Artificial intelligence model for analyzing process alarms and generating corresponding rationalized actions

Abstract

The disclosure relates to an artificial intelligence model that analyzes process alarms and generates corresponding rationalized actions. An alarm system within a rationalized industrial plant includes an alarm system computer in communication with one or more alarm system databases. The alarm system computer is configured to execute the machine-learned model to analyze at least one of the alarm data and the process data received from the alarm system database to output a current state of the alarm system within the industrial plant. A predetermined alarm system principle of the plant is provided as input, and the machine-learned model identifies and performs one or more rationalized actions for the alarm system based at least on the current state and the predetermined alarm system principle to output a rationalized future state of the alarm system. At least one of the one or more rationalized actions includes modifying an alarm priority within the alarm system based at least on an urgency multiplier and a severity score or consequence value.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. Provisional Patent Application Serial No. 63 / 597,759 filed on November 10, 2023 and U.S. Non-Provisional Patent Application Serial No. 18 / 606,797 filed on March 15, 2024. Technical Field

[0003] The present disclosure generally relates to systems and methods for rationalizing alarm systems within industrial plants. Background Art

[0004] Alarm systems are used in industrial plants to inform plant operators of the status of operating conditions in the industrial plant. In addition, alarm systems are used to signal plant operators when process intervention is required in the industrial plant to correct or prevent abnormal or unsafe operating conditions.

[0005] Plant personnel can perform an alarm rationalization process to improve the effectiveness of alarm systems in industrial plants. Generally speaking, many processes are involved in alarm rationalization, which require a lot of time and resources to perform, but the alarm rationalization process is crucial for improving alarm systems. In particular, some of the main goals of the alarm rationalization process are to determine the alarm priority, alarm set point, root cause, and corrective measures for process alarms in the alarm system. The traditional alarm rationalization process can involve assigning priority classifications (e.g., emergency, high, medium, and low) to each alarm based only on the consequences associated with the uncleared alarms. Using such methods, the traditional alarm rationalization process usually provides atomic, highly variable, and inaccurate results, which is suboptimal for improving the effectiveness of the alarm system. Thus, due to the variations and inaccuracies in the traditional alarm rationalization process, the results of the process are often false. When performing tasks such as determining the priority classification of alarms, false statements in alarm rationalization can be devastating. For example, false priority classifications of alarms can cause deviations in the alarm priorities of the alarm system. In such circumstances, plant operators are more likely to mishandle the order in which the alarms are activated, which can lead to process upsets, plant downtime, or even accidents that hamper the performance and safety of the industrial plant.

[0006] Therefore, there is a need for an improved framework for an alarm rationalization process, wherein such a process is optimized for configuring and managing alarms within an alarm system of an industrial plant to improve the overall effectiveness of the alarm system. Summary of the invention

[0007] Aspects of the present disclosure allow for deployment of an artificial intelligence based alarm rationalization system for rationalizing alarm systems within an industrial plant.

[0008] One aspect of the present disclosure includes a computer-implemented method for rationalizing an alarm system within an industrial plant, comprising providing an alarm system input through at least one or more alarm system databases, the alarm system input including at least one of a training input and an operational input regarding the alarm system. A predetermined alarm system principle for the industrial plant is provided as an input. A processor executes a machine learning model for analyzing the alarm system input to output a current state of the alarm system. One or more rationalization actions for the alarm system are identified and performed based on the current state and at least one of the predetermined alarm system principles to output a rationalized future state of the alarm system. At least one of the rationalization actions includes modifying an alarm priority within the alarm system based on at least an urgency multiplier, a consequence value, and a severity score.

[0009] In another aspect, a computer-implemented method for rationalizing an alarm system within an industrial plant includes providing an alarm system input through an alarm system database, the alarm system input including at least one of an operational input and a training input for developing an alarm system within the industrial plant. A predetermined alarm system philosophy for the industrial plant is provided as an input. A processor executes a machine learning model for identifying and performing one or more rationalization actions based on the alarm system input and the predetermined alarm system philosophy to output a rationalized future state of the alarm system including one or more alarms. At least one of the one or more rationalization actions includes creating an alarm priority for the alarm system based on a consequence value, a severity score, and an urgency multiplier.

