Method and system for diagnostic messaging

Abstract

The invention relates to a method and a system for diagnosing messages (M1, M2, ...) generated during the operation of a technical installation (1). A process (2) of a technology running in the installation (1) is controlled by means of a process control system (3), and a data set with values of process variables (L, T, F, ...) representing the operation of the installation is detected and stored in a data memory (15). A diagnostic statement about the operation of the installation (1) is determined based on the data set and at least one predetermined, so-called self-organizing map. At the nodes of the self-organizing map, n-tuples corresponding to the data set of predetermined values of the process variables (L, T, F, ...) for error-free operation of the installation are stored. The invention is characterized in that, after the occurrence of a message (M1) for a current and / or historical time interval, a minimum distance between the data set of the current time interval (t_akt) and the data set of the nodes (K) of the self-organizing map in which the message (M1) occurred is determined, and one or more corresponding nodes (Kres) are selected. Symptoms are then determined by forming differences between the time interval and a previously determined data set for one or more nodes (Kres), with only differences in process variables exceeding a preset threshold value (S_Sym) being considered. The process variables (L, T, F, ...) associated with the previously determined symptom are output. This supports the facility operator in diagnosing process messages.

Description

Technical Field

[0001] The invention relates to a method for diagnosing messages generated during the operation of a technical installation, wherein the technical processes running within the installation are controlled by means of a process control system. The invention also relates to a corresponding system with an associated computer program for diagnosing messages of a process control system. Background Art

[0002] In manufacturing and process technology facilities, technical processes are controlled using automation systems. The flow of individual production steps or process steps can be subject to fluctuations caused by production-related factors as well as by faults within the technical facilities. Therefore, process monitoring is necessary for safe operation and the achievement of qualitative targets. Abnormal deviations caused by faults must be distinguished from normal production or process fluctuations. These abnormal deviations caused by faults are called anomalies.

[0003] The presence of an anomaly is in principle a binary conclusion, but is usually combined with at least one other value, for example a value which quantifies the degree of deviation from the norm.

[0004] However, if an anomaly exists, scalar knowledge about it is not sufficient to understand it and plan the next steps. It is important to provide symptoms. A symptom is an indicator or indication of a deviation from the normal state. In the VDI / VDE Guideline 2651 of May 2017, page 1, "Plant Asset Management (PAM) in der Prozessindustrie", page 19, the term "symptom" is explained: Therefore, a symptom is derived from a characteristic (= characteristic variable, i.e., pressure, temperature, or material level) by comparison with a reference variable. Thus, for example, the characteristic "winding temperature of the motor" alone is not a symptom, because the absolute value does not indicate a possible overload. Only comparison with a maximum permissible value or a reference value (e.g., a nominal value) can provide conclusions about the deviation. Therefore, symptoms can be derived, for example, by comparison with target values, limit values, nominal values, or empirical values, by trend monitoring, or by comparison with a model.

[0005] The subsequent diagnosis or cause determination evaluates the symptom as a result of monitoring and determines the influencing force or process variable that triggered the symptom.

[0006] Process messages generated by automation systems alert operators in process plants to deviations from normal operation caused by faults. In addition to user intervention (operational messages), process messages typically indicate departures from the normal range of individual process variables (warnings), reaching critical values (alarms), and diagnostic information from the automation system (system messages). Due to their critical nature, alarm messages, in particular, require immediate, or at least timely, action by operators of the technical plant to prevent quality losses, production failures, plant downtime, and, in the worst case, accidents involving personnel.

[0007] Therefore, alarms in process technology systems are one of the most important auxiliary measures for operators to monitor processes and detect deviations from normal operation. Alarms are usually implemented as fixed or variable limit values for measured or derived (calculated from other measured variables) variables. If a limit value is violated, the operator is notified by means of an alarm.

[0008] However, in many cases, alarm information alone is insufficient to provide a detailed understanding of a variable's deviation from its target value, or even the cause of the deviation. Even worse, a single error condition often triggers multiple alarms, making it difficult to identify the correct response in a timely manner. In the worst-case scenario, an operator's incorrect intervention can exacerbate the deviation and lead to disaster.

