable (v n At least one other variable (v) in ) j The variable value (mv) j Exceeding the maximum allowable range, exceeding a given threshold, or falling below a given threshold are elimination criteria. With included in the specified variable (v) n The elimination criteria are related to at least one parameter in the measuring device (Mi) or by a measuring instrument (Sm) installed at the specific application.

18. The method according to any one of claims 1 to 5, wherein, The measurement system (200) is an analysis system, wherein the measuring device (Mi) measures the measured variable (m) of the medium (5) flowing through the flow chamber (11). i ), where the specified variable (v n ) includes the flow rate (F) of the medium (5) flowing through the flow chamber (11), and wherein: The likelihood criterion (Cik) includes methods for determining the measured variable (m) based on the measured flow rate (F). i One of the measured values ​​(mv) i At least one criterion of the likelihood value (PCik) of the above, and / or When the flow rate (F) through the flow chamber (11) decreases to the level at which the corresponding measured variable (m) is measured... i When the absolute minimum flow rate required is below, the at least one measured variable (m) i At least one or each of the measured values ​​(mv) i The total likelihood (Ptot(m)) i Set it to zero.

19. The method according to any one of claims 1 to 5, comprising the method steps of: [the following steps are described in the original text, but the provided text is incomplete and requires further context to translate accurately.] i An expert operator obtains labeled training data and performs an iterative process to optimize a machine learning method, which is configured to optimize the total likelihood (Ptot(m)). i The determination and / or optimization of at least one of the following: an application-specific threshold, the application-specific threshold range, a value used to determine the likelihood value (PCik), and a value used to determine the total likelihood (Ptot(m)). i The lookup table and function for the likelihood measure (Rin).

20. The method according to any one of claims 1 to 5, wherein: The method described is a computer-implemented method. Each likelihood result (PR) is determined and provided by a computing device (19) configured to determine and provide the likelihood result (PR) based on recorded data (D) and on a computer program (SW) implemented on the computing device (19); and The computing device (19) is included in the measurement system (300), or in a transmitter (33) connected to or connected to the measurement device (Mi), or in a device (27), edge device or upper unit located near the measurement system (200) and connected to or communicating with the measurement device (Mi), or embodied in the cloud.

21. A computer program (SW) product comprising instructions that, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 20 based on data (D) provided to the computer.

Images