[0010] In another aspect, an alarm rationalization system for optimizing an alarm system within an industrial plant includes one or more alarm system databases configured to store at least one of alarms and process data. A predetermined alarm system principle is configured to manage alarm system decisions within the industrial plant. A processor is configured to execute a machine-learned model. When executing the machine-learned model, the processor is configured to receive at least one of the alarm data and the process data as input to analyze and provide a current state of the alarm system within the industrial plant as output. The processor is configured to receive the predetermined alarm system principle as input, and when executing the machine-learned model, the processor is further configured to perform one or more rationalization actions on the alarm system based on the current state and the predetermined alarm system principle to provide a rationalized future state of the alarm system as output. A controller in communication with the processor is configured to receive an alarm setting change based on the rationalized future state from the processor as input, and automatically change the alarm setting of the alarm within the alarm system. At least one of the rationalization actions includes modifying the alarm priority within the alarm system based on a consequence value or a severity score and an urgency multiplier. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1is a graphical representation of clearing times for alarms according to an embodiment.

[0012] Figure 2 is an illustration of an example of determining alarm priority based on clearing time according to an embodiment.

[0013] Figure 3 is a schematic block diagram of an alarm rationalization system within an industrial plant according to an embodiment.

[0014] Figure 4 is a schematic diagram illustrating a machine-learned model of an alarm rationalization system according to an embodiment.

[0015] Figure 5 is a schematic diagram illustrating the workflow of an alarm rationalization system according to an embodiment.

[0016] Figure 6 is an illustration of operational inputs using a machine-learned model in an industrial plant, according to an embodiment.

[0017] Fig. 7A is an illustration of an example of an alarm severity score or consequence value table, according to an embodiment.

[0018] Figure 7B is a diagram of an example of an alarm breakpoint table according to an embodiment.

[0019] Figure 7C is an illustration of an example of an alarm urgency assignment table, according to an embodiment.

[0020] Fig.7D is a diagram of an example of an alarm priority table according to an embodiment.

[0021] Figure 8 is a diagram of an alarm rationalization process according to an embodiment.

[0022] Corresponding reference numerals indicate corresponding parts throughout the drawings. DETAILED DESCRIPTION

[0023] Alarm management within an industrial plant involves determining best practices and tools to improve the effectiveness of alarm systems within the plant so that the operational performance and safety of the plant are enhanced. One aspect of alarm management involves alarm rationalization. Alarm rationalization includes a series of alarm rationalization processes to be performed to improve the effectiveness of the alarm system. One rationalization process includes determining alarm priorities, in which an order is determined for resolving activated alarms. Thus, alarm prioritization provides a strategy that enables plant operators to effectively focus their efforts on resolving activated alarms to clear alarms before process upsets and hazards occur.

[0024] One difficulty with conventional alarm rationalization processes is determining accurate alarm priority. Conventional processes for determining alarm priority are primarily based on the consequences associated with an alarm not being cleared. However, by considering the consequences or severity of an alarm not being cleared within the alarm system and the urgency of the alarm, a more accurate alarm priority can be obtained. Urgency can be expressed if the time or view, understand, decide, and act (SUDA) time required to clear an alarm is known. Figure 1 An example of a breakdown of the SUDA time for an alarm is provided, starting with a fault occurring in a process in an industrial plant. From here, the alarm has an internal diagnostic time, which includes the time it takes for the alarm system to detect the fault and notify the plant operator about the detected fault. Next is the time it takes for the corrective action to be performed within the industrial plant, and the time it takes for the process to react to the corrective action and become stable or activate an alarm trip. Figure 2 An example of using only the estimated SUDA time to determine the alarm priority of an alarm system is provided, without considering the consequences or severity of the uncleared alarms. The alarm rationalization system and method according to the present disclosure considers consequences, severity, and urgency to determine a more accurate alarm priority, as will be described in further detail below.

[0025] Aspects of the present disclosure improve the accuracy and efficiency of determining alarm priorities for an alarm system at least in part by providing an improved system and method for rationalizing an alarm system within an industrial plant. As will be explained in further detail below, the present disclosure provides an alarm rationalization system and method for optimizing an alarm system within an industrial plant. In addition, the alarm rationalization system and method disclosed herein are configured to continuously analyze and update the alarm priorities of the alarm system, recommend effective alarm setting changes to the plant operator, recommend corrective actions required to clear activated alarms to the plant operator, add or remove alarms within the alarm system, and recommend special alarm treatments to the plant operator.