[0009] For these reasons, operators must first analyze the cause of a process message in order to take appropriate action. To identify the cause, they must analyze various measured variables, time curves, and KPIs. However, this varies depending on the message and cause, making it impossible to always associate relevant situations with alarms in advance. Navigating through the various values in the different overview views on the operator station is time-consuming and can also lead to confusion between components or signals. However, depending on the problem, detailed analysis can be difficult, especially with the increasing engineering dependencies between linked subprocesses. Therefore, it is particularly desirable to be able to identify individual signals and / or process variables as the cause of a message. Summary of the Invention

[0010] The object of the present invention is therefore to provide a method for diagnosing process messages of a process control system of a technical installation and a corresponding suitable system, characterized by improved cause analysis when messages occur. Based on this, the object underlying the invention is to provide a suitable computer program.

[0011] This object is achieved by the features of the invention.Furthermore, the object is achieved by a computer program according to the invention and a computer program product according to the invention.

[0012] Configurations of the invention, which can be used individually or in combination with one another, are the subject matter of the individual exemplary embodiments.

[0013] The present invention proposes an improved method for diagnosing messages generated in a process control system during the operation of a technical installation. According to the invention, the method uses a self-organizing map to determine diagnostic information, based on which the normal behavior of the installation operation can be mapped and thus determined. By determining the behavior of corresponding observed data sets at specific time intervals that deviates from normal installation operation in a purely data-driven manner, the causes of the triggering of messages and, therefore, the deviations from the normal operation of the installation can be determined more precisely and reliably.

[0014] “Self-organizing maps”, Kohonen maps or Kohonen networks (after Teuvo Kohonen; in English: self-organizing map, SOM or self-organizing feature map, SOEM) are well known in the prior art. According to Wikipedia as of February 20, 2020, a self-organizing map is an artificial neural network. As an unsupervised learning method, it is a powerful data mining tool. Its operating principle is based on the biological knowledge that a large number of structures in the brain have a linear or planar topology. The article “Monitoring of Complex Industrial Processes based on Self-Organizing Maps and Watershed Transformations” by Christian W. Frey from the Fraunhofer Institute IOSB (available on the Internet on March 11, 2016 at the address www.iosb.fraunhofer.de / servlet / is / 22544 / Paper_ICIT2012_Frey.pdf? command=downloadContent&filename= Paper_ICIT2012_Frey.pdf A diagnostic method for monitoring complex industrial processes is known in the literature. A self-organizing map is trained for error-free process behavior based on process variables, i.e., the detected values of measured variables and the values of controlled variables output to the process. Subsequent operating behavior is compared with the trained error-free behavior based on the self-organizing map defined in accordance with the method. Deviations are thus identified and analyzed for possible errors and causes in the process operation.

[0015] EP 3 279 756 B1 also discloses a diagnostic device and a method for monitoring the operation of a technical installation, wherein reliable diagnostic information is obtained in a step chain control by using one or more self-organizing maps.

[0016] Before a self-organizing map can be used, it must first be trained using so-called good data. During the automated learning process, for each node of the self-organizing map, a data memory stores multiple data sets representing error-free operation of the plant. For each node of the map, a learning method calculates an n-tuple containing the values of the process variables that are operating correctly, based on these data sets, and stores it at the node—that is, stores it in association with the node. The map thus predefined can be used directly for further analysis. To train the self-organizing map, the values of the process variables that are operating normally and correctly are used as training data. During training, the self-organizing map stores typical values of the process variables that are operating normally as good values at each node. For example, at the beginning of the training method, a map with a size of 8×12 nodes can be used. Obviously, other sizes can also be used. After training, the map size can be checked using the training data. Therefore, the new diagnostic method advantageously requires little operator knowledge of the plant to be monitored and can be used universally in practice. If, during a diagnostic run, the values of the process variables deviate significantly from the good values stored at the nodes of the previously trained self-organizing map, this indicates an error in the monitored plant. Only for a subsequent analysis of the cause of the error may more detailed knowledge of the process running in the respective monitored installation be necessary.

