Method for handling alarms in a processing system

By constructing a subgraph of multi-node directed graph and using time series analysis, combining artificial neural networks and machine learning, the problem that operators find it difficult to quickly understand the state in the factory is solved, and fast data access and efficient alarm processing are achieved, simplifying Anomaly investigation and maintenance process were carried out.

CN113534752BActive Publication Date: 2024-11-26ABB (SCHWEIZ) AG
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202110403105.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2021-03-31
Filing Date
2021-04-14
Publication Date
2024-11-26
Estimated Expiration
2041-04-14

AI Technical Summary

Technical Problem

In industrial plants or processing systems, it is difficult for operators to quickly understand the status of the plant or production process when handling alarms or critical stages, especially when data and topology are complex, resulting in information spread across multiple screens and data sources , increasing the complexity of analysis and the risk of error.

Method used

By constructing a subgraph of a multi-node directed graph, selecting the starting node and outputting the node and its properties, combining time series and event analysis, using artificial neural networks and machine learning methods to highlight related data points, generate dynamic process graphs and interactive methods User interface, providing fast data access and analysis.

Benefits of technology

A quick and clear overview of the factory or production process is achieved, reducing maintenance time and cost, improving the quality and efficiency of alarm handling, and simplifying abnormality investigation and root cause analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113534752B_ABST
    Figure CN113534752B_ABST
Patent Text Reader

Abstract

The present invention is entitled "Method for Alarm Handling in a Processing System". The present invention relates to the field of alarm handling or the field for providing attributes of elements in an industrial plant or a processing system, in particular for a processing system that can be represented as a graph. To this end, the processing system is represented as a directed graph comprising a plurality of nodes and directed edges, wherein each node represents an element, each node has attributes, and each directed edge represents a relationship between two elements of the plurality of elements. The method comprises the following steps: selecting a node from a plurality of nodes as a starting node; constructing a subgraph, wherein the subgraph consists of all nodes forward-connected from at least one directed edge of the starting node; and outputting all nodes and attributes of the nodes of the subgraph.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of alarm handling in industrial plants or process systems, in particular for process systems that can be represented as graphs. The invention also relates to a computer program product, a computer readable storage medium and the use of the method. Background Art

[0002] Plant operations personnel, such as operators, process engineers, automation engineers or dispatchers often need to get an overview of the status of a plant or production process, especially in the event of an alarm or other critical or interesting phase during plant operation. In some cases, it may be useful to gain a better understanding of the data of the process system and / or the topology of the plant and / or the connectivity of the process elements of the plant. Summary of the invention

[0003] Therefore, the object of the present invention is to provide a method for obtaining the state or another property of at least some of the elements of interest in a plant or a production process. This object is achieved by the subject matter of the independent claims. Further embodiments are clear from the dependent patent claims and the following description.

[0004] One aspect relates to a method for providing properties of elements in a processing system including a plurality of elements, wherein the processing system is represented as a directed graph including a plurality of nodes and directed edges, and wherein each node represents an element, each node has properties, and each directed edge represents a relationship between two elements of the plurality of elements. The method comprises the following steps:

[0005] Select a node from multiple nodes as a starting node;

[0006] constructing a subgraph, wherein the subgraph consists of all nodes that have at least one directed edge forward connected from the start node; and

[0007] Output all nodes and node attributes of the subgraph.

[0008] Elements or process elements in a process system, in particular an industrial process system, or elements or process elements thereof may include, for example, pumps, compressors, containers or pressure vessels, tanks, heat exchangers, furnaces, fans, cooling towers, valves, etc.

[0009] The attributes (representation elements) of a node may include the "state" of the node, such as process values, like temperature, pressure, etc., and / or composite or derived values, such as "safe, serious, dangerous, damaged..."; this may include a classification of (one or more) attributes. Each of the nodes includes or has attributes. Attributes may be constant or empty. Attributes may include one attribute, multiple attributes, and / or a list of attributes.