[0026] Reference now Figure 3, an exemplary embodiment of an alarm rationalization system according to the present disclosure is generally indicated by reference numeral 10. The system 10 broadly includes a plurality of alarms 101 within an industrial plant 111, an alarm system database 201 including at least an alarm database 102 and a plant database 105, and one or more alarm system computers 106 that execute processor executable instructions from a machine-learned model 103 based at least on a predetermined alarm system principle 104. The alarms 101 reside within the alarm database 102 and are configured to notify plant operators of operating conditions within the industrial plant 111. As will be described below, the alarm system computer 106 is operably connected to the alarm system database 201, and the alarm system computer is configured to execute the machine-learned model 103 for determining a rationalized future state of the alarm system within the industrial plant 111. Before turning to an exemplary method of training the machine-learned model 103 and using the alarm rationalization system 10 within the industrial plant 111, the various components of the alarm rationalization system 10 will now be described in more detail.

[0027] Reference now Figure 4 , an exemplary embodiment of an alarm rationalization system 10 is shown, wherein an alarm system database 201 includes at least an alarm database 102 and a plant database 105. The alarm database 102 is configured to store configured alarms. In the illustrated embodiment, the plant database 105 includes a historical database, and the alarm database 102 includes a master alarm database (hereinafter referred to as a "MADB"). The historical database 105 is configured to store historical operating data from the industrial plant 111, such as historical alarm data 202 (e.g., alarm history), historical operator actions 203 (e.g., operator action log data), and historical process data 204 (e.g., analog value trends). The MADB 102 is configured to store configured alarm data, such as a list of alarms, alarm groups, alarm types, alarm priorities, alarm classes, alarm activation points, corrective actions, and consequences of alarm activation. In an exemplary embodiment, the alarm system database 201 is also configured to store data related to existing alarm key performance indicators 205, control loop details 206, existing alarm rationalization efforts 207, and customer performance guidelines 208.

[0028] In an exemplary embodiment, the alarm system database 201 is also configured to store predetermined alarm system principles 104 for the industrial plant 111. The alarm system principles 104 include detailed guidelines for alarm systems and alarm rationalization. For example, the alarm system principles 104 include a list of rationalization rules that are consistent with performance goals such as KPIs of the industrial plant 111, and these rules are derived from plant management and alarm system standards (such as ISA 18.2). The alarm system principles 104 are configured to define alarm rationalization methods for managing alarm system decisions and performing rationalization actions within the industrial plant 111. For example, the alarm system principles 104 include change management procedures (hereinafter referred to as "MOC") for performing rationalization actions identified by the machine-learned model 103 within the industrial plant 111.

[0029] In one embodiment, the alarm system principle 104 includes an alarm severity score or consequence value table 301, an alarm breakpoint table 302, and an alarm urgency assignment table 303. The alarm severity score table 301 includes consequences or severity impacts, such as safety, environmental and economic impacts, severity classifications of severity impacts, such as none, low, medium, high and urgent, and severity scores for each severity impact and severity classification combination. The alarm breakpoint table 302 includes priority classifications, such as urgent, high, medium, low and no alarm. In addition, the alarm breakpoint table 302 includes breakpoint value thresholds for each priority classification. The alarm urgency assignment table 303 includes multiple time thresholds defining the available time for clearing the alarm, urgency classifications associated with each time threshold (such as urgent, high, medium and low), and urgency multipliers associated with each urgency classification.

[0030] In general, the alarm system computer 106 is configured to execute the machine-learned model 103 to rationalize the alarm system within the industrial plant 111 based on data obtained from the alarm system database 201. The alarm system computer 106 includes a processor, a memory, a user input, a display, and other related elements. The alarm system computer 106 may also include a circuit board and / or other electronic components, such as a transceiver or external connection for communicating with other computing devices of the system 10. For example, the alarm system computer 106 includes components such as a wireless transceiver and / or a wired connector that connects the alarm system computer to the alarm system database 201. Thus, the memory of the alarm system computer 106 stores process executable instructions such as the machine-learned model 103, and the processor is configured to execute the instructions to generate a rationalized future state of the alarm system within the industrial plant 111. In one embodiment, the processor is configured to populate at least one of the alarm system databases 201 with data about the rationalized future state. The plant operator can view the rationalized future state data via the display. In the exemplary embodiment, the alarm rationalization system 10 also includes one or more controllers operably connected to the alarm system computer 106, and the controllers are configured to automatically perform actions within the industrial plant 111. For example, the controllers are configured to receive inputs from the alarm system computer 106, such as alarm setting changes derived from the rationalized future state data, and automatically change the settings of the alarms 101 within the industrial plant 111 based on the inputs.