[0017] The present invention utilizes it to carry out message diagnosis:

[0018] In its simplest implementation variant, in a first step, the data set of the process variable of the current time interval in which the message occurred is checked to determine in which area of the SOM the data set is located, i.e., where the current data set is located compared to normal operation, as represented by the nodes of the SOM. This is done by determining the minimum distance between the data set of the current time interval and the data sets of all nodes of the self-organizing map.

[0019] In this case, any mathematical distance measure can be chosen. Thus, for example, the Cartesian distance or the Manhattan distance can be used for the distance between two points.

[0020] The node resulting from the calculation is now selected as a reference for the subsequent symptom determination. The reference node is often also called a winning neuron or winning node, since it corresponds to the current normal operation of the facility, i.e. error-free operation. The symptom associated with the current time interval is determined by subtracting the data set from the current time interval and the previously determined reference node, wherein only the difference in the process variables exceeding a preset threshold value is taken into account. The process variables related to the previously determined symptom are output. In this embodiment, the method according to the invention thus makes it possible to determine process variables that are characteristic for the current facility state and to support the facility operator when diagnosing process messages. The same advantages apply to a corresponding system for diagnosing messages of a process control system of a technical facility, wherein the system comprises: at least one data memory in which at least one data set characterizing the facility operation with values of process variables can be stored; and at least one evaluation device, and the evaluation device is designed to carry out the method according to the embodiment described further above.

[0021] The method according to the invention is particularly suitable for continuous processes, since specific operating points of normal operation in continuous processes are reflected in the nodes of the self-organizing map.

[0022] The term "system" can refer to a hardware system, such as a computer system consisting of servers, networks, and storage units, or a software system, such as a software architecture or larger software program. Hybrid systems of hardware and software are also possible, such as IT infrastructure, such as cloud infrastructure and its services. The components of such infrastructure typically include servers, storage, networks, databases, software applications and services, data directories, and data management. Virtual servers, in particular, also fall into this category of system.

[0023] In another embodiment of the present invention, in addition to the data set of process variables for the current time interval in which the message occurred, the frequency of past occurrences of the message is also examined. Thus, historical data sets for past time intervals containing the same message are also considered when determining the symptom. In this case, by determining the minimum distance between the data set for the past time interval and the data set for the node, a number of winning nodes are identified in the self-organizing map, which correspond to error-free operation of the facility. Accordingly, historical symptoms are also determined by subtracting the data set of the historical data interval from the previously determined winning nodes, only considering process variable differences that exceed a preset threshold. The process variables associated with the previously determined historical symptoms are output. By considering historical time intervals with the same message, the selection of process variables associated with messages, particularly alarms, becomes more robust. For example, if the same process variable is displayed for the current time interval and for the historical time intervals after the symptom is determined, the facility operator can assume that the same interaction of the process variables always leads to a display disturbance.

[0024] In a particularly advantageous embodiment variant, the robustness of message diagnosis is further increased by also considering individual messages in relation to other messages. After a message appears, in this embodiment variant, it is first determined whether the message is part of a message chain and how often message chains have occurred in the past. In the further course of the method, the same method as for individual messages is used for any selected time interval containing a message chain: After determining the winning node of the self-organizing map for the dataset containing the current and historical time intervals of the message chain, current and historical symptoms are determined and evaluated.

[0025] Message chains are messages that typically appear in a defined sequence or following a recurring pattern in the message archive. Message chains often provide valuable information about operational behavior, as one event is the result of another. Based on this information and on facility and process knowledge, operators can then make better decisions regarding the further operation of the facility.

[0026] For example, EP 3 454 154 A1 discloses an automated method for identifying statistical correlations between process messages. The method for determining message chains is designed to analyze a given data set (e.g., all messages of a message archive within a specific time range) at once and to generate a closed data set with results, which consist of a table with message chains and a matrix with the corresponding transition probabilities between specific messages.