[0010] Selecting a node as a starting node may include selecting one or more arbitrary nodes, the selection may be triggered or inspired by an alarm, or the selection may be made to obtain a piece of information to understand a plant or production process. Constructing a subgraph may include following or continuing forward along directed edges from a starting node to an ending node or to a node that has been touched (e.g., in a cycle).

[0011] Outputting all nodes and properties of the nodes of the subgraph may include showing or transmitting a list of these nodes and their properties, and / or performing some post-processing, such as showing the subgraph as a topology graph with colored nodes, the color depending, for example, on their current criticality, and / or showing the topology graph as an interactive diagram, for example with "clickable nodes" that show a list of properties on demand.

[0012] This may help provide a quick and clear overview of the current state of a plant or production process. This may help to quickly enter the assessment process and / or get quick access to process data, for example, by visualizing and interactively displaying relevant attributes and / or additional data that may be related to the current state. For example, by analyzing which process parts need to be shut down or can be operated within safe parameters, this may help to speed up maintenance and / or respond quickly in alarm situations. Other examples may include conducting efficiency surveys, for example, how to improve the efficiency of heat exchangers, event handling, such as root cause analysis of failures, defects or emergencies of compressors in compressor stations, and / or other actions or uses. For example, this may be useful when a plant operator wants to browse the life data of the plant, such as analyzing the current state relative to a given or expected state. This may further help to understand the process topology of a well-directed process. In addition, time and cost can be reduced for tasks such as manual anomaly investigations, complex and multi-device equipment failure mitigation, and / or root cause analysis.

[0013] In various embodiments, the method further comprises the step of outputting a subset of the process values ​​of the directed graph, wherein at least one attribute of at least one node of the directed graph is a process value. Outputting may include, for example, displaying on an alarm screen, an analysis module, and / or other actions, visually transmitting, post-processing the attribute(s).

[0014] In various embodiments, the selected node has attributes related to the alarm.

[0015] "Related" may include having been triggered or excited by an alarm. An example may be a pressure increase that has become critical, and the method shows which elements the node affects, for example, for an operator. Or, for example, if a temperature increase (i.e., a property) of a container is indicated as an alarm trigger, the subgraph starts with the process element container. Alarms may include composite or derived values ​​such as classifications, for example, of pressure or temperature, such as "safe / critical / dangerous / damaged".

[0016] In various embodiments, alarms are provided by analyzing time series of process values. For example, an alarm trigger may not be a current state, but may depend on rapid changes in temperature, pressure, etc. This may be performed by modules that use one or more time series and event analysis to highlight particularly relevant equipment and process values ​​in dynamically generated process graphics.

[0017] In some embodiments, the analysis is performed by an artificial neural network ANN. The ANN can be trained by providing a known time series, such as a time series associated with an alarm or already associated with an alarm. This can advantageously help to more easily distinguish between a critical increase in temperature (etc.) and an increase in non-critical temperature (etc.). Training (and / or application of trained ANN) may include the use of time series and event analysis, which, for example, uses standard anomaly detection methods (local outlier factors, a class of SVMs, autoencoders (AutoEncoder) and / or others) to identify process values ​​and events that should be displayed and / or highlighted to the operator. Training may include analysis of historical persistent disturbances (particularly oscillatory disturbances). A further method, such as for obtaining more insights into causality and timing, is to analyze the cross-correlation and cross-entropy functions of changes around setpoint changes in order to find out the intensity of the propagation of the change and the time it takes to propagate. Cross-correlation may exist between an increase in pressure in an element and a decrease in pressure in an element (as an example). Therefore, machine learning by analyzing time series and event data may help highlight interesting and relevant data points to operators, for example, particularly relevant equipment and process values ​​in dynamically generated process graphics.

[0018] In some embodiments, the time series of process values ​​includes historical data. Using historical data may provide changes to take into account the development of key data, so that a piece of the "future" is considered when applied to the current situation or process value. In particular, the analysis of historical persistent disturbances (especially oscillatory disturbances) by means of, for example, cross-correlation analysis or cross-entropy can improve the quality of alarm processing.