[0031] The machine-learned model 103 is configured to obtain data from the alarm system database 201 as alarm system inputs, including at least one of operational inputs and training inputs, to determine or output a current state of an alarm system within the industrial plant 111 . Figure 4 Examples of data used as training input to train the machine-learned model 103 are provided, and Figure 6 Examples of data used as operational input to use the machine-learned model in operation are provided. In addition, the machine-learned model 103 is configured to analyze the operational input, the training input, and the determined current state of the industrial plant 111. For example, the machine-learned model 103 is configured to perform at least one of the following: estimating an average rate of change of disturbances, evaluating alarm performance, identifying an active alarm priority distribution, and identifying a configured priority distribution based on at least a predefined or target distribution 304.

[0032] The machine-learned model 103 is also configured to identify and execute one or more rationalization actions for the alarm system based on at least the current state and the alarm system principles 104 to determine a rationalized future state of the alarm system. In an exemplary embodiment, the machine-learned model 103 is also configured to evaluate a change management evaluation for each rationalization action to determine which rationalization action to evaluate based on the MOC procedure. In one example, the rationalization actions include modifying alarm priority, disabling alarms, enabling alarms, modifying alarm logic, validating alarm logic, and not performing any changes to individual alarms within the alarm system.

[0033] The machine-learned model 103 is configured to modify the alarm priority of the alarm system by weighting the reaction time (which includes the available time to clear the alarm) with other factors that affect the alarm priority (such as alarm severity, alarm consequence, and alarm urgency) to determine the alarm priority of the alarm system. Thus, embodiments according to the present disclosure provide more comprehensive and accurate alarm priorities for the alarm system than conventional processes for determining alarm priorities because more factors that affect alarm priority are taken into account.

[0034] Figures 7A-7D An example of modifying alarm priority based on at least a consequence value or severity score and an urgency multiplier is provided. Each alarm 101 in the alarm system includes a unique identification tag. Each alarm 101 is assigned a severity score or consequence value for each severity impact determined from the alarm severity score table 301. In one embodiment, the severity scores of each alarm are added to determine the breakpoint value of each alarm. Moreover, each alarm has a reaction time. The reaction time of each alarm is compared with a plurality of time thresholds from the alarm urgency assignment table 303 to determine the urgency multiplier. The breakpoint value is multiplied by the urgency multiplier to obtain a modified breakpoint value taking into account urgency. The modified breakpoint value is compared with the breakpoint threshold of the alarm breakpoint table 302 to determine the priority classification of the alarm 101. The priority classification of each alarm 101 in the alarm system is used to determine the alarm priority of a rationalized future state. The machine-learned model 103 is configured to fill the MADB 102 with data related to the alarm priority of a rationalized future state.

[0035] In one embodiment, the enable and disable alarm rationalization actions are based on the protocol outlined by the MOC procedure. If the disabled alarm belongs to the category of enabled alarms, the machine-learned model 103 will assign an enable alarm action. In the event that the current state is analyzed and it is determined that the current state does not have a sufficient number of alarms, the enable alarm rationalization action is performed to add alarms to the alarm system based on the MOC procedure. Conversely, when the current state is analyzed and it is determined that it has useless alarms, the disable alarm rationalization action is performed to remove alarms from the alarm system based on the MOC procedure.

[0036] The machine-learned model 103 is configured to add and / or modify the high-level alarm logic of the alarm system by correlating related alarms and events that are activated within the same time frame. Thus, compared to conventional processes for determining when an alarm should be activated, embodiments according to the present disclosure provide more comprehensive and accurate alarm activation for alarms within an alarm system. As an example, the machine-learned model 103 is configured to monitor the activated alarm history 202 and the operator action log 203 to identify that if a unit fails, the alarm is always activated. The machine-learned model 103 will recommend suppressing the alarm as long as the unit fails.

[0037] Furthermore, the machine-learned model 103 is configured to validate the high-level alarm logic of the alarm system by correlating related alarms and events that are activated within the same time frame. Thus, compared to conventional processes for determining when an alarm should be activated, embodiments according to the present disclosure provide more comprehensive and accurate alarm activation for alarms within an alarm system. As an example, the machine-learned model 103 is configured to monitor the activated alarm history 202 and operator action log 203 to identify that if a unit fails, then the alarm is always activated. The machine-learned model 103 will validate the alarm suppression logic and ensure that important or critical alarms are not missed due to suppression.