[0027] The determination of message chains can, for example, be handled by a dedicated software module. In this case, it is particularly advantageous if the evaluation unit of the system for diagnosing messages according to the present invention is connected to an analysis device designed to determine and analyze message chains. Alternatively, the determination of message chains can be integrated into the message diagnosis, for example to achieve a simplified data structure.

[0028] If a symptom is very similar at all times in a message chain or within a considered time interval, the combined set of process variables involved in the symptom can be visualized along with the current symptom. Therefore, when a large number of different process variables are obtained from the determination of current and historical symptoms, it is recommended to cluster the symptoms in order to reduce the number of process variables involved in the symptom. Therefore, in another advantageous embodiment of the present invention, the symptoms are clustered, and then a cluster containing, for example, the current symptom is selected, and the process variables involved in the current symptom are output. Any known clustering method is suitable for this purpose.

[0029] In another preferred embodiment of the method and system according to the present invention, all time intervals can be variably determined or adjusted according to the process dynamics. This allows the user of the diagnostic application to have unlimited flexibility when evaluating the data. If time intervals are selected over multiple time units (e.g., seconds or minutes), trend curves of the process variables can be advantageously displayed.

[0030] In the case of message chains, the time interval can be selected based on the identified chain. Thus, for example, the time interval can begin at the beginning of the message chain, i.e., at the time of the first message, and end at the time of the last message. However, depending on the process dynamics, the analyzed time period can also be extended before and after the occurrence of the message chain (e.g., 10 minutes before the chain starts to 5 minutes after the chain ends).

[0031] In a further preferred embodiment of the invention, the symptom is determined for an arbitrarily selected point in time of the time interval. This has the advantage that fewer data sets need to be calculated when determining the symptom, thus enabling a faster determination, ie reducing the calculation time.

[0032] In another advantageous variant, a threshold value is variably determined to identify the symptom. The threshold value used to determine the symptom indicates how large the difference between the data set at the time interval of the message and the data set of the winning node is allowed to be at a certain point in time. If the threshold value is selected relatively low, the data set associated with the message is similar to good data, that is, a data set of normal operation stored in the node of the SOM. The symptom is then determined or inconclusive because the detected deviation from normal operation is too small. If the threshold value is selected too high, too many process variables may be taken into account when forming the symptom, which may distort the message diagnosis. Similarly, as with the training of nodes in a self-organizing map, the threshold value can be automatically determined based on a data set that operates correctly. To this end, the corresponding winning node can be determined using the data set, and the corresponding distance between the data set and the associated winning node can be determined. The threshold value can then be calculated and predefined by increasing the safety margin by, for example, 5 to 50%, preferably 15%, in each case to avoid incorrect diagnoses.

[0033] In another advantageous embodiment, the message diagnosis is refined by combining additional information with the symptom determination. Advantageously, measurement data from similar historical events can also be displayed in parallel. If the causes and solutions are stored in a shift log, this information can be even more useful to the facility operator. Data from similar events themselves, for example, allows conclusions to be drawn about how a process will continue to behave. This can also be very useful when diagnosing messages, such as alarms.

[0034] In one embodiment variant, the system according to the present invention is part of a computer system that is physically separated from the premises of the technical installation. The connected second system then advantageously has an evaluation unit that has access to components of the technical installation and / or data storage devices connected thereto and is designed to visualize the analysis results and transmit them to a display unit. This allows, for example, coupling to a cloud infrastructure, which further increases the flexibility of the overall solution.

[0035] A local implementation on a computer system of a technical installation can also be advantageous. Thus, for example, an implementation on a server of a process control system is particularly suitable for safety-related processes.

[0036] The aforementioned object is correspondingly also achieved by a user interface (GUI, graphical user interface) displayed on a display unit, which is designed to display analysis results of all variants of the system according to the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The invention will be described and explained in more detail below based on the drawings and on the basis of exemplary embodiments.