[0019] In various embodiments, the method further comprises the step of outputting a process state of the processing system, wherein the process state comprises at least one of: a subgraph, a process value of a directed graph and / or a time series of process values.

[0020] The process state may be an indicator of the state of the entire process system or a subsystem (e.g., a subsystem defined by a submap). The process state is generated, may be displayed and / or communicated. This may include process equipment, process values, alarms and events relevant to the current analysis task. Process topology information may be derived from the process flow diagram and made available for use during plant operation, as described above and / or below.

[0021] There are scenarios where information is spread across many screens and different data sources, such as process history, (one or more) process flow charts, and / or control system descriptions. The operator may then need to switch between different screens showing different parts of the process in order to collect all relevant information. This switching may make the task of alarm analysis more challenging because the operator needs to remember all relevant information when switching between screens. In such scenarios, the collection of information may become time-consuming, error-prone, and cumbersome for operators. The embodiments described herein may use computer-readable process topology information and process control system information, traverse the topology to identify relevant subgraphs, use machine learning methods, such as anomaly detection for time series and event analysis, to highlight relevant process values ​​and events, or filter out irrelevant process values ​​and events, and may generate task-specific graphical representations of subgraphs with embedded trend and event visualizations and / or may generate task-specific user interfaces, such as operator screens. The embodiments may also be suitable for small screens. The embodiments may provide an API for accessing topology information together with real-time process values ​​and process control system device properties. The information provided may be further used for root cause analysis of alarms and / or emergency situations.

[0022] In some embodiments, the process system is represented by a process flow diagram (PFD) and / or a piping and instrumentation diagram (P&ID). The PFD (etc.) can be used as the basis for a directed graph including a plurality of nodes and directed edges. It may also be useful to have standardized exchange formats for smart P&IDs. Advantageously, this provides a common format for describing a process or plant.

[0023] One aspect relates to a computer program product comprising instructions which, when executed by a computer, cause the computer to perform the method described above and / or below.

[0024] One aspect relates to a computer-readable storage medium on which is stored a computer program or a computer program product as described above.

[0025] One aspect relates to the use of a method as described above and / or below for efficiency surveys of plants or subsystems, for access, in particular rapid access, to process data, and / or for event handling of elements. This may help to speed up maintenance and / or react quickly in alarm situations.

[0026] One aspect relates to an industrial plant or a processing system, comprising a computer on which instructions are stored, which, when the computer executes the program, cause the computer or the industrial plant to perform the method described above and / or below.

[0027] For further clarification, embodiments of the present invention are described with the aid of the embodiments shown in the accompanying drawings. These embodiments are to be considered as examples only and not restrictive. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Schematically illustrates an example of identifying problems by monitoring process values ​​and alarms in a process control system according to an embodiment;

[0029] Figure 2 A flow chart of a method according to an embodiment is schematically shown.

[0030] Figure 3 A flow chart of a further method according to an embodiment is schematically shown. DETAILED DESCRIPTION

[0031] Figure 1 An example of identifying a problem by monitoring process values ​​and alarms in a process control system 100 according to an embodiment is schematically shown. The automated monitoring system 110 (or an operator) can identify a problem, for example, by monitoring process values ​​and alarms in the process control system 100. This sends out an agent 120 to find some context and / or additional information related to the problem, such as finding relevant process values ​​and process devices or elements. The agent 120 performs a query on the module 140 to obtain the process topology and / or additional information. The query can be specified based on the connectivity information in the process flow, all meta information from the PFD or intelligent P&ID diagram (e.g., media type, pipe diameter, alarm level, etc.), and attributes derived from the process control system (e.g., device attributes). Then, a node in the directed graph related to the topology description is selected as the starting node. Thus, the directed graph is traversed to construct a subgraph 150, thereby finding relevant process values ​​and device labels, i.e., nodes and their attributes.