[0038] In an exemplary embodiment, the machine-learned model 103 is also configured to determine a rationalization effect of performing the rationalization action. In one example, the current state is compared to a rationalized future state to determine a rationalization effect, wherein the rationalization effect defines an increment between the current state and the rationalized future state. In another example, the rationalized future state is compared to a target state provided as an input to determine a rationalization effect of the performed rationalization action, wherein the rationalization effect indicates an increment between the rationalized future state and the target state. For example, the target state may include a target distribution of alarm priorities based on the ISA18.2 alarm system standard. The machine-learned model 103 is configured to evaluate a distribution of alarm priorities for the rationalized future state and compare the distribution to a predefined or target distribution 304 to determine a rationalization effect of modifying the alarm priorities of the alarm system.

[0039] The machine-learned model 103 is also configured to determine whether to re-execute the rationalization action or propose a rationalized future state based on the rationalization effect. In one example, if the rationalization effect meets or exceeds the expected rationalization effect, the machine-learned model 103 determines to propose a rationalized future state, and if the rationalization effect does not meet or exceed the expected rationalization effect, the machine-learned model determines to re-execute the rationalization action. When re-executing the rationalization action, the machine-learned model 103 is configured to execute at least one of a rationalization action that has not yet been executed, a modification of a rationalization action that has been executed, and the same rationalization action. In one example, the machine-learned model 103 is configured to define urgency classifications for multiple time thresholds and define urgency multipliers for each urgency classification to reconstruct the alarm urgency assignment table 303. In another example, the machine-learned model 103 is configured to define severity classifications for multiple severity impacts and define severity scores for each severity impact to reconstruct the alarm severity score table 301. Furthermore, the machine-learned model 103 is configured to define priority classifications for a plurality of breakpoint thresholds to reconstruct the alarm breakpoint table 302 .

[0040] The machine-learned model 103 is also configured to generate corrective actions for clearing activated alarms within the alarm system based on the rationalized future state, and communicate the corrective actions to the plant operator via the display. Moreover, the machine-learned model 103 is configured to generate alarm settings for alarms 101 within the alarm system based on the rationalized future state, and communicate the alarm settings to the plant operator via the display. Furthermore, in both cases, the machine-learned model 103 can populate the MADB 102 with the generated corrective actions and the generated alarm settings.

[0041] Reference now Figure 5 , shows an exemplary embodiment of a periodic workflow of an alarm rationalization system 10 within an industrial plant 111. In particular, the workflow includes steps defined by a machine-learned model 103 for execution by an alarm system computer 106. In step 1, the current state is determined based on input from the alarm system database 201. For example, input regarding historical data (such as alarms and process data) is received from the plant database 105. The average rate of change of disturbances of the alarm variables is determined. The current performance of the alarms is evaluated, and the active alarm priority distribution is identified. In one example, the alarm performance is found by multiplying the alarm frequency (top 10) by the repeat alarms (alarms occurring at the same time), and dividing by the continuous alarms, jittering alarms, and alarm priority distribution.

[0042] In step 2, constraints are identified. For example, alarm system inputs (such as data from alarm database 102 and control loop details 206) are used to identify a configured priority distribution, establish alarm priorities based on consequences modified by reaction time, and identify rationalization actions based on alarm classification.

[0043] In step 3, rationalization actions are identified based on the alarm system inputs, such as disabling the alarm, enabling the alarm, modifying the alarm priority, modifying the alarm group, modifying the alarm logic (e.g., excluding the distributed control system), verifying the alarm logic (e.g., verifying the emergency shutdown (ESD) logic for opening / closing), rationalizing the alarm thresholds, and not performing any changes.

[0044] In step 4, rationalization actions and change management procedures are performed.

[0045] At step 5, the alarm performance is re-evaluated. For example, the priority distribution is compared to the target distribution 304 and operational feedback is considered when re-evaluating the alarm performance. A determination is made as to whether the performance is adequate or inadequate.

[0046] At step 6, if the performance is insufficient, new alarm settings are proposed and input back into the workflow loop along with data from the marked master alarm database (MADB). In an exemplary embodiment, the alarm settings are restructured in an understandable format.

[0047] Computer-implemented methods of training and using a machine-learned model 103 in an industrial plant 111 will now be described. In the first method described below, the machine-learned model 103 is trained to rationalize an alarm system with pre-existing alarms using a start-from-scratch approach. In the second method described, the machine-learned model 103 is trained to develop a rationalized alarm system using a start-from-scratch approach.