[0038] The accompanying drawings show

[0039] Figure 1 An example of a technical installation is shown,

[0040] Figure 2 An example of the process is shown. DETAILED DESCRIPTION

[0041] Figure 1 A simplified schematic diagram is shown as an example of a process plant 1, in which a process 2 is controlled and / or monitored by means of an automation system or process control system 3. The process control system 3 comprises a planning and engineering tool 4, an operating and monitoring device 5, and a plurality of automation devices 6, 7, 8 which are connected to one another via a bus 9 for data communication. The automation devices 6, 7, 8 control the method-technical process 2 according to the provisions of an automation program, one of which 10 is shown as an example in FIG. Figure 1. For example, the automation program 10 usually consists of a plurality of functional modules, which can interact with other functional modules distributed in the automation system 3. In order to control the process 2, a variety of field devices 11, 12, 13, 14 are used for process coordination. Measuring transducers are used to detect process variables, such as temperature T, pressure P, flow rate, material level L, density of the medium or gas concentration. By means of regulating mechanisms, the process flow can be influenced accordingly according to the detected process variables, for example according to the presets of the automation program 10. As examples of regulating mechanisms, control valves, heating devices or pumps are listed. In order to monitor the operation of the facility 1, a large number of data sets for characterizing the operation of the facility are detected and stored in a data memory 15. The data sets containing the values of the process variables at a specific time are evaluated by means of an evaluation device 16 in order to determine a diagnostic statement and display it to the operator, so that appropriate measures can be taken to deal with the error.

[0042] In an automated environment, the evaluation device for executing the method according to the present invention can advantageously be designed as a software function module that can be linked to the function modules of the automation program in a graphical user interface of an engineering system and, for example, loaded into an automation system. Detected deviations of process variables, indicating errors in the system, are then displayed on a so-called faceplate for implementing the human-machine interface on an operating and monitoring device 5 of the automation system. If desired, the operator can make changes to the self-organizing map, threshold values, or other parameters via the graphical user interface of the operating and monitoring device.

[0043] The diagnostic system according to the present invention for monitoring the operation of technical installations, particularly data storage and evaluation devices, can be implemented particularly advantageously in a non-local software environment for cloud-based facility monitoring. Data from customer installations is collected, aggregated, and transmitted to a service operations center using software agents, where it is stored on a remote service computer. There, in the cloud environment, the data is semi-automatically evaluated using various data analysis software applications. If required, trained experts in remote service can efficiently work on this database. The results of the data analysis can be displayed on the remote service computer's monitor and / or made available on Sharepoint, allowing the end customer, i.e., the operator of the technical installation, to view them, for example, in a browser.

[0044] The method according to the present invention is therefore preferably implemented in software or a software / hardware combination. Thus, the present invention also relates to a computer program having program code instructions executable by a computer for implementing the diagnostic method. In this context, the present invention also relates to a computer program product, in particular a data carrier or storage medium, having such a computer program executable by a computer. As described above, such a computer program can be stored or preloaded in the memory of an automation device to automatically monitor the operation of a technical installation during operation of the automation device, or the computer program can be stored or loaded in the memory of a remote service computer for cloud-based monitoring of a technical installation.

[0045] With the help of Figure 2 An exemplary embodiment for the diagnosis of messages of a process control system according to the invention is described in more detail.

[0046] Consider a reactor R, in which a liquid mixture is produced in this exemplary embodiment, wherein an exothermic reaction occurs. Raw materials are continuously conveyed and products are discharged. The conveying rate is represented by the process variable F_in, and the outflow rate by F_out. The material level L in the reactor R can fluctuate during the reaction. In the reactor R, the mixture is stirred at the stirrer speed n and an exothermic reaction occurs. Therefore, the reactor must be cooled. The cooling power is represented by Q_c. The external temperature T_amb and the temperature T inside the reactor are continuously monitored. The pH value of the mixture in the reactor is also monitored as the process variable Q_ph. All physical measured variables mentioned are data sets with process variable values, which are detected by means of corresponding sensors and measuring transducers and stored in a data memory.

[0047] Assuming that a corresponding system or software component is installed for diagnosing the process in question, a self-organizing map (SOM) is mapped in the software component. Furthermore, it is assumed that the self-organizing map is trained using historical "good data," i.e., a training dataset with values of process variables that characterize error-free operation of the technical installation. During training, the SOM stores typical values of the process variables that characterize error-free operation of the technical installation at each node as good values. Therefore, each node is an n-tuple containing predetermined values of n process variables for error-free operation.