[0032] The list of identified process values ​​and process tags may be passed to an (optional) time series and event analysis module 160 that uses standard anomaly detection methods, such as local outlier factors, one-class SVMs, autoencoders, to identify process values ​​and events that should be highlighted to an operator. The time series and event analysis may also perform correlation and / or cross-correlation analysis using standard methods to identify process values ​​that may be causally related to the identified atypical process values ​​or events, such as by combining information about cross-correlations and topological relationships. The results of modules 140 and / or 160 are passed to a display, visualization, or task-specific user interface 160 to provide rapid, and in some embodiments interactive, analysis and / or assessment of alarms or conditions of interest.

[0033] Figure 2 A flow chart 200 of a method according to an embodiment is schematically shown. In step 201, a trigger is sent by the system and / or an operator. In step 202, a query is executed for which a directed graph related to the system topology is traversed in order to construct a subgraph. In optional step 203, time series and event analysis is performed. This is to highlight abnormal and / or relevant values ​​and / or filter out normal and / or irrelevant values, thereby advantageously reducing the information that can be provided to the operator. In step 204, a task-specific process and / or user interface is generated, which reflects relevant trends, events and alarms.

[0034] Figure 3 A flowchart 300 of a method for providing attributes of an element in a processing system including a plurality of elements is schematically shown, wherein the processing system is represented as a directed graph including a plurality of nodes and directed edges, wherein each node represents an element, each node has attributes, and each directed edge represents a relationship between two elements in the plurality of elements. In step 301, a node in the plurality of nodes is selected as a starting node. In step 302, a subgraph is constructed, wherein the subgraph consists of all nodes forward connected from at least one directed edge of the starting node. In step 303, all nodes and attributes of the nodes are generated for the subgraph output, i.e., a display, visualization and / or a task-specific user interface.

Claims

1. A computer-implemented method for providing properties of an element in an industrial process system comprising a plurality of elements, in, The processing system is represented as a directed graph based on a process flow diagram and / or a piping and instrumentation diagram, the directed graph comprising a plurality of nodes and directed edges, Wherein each node represents an element, each node has attributes, and each directed edge represents a relationship between two elements of the plurality of elements, the method comprising the following steps: selecting one of the plurality of nodes as a starting node, wherein the selected node has an attribute related to the alarm; constructing a subgraph, wherein the subgraph consists of all the nodes forward-connected from the starting node by at least one directed edge; and All nodes of the subgraph and the attributes of the nodes are output.

2. The method according to claim 1, further comprising: The following steps are involved: outputting a subset of the process values ​​of the directed graph, At least one attribute of at least one node of the directed graph is a process value.

3. The method according to claim 1, in, The alarm is provided by analyzing the time series of the process values.

4. The method according to claim 3, in, The analysis is performed by means of an artificial neural network.

5. The method according to claim 3 or 4, in, The time series of process values ​​comprises historical data.

6. The method according to claim 1, further comprising: The following steps are involved: outputting a process status of the processing system, The process state includes at least one of the following: the subgraph, a process value of the directed graph and / or a time series of process values.

7. A computer program product comprising a computer program having instructions which, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 6.

8. A computer-readable storage medium having stored thereon a computer program with instructions, which, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 6.

9. Use of the method according to any one of claims 1 to 6 for efficiency surveys of plants or subsystems, for access to process data and / or for event handling of elements.

10. An industrial plant or processing system comprising a computer and a storage device having stored thereon a computer program with instructions, and when the computer program is executed by the computer, the instructions cause the computer to perform the method according to any one of claims 1 to 6.

Citation Information

Patent Citations

  • Business topology information display method and system of cloud monitoring system

    CN106452927A

  • Prediction model testing framework

    US10372572B1

  • Enabling network intrusion detection by representing network activity in graphical form utilizing distributed data sensors to detect and transmit activity data

    US20070209075A1

  • Method and Device for Supporting Fault Diagnosis of an Industrial Installation

    US20130235041A1

  • System and method for custom graph generation

    US20170310552A1