[0048] First, in a method that starts from the prior art, an alarm system input (including at least one of a training input and an operational input, and a predetermined alarm system principle) from at least one of the alarm system databases 201 is provided as input to the alarm system computer 106. The processor executes the machine-learned model 103, and the machine-learned model obtains the alarm system input for the alarm system within the industrial plant 111. Next, the machine-learned model 103 analyzes the alarm system input to analyze and provide as output the current state of the alarm system within the industrial plant 111. For example, this may include at least one of: evaluating the performance of the current state of the alarm system, estimating the average rate of change of alarm interference, identifying an active alarm priority distribution, and identifying a configured priority distribution.

[0049] Next, the machine-learned model 103 identifies one or more rationalization actions for the alarm system based on at least one of the alarm system input, the current state, the alarm key performance indicators, and the alarm system principle 104 to perform the action to obtain a rationalized future state of the alarm system. Suitably, the machine-learned model 103 performs change management procedures and rationalization actions. For example, the machine-learned model 103 modifies the alarm priority of the alarm system based on at least the consequence value or the severity score and the urgency. Accordingly, the machine-learned model 103 compares the rationalized future state with the target state to determine the rationalization effect of performing the one or more rationalization actions. At this point, the machine-learned model 103 determines whether to re-execute the method or propose a rationalized future state based on the rationalization effect. In one example, if the rationalization effect meets or exceeds the expected rationalization effect, then the machine-learned model 103 determines to propose a rationalized future state, and if the rationalization effect does not meet or exceed the expected rationalization effect, then the machine-learned model determines to re-execute the method.

[0050] In the event that the machine-learned model 103 determines to re-execute the method, the machine-learned model 103 obtains further alarm system inputs to determine the current state and performs further rationalization actions based on at least the alarm system inputs, the current state, and the alarm system principle 104. In addition, the machine-learned model 103 identifies rationalization actions that have not been previously performed. For example, the machine-learned model 103 is configured to modify alarm setting values, such as consequence values, urgency multipliers, and time thresholds when performing further rationalization actions. In an exemplary embodiment, the machine-learned model 103 repeats this process until the desired rationalization effect is met or exceeded, which can include plant KPIs. In the event that the machine-learned model 103 determines to propose a rationalized future state, the machine-learned model generates alarm settings and corrective actions based on the rationalized future state for clearing activated alarms within the industrial plant 111. The machine-learned model 103 populates the MADB 102 with data about the rationalized future state. The rationalized future state is presented to the plant operator via a display of the alarm system computer 106.

[0051] In the method from scratch, the method similarly starts with at least one of an alarm system input from at least one of the alarm system databases 201 and a predetermined alarm system principle being provided as input to the alarm system computer 106. The processor executes the machine-learned model 103, and the machine-learned model 103 obtains the alarm system input from the alarm system database 201 for developing an alarm system including a plurality of alarms. Next, the machine-learned model 103 identifies and performs one or more rationalization actions based on the alarm system input and at least one of the alarm system principles 104 of the industrial plant 111 to determine a rationalized future state of the alarm system. The rationalization actions include at least one of the following: creating an alarm priority for the alarm system based on at least a consequence value and an urgency multiplier, enabling an alarm, developing an alarm group, developing an alarm logic, developing an alarm threshold, and validating the alarm logic. Next, in an exemplary embodiment, the machine-learned model 103 compares the rationalized future with the target state to determine a rationalization effect of performing the one or more rationalization actions. Additionally, the machine-learned model 103 determines whether to re-execute the method based on the rationalization effect. Once the desired rationalization effect is maintained, the machine-learned model 103 generates alarm settings and corrective actions based on the rationalized future state. The rationalized future state, alarm settings, and corrective actions are presented to the plant operator via a display of the alarm system computer 106.

[0052] In an exemplary embodiment, in both the start-from-scratch approach and the start-from-scratch approach, an additional step of evaluating a performance evaluation based on at least one of customer and operator feedback is involved for training the machine-learned model 103 for use by the industrial plant. Moreover, in both cases, the machine-learned model 103 is retrained based on the performance evaluation.

[0053] Reference now Figure 8 , the alarm management process is generally indicated by reference numeral 800. At step 801, a master list of alarms is made and exported from the alarm system. In the following steps, this list is marked in red. Step 802 requires that all required alarms, such as regulatory, hazard and operability (HAZOP), incident, procedure and training alarms, be retained. At step 803, common repeating elements in the alarm list are identified and alarms are issued at step 804. At step 805, the remaining alarms in the alarm list are analyzed, and at 806, unresolved issues in the list are found and fixed.