[0048] In this example, let's assume that, for example, the alarm "Temperature T in reactor too high" is displayed on a screen of a process control system. In a real process with multiple process variables, it would be very unclear if all measured variables associated with the process variables were displayed. In accordance with the method according to the present invention, only those measured variables that are causally associated with one or more occurring alarms are displayed.

[0049] Following the alarm, the warning "Very high fill level" and the message "Compressed air compressor started" are also displayed. According to the method according to the invention, a message chain (M1, M2) with M1 = "Temperature T in reactor too high" and M2 = "Very high fill level in reactor" is first identified based on a chain analysis. Since the compressed air compressor is started independently of the process as soon as the compressed air supply falls below a certain pressure, this means that the message "Compressed air compressor started" is irrelevant to the evaluation. The frequency of the message chain (M1, M2) occurring in the past is then determined. In this embodiment, it is assumed that the message chain occurred 27 times in the past year. If no message chain is identified, only the frequency of occurrence of the alarm is determined. For all time intervals in which the message chain occurs, a comparison with historical good data is now performed. For example, in a message chain, the time point at which the message chain begins can be determined as the beginning of the time interval to be considered. Generally, the time interval can be determined variably or adjusted according to the process dynamics. For example, the time interval can extend from 10 minutes before the start of the message chain to 5 minutes after the end of the chain.

[0050] For all time intervals (here, the current time interval t_akt and 27 historical time intervals dti, where i=27), i+1 "winner nodes" are now determined by, for example, determining the smallest Cartesian distance (here, between a data set containing process variable values for time intervals i+l=28) and the n-tuple at the node of the SOM. The 28 resulting nodes Kresi are selected. Current and historical symptoms are then determined by subtracting the data sets from the time intervals from the previously determined nodes Kresi, only considering differences in process variable values that exceed a predefined threshold value S_Sym. If the symptom can be clearly identified using the predefined threshold value, there is no need to lower the threshold value.

[0051] In the case under consideration, the symptoms of the 27 historical data sets differed, so clustering was performed. This resulted in two clusters, which means that there seem to be two process situations that have occurred many times in history and differ in their symptoms:

[0052] Cluster 1:

[0053] ΔL=+5, ΔT=+10, ΔF_in=+2, ΔF_out=0, ΔQ_c=+3, Δn=0, ΔQ_ph=O, ΔT_amb=0

[0054] and

[0055] Cluster 2:

[0056] ΔL=+6, ΔT=+11, ΔF_in=0, ΔF_out=-3, ΔQ_c=+3, Δn=0, ΔQ_ph=0, ΔT_amb=0

[0057] Within a cluster, the behaviors are respectively similar, so the focus of the corresponding cluster is interpreted as a symptom. Alternatively, historical situations contained in the cluster can also be selected.

[0058] In this example, the current symptom is in cluster 2. Therefore, the process variables of fill level L, temperature T in the reactor, outflow rate F_out, and cooling power Q_c are displayed. The plant operator can conclude from this that a reduced outflow indicates an excessively high temperature in the reactor, so that, for example, the outflow must be increased manually.

[0059] Although the present invention has been described in detail by way of preferred embodiments, the present invention is not limited to the disclosed examples and variations thereof will be apparent to those skilled in the art without departing from the scope of the present invention.

Claims

1. A method for diagnostic messages, which are generated during operation of a technical installation, wherein: Controlling a method-technical process running in the installation by means of a process control system (3) and recording a data set characterizing the operation of the installation with values of process variables and storing the data set in a data memory (15), wherein diagnostic information about the operation of the installation is determined using the data set and at least one predetermined so-called self-organizing map, and n-tuples corresponding to the data set are stored at at least one predetermined node of the self-organizing map, said n-tuples having predetermined values of process variables for error-free operation of the installation. It is characterized by: - after a message appears, determining, for the current time interval in which the message appears, the minimum distance between the data set of the current time interval and the data sets of all nodes of the self-organizing map, and selecting the corresponding node, - after the occurrence of the message, further determining: a number i characterizing how often the message has appeared in the past, whether the message is part of a message chain, and a number j characterizing how often the message chain has appeared in the past, - determining, for a number of historical time intervals in which the message appeared and for all time intervals in which the message chain appeared, the minimum distance between the data sets of the current time interval and the historical time intervals and the data sets of the nodes of the self-organizing map, and selecting the corresponding node, - determining current symptoms and historical symptoms by forming differences from data sets of said current time interval and said historical time interval with said previously selected nodes, wherein only said differences are considered for process variables exceeding a preset threshold, and - outputting a process variable related to the current symptom and the historical symptom previously determined.