[0054] The deployment of the AI-based alarm rationalization system 10 in a process control system employing aspects of the present disclosure promotes more sustainable plant operations by efficiently utilizing and maintaining plant resources. The AI-based alarm rationalization system 10 provides continuous alarm rationalization throughout the life cycle of the plant, reduces alarm rationalization time and cost, and provides more process insights that were previously unavailable, which provides a more predictive alarm management system.

[0055] Having described the invention in detail, it will be apparent that modifications and variations are possible without departing from the scope of the invention as defined in the appended claims.

[0056] Embodiments of the present disclosure include special-purpose computers, which include a variety of computer hardware, as described in more detail herein and can be run with other special-purpose computing system environments or configurations, even if described together with an example computing system environment. The computing system environment is not intended to imply any limitation on the scope of use or function of any aspect of the present invention. In addition, the computing system environment should not be interpreted as having any dependency or requirement related to any one component or component combination shown in the example operating environment. Examples of computing systems, environments and / or configurations that can be applicable to various aspects of the present disclosure include, but are not limited to, personal computers, server computers, handheld or laptop devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, mobile phones, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0057] Aspects of the present disclosure may be described in the general context of data and / or processor executable instructions (such as program modules), wherein instructions are stored in one or more tangible, non-transient storage media and executed by one or more processors or other devices. In general, program modules include, but are not limited to, routines, programs, objects, components, and data structures that perform specific tasks or implement specific abstract data types. Aspects of the present disclosure may also be practiced in a distributed computing environment, wherein tasks are performed by remote processing devices linked through a communication network. In a distributed computing environment, program modules may be located in local and remote storage media including memory storage devices. For purposes of illustration, programs and other executable program components may be shown as discrete blocks. However, it should be appreciated that such programs and components reside in different storage components of a computing device at different times and are executed by the data processor of the device.

[0058] In operation, a processor, computer, and / or server may execute processor-executable instructions (e.g., software, firmware, and / or hardware), such as those instructions that implement various aspects of the present invention as shown herein. Processor-executable instructions may be organized into one or more processor-executable components or modules on a tangible processor-readable storage medium. Moreover, embodiments may be implemented with any number and organization of such components or modules. For example, aspects of the present disclosure are not limited to specific processor-executable instructions or specific components or modules shown in the figures and described herein. Other embodiments may include different processor-executable instructions and components with more or less functionality than shown and described herein.

[0059] Unless otherwise noted, the execution order of the operations according to various aspects of the present disclosure shown and described herein is not necessary. That is, unless otherwise noted, the operations can be performed in any order, and embodiments can include more or less operations than disclosed herein. For example, it can be expected that performing a specific operation before, simultaneously with, or after another operation is within the scope of the present invention.

[0060] Not all components depicted as shown or described are required. In addition, some implementations and embodiments may include additional components. Changes in the arrangement and type of components may be made without departing from the spirit or scope of the claims as set forth herein. In addition, different or fewer components may be provided, and components may be combined. Alternatively or additionally, a component may be implemented by several components.

[0061] When introducing elements of the present invention or the embodiments thereof, the articles "a," "an," and "the" mean that there are one or more of the elements. The terms "comprising," "including," and "having" are intended to be inclusive and mean that there may be additional elements other than the listed elements.

[0062] In view of the above, it will be seen that the several objects of the invention are achieved and other advantageous results attained.

[0063] As various changes could be made in the above product without departing from the scope of the invention, it is intended that all matter contained in the above description and shown in the accompanying drawings shall be interpreted as illustrative and not in a limiting sense.

[0064] The Abstract and Summary are provided to help the reader quickly ascertain the nature of the technical disclosure. They are submitted with the understanding that they will not be used to interpret or limit the scope or meaning of the claims. The Summary is provided to introduce a selection of concepts that are further described in the Detailed Description in a simplified form. The Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to aid in determining the claimed subject matter.

Claims

1. A computer-implemented method for rationalizing an alarm system within an industrial plant, the method comprising: providing, via at least one or more alarm system databases, an alarm system input, the alarm system input comprising at least one of a training input and an operational input regarding the alarm system; Providing as input predetermined alarm system principles for industrial plants; The machine-learned model is executed by the processor to: analyzing alarm system inputs to output the current state of the alarm system; identifying and performing one or more rationalization actions for the alarm system based on at least one of a current state and a predetermined alarm system principle to output a rationalized future state of the alarm system; and Wherein at least one of the rationalization actions comprises modifying an alarm priority within the alarm system based on at least an urgency multiplier, a consequence value, and a severity score.