2. The method according to claim 1, It is characterized by: performing clustering of the current symptoms and the historical symptoms, A cluster containing at least one current symptom is selected, and a process variable related to the at least one current symptom is output.

3. The method according to claim 1 or 2, characterized in that The current time interval and the historical time interval can be determined variably, or can be set dynamically according to a process.

4. The method according to claim 1 or 2, characterized in that Symptoms are determined for any selected time points of the current time interval and the historical time interval.

5. The method according to claim 1 or 2, characterized in that Threshold values can be variably determined to determine the current symptom and the historical symptom.

6. The method according to claim 2, characterized in that For a clustered symptom, the historical process variables of the historical time intervals related to the symptom in the cluster are also output.

7. The method according to claim 1 or 2, characterized in that Information from the shift log of the technical installation is also output with the symptom, wherein information from a short time interval before and / or after the symptom is also taken into account.

8. A system for diagnosing messages of a process control system (3) of a technical installation, wherein The system comprises a data memory (15) in which at least one data set characterizing the operation of the installation and having values of process variables can be stored; and an evaluation device (16), wherein the evaluation device (16) is designed to: determining diagnostic information about the operation of the installation using the data set and at least one predetermined so-called self-organizing map, wherein n-tuples corresponding to the data set are stored at nodes of at least one predetermined self-organizing map, said n-tuples having predetermined values of process variables for error-free operation of the installation, It is characterized by: - the evaluation device (16) is connected to an analysis device, which is designed to detect and analyze the message chain, and The evaluation device (16) is further designed to carry out the method according to claim 1 after the occurrence of the message, and The evaluation device is further designed to output a process variable related to a previously determined symptom by means of the method according to claim 1 .

9. The system according to claim 8, It is characterized by: The evaluation device (16) is further designed to perform a clustering of previously determined symptoms and to select a cluster containing at least one current symptom and to output a process variable related to the at least one current symptom.

10. The system according to claim 8 or 9, It is characterized by: The evaluation device (16) is further designed to variably determine the current time interval and the historical time interval or to dynamically set the current time interval and the historical time interval according to a process.

11. The system according to claim 8 or 9, It is characterized by: The evaluation device (16) is further designed to determine symptoms for any selected time points of the current time interval and the historical time interval.

12. The system according to claim 8 or 9, It is characterized by: The evaluation device (16) is further designed to be able to variably determine threshold values for determining the current symptom and the historical symptom.

13. The system according to claim 8 or 9, It is characterized by: The evaluation device (16) is further designed to output the historical process variable for the historical time interval for symptoms of clustering.

14. The system according to claim 8 or 9, It is characterized by: The evaluation device (16) is connected to a component having a shift log, and the component is connected to an analysis device, which is designed to evaluate the information of the shift log of the technical installation in conjunction with symptoms and output this information to the evaluation device (16).

15. The system according to claim 8 or 9, It is characterized by: The evaluation device (16) is also connected to a display device, which is designed to display a selection of the following information: messages or determined message chains, the frequency of occurrence of messages or message chains, a selection of current and historical time intervals and process variables, process variables related to historical symptoms and / or process variables related to current symptoms. 16 . A computer program product comprising a computer program, the computer program comprising program code instructions executable by a computer, the program code instructions implementing the method according to claim 1 when the computer program is executed on a computer.

17. The computer program product according to claim 16, wherein: The computer program product is a data carrier or a storage medium.

Images