2. A computer-implemented method as described in claim 1, wherein the alarm system input includes at least one of the following: existing alarm key performance indicators, existing alarm rationalization efforts, customer performance guidelines, simulated value trends, master alarm database data, alarm history data, operator action log data, and control loop details.

3. The computer-implemented method of any of claims 1-2, wherein the performing one or more rationalization actions further comprises at least one of: disabling an alarm, enabling an alarm, modifying an alarm group, modifying alarm logic, validating alarm logic, and performing no changes.

4. The computer-implemented method of any one of claims 1 to 3, further comprising: Receives a target state as input; while executing the machine-learned model, comparing a rationalized future state of the alarm system to at least one of a current state and a target state to output a rationalized effect of performing the one or more rationalized actions; as well as While executing the machine-learned model, a determination as to whether to re-execute the method is output based on the rationalization effect.

5. The computer-implemented method of any of claims 1-4, further comprising, while executing the machine-learned model, outputting an alarm setting based on a rationalized future state of the alarm system, and further comprising, while executing the machine-learned model, outputting a corrective action for clearing an activated alarm within the alarm system based on the rationalized future state.

6. The computer-implemented method of any one of claims 1-5, further comprising, when executing the machine-learned model, outputting urgency classifications for a plurality of time thresholds and outputting an urgency multiplier for each of the urgency classifications.

7. The computer-implemented method of any one of claims 1-6, further comprising outputting a severity classification for a plurality of severity impacts and a severity score for each of the severity impacts when executing the machine-learned model, and further comprising outputting a priority classification for a plurality of breakpoint thresholds when executing the machine-learned model.

8. The computer-implemented method of any one of claims 1 to 7, further comprising: Receive customer and operator feedback as input; outputting a performance evaluation of the alarm system based on at least one of customer and operator feedback while executing the machine-learned model; as well as Retrain machine-learned models based on performance evaluation.

9. A computer-implemented method for rationalizing an alarm system within an industrial plant, the method comprising: providing, via an alarm system database, an alarm system input including at least one of an operational input and a training input for developing an alarm system within the industrial plant; Provide the principles of the intended alarm system for industrial plants as input; The machine-learned model is executed by the processor to: identifying and performing one or more rationalization actions based on the alarm system input and predetermined alarm system principles to output a rationalized future state of the alarm system including the one or more alarms; Wherein at least one of the one or more rationalization actions comprises creating an alarm priority of the alarm system based on a consequence value, a severity score, and an urgency multiplier.

10. The computer-implemented method of claim 9, wherein the performing one or more rationalization actions further comprises at least one of: enabling alerts, developing alert groups, developing alert logic, validating alert logic, and developing alert thresholds.

11. The computer-implemented method of any of claims 9-10, further comprising, while executing the machine-learned model, outputting alarm settings for the one or more alarms based on a rationalized future state of an alarm system.

12. The computer-implemented method of any of claims 9-11, further comprising, while executing the machine-learned model, outputting corrective actions for the one or more alarms when the one or more alarms are activated within an alarm system based on a rationalized future state.

13. The computer-implemented method of any one of claims 9 to 12, further comprising: Receives a target state as input; When executing the machine-learned model, comparing a rationalized future state of the alarm system to a target state to output a rationalized effect of executing the one or more rationalized actions; as well as While executing the machine-learned model, a determination as to whether to re-execute the method is output based on the rationalization effect.

14. An alarm rationalization system for optimizing an alarm system within an industrial plant, the system comprising: one or more alarm system databases configured to store at least one of alarms and process data; A predetermined alarm system principle configured to manage alarm system decisions within an industrial plant; a processor configured to execute the machine-learned model, wherein in executing the machine-learned model, the processor is configured to receive as input at least one of alarm data and process data to analyze and provide as output a current state of an alarm system within the industrial plant, the processor is configured to receive as input a predetermined alarm system principle, and in executing the machine-learned model, the processor is further configured to perform one or more rationalization actions on the alarm system based on the current state and the predetermined alarm system principle to provide as output a rationalized future state of the alarm system; as well as Wherein at least one of the rationalization actions comprises modifying an alarm priority within the alarm system based on a consequence value or a severity score and an urgency multiplier.

15. The alarm rationalization system of claim 14, further comprising a controller operably connected to the processor, the controller configured to receive as input from the processor an alarm setting change based on the rationalized future state and automatically change the alarm settings of the alarms within the alarm system.

Images