Identification of facility status and operating mode in specific event scenarios

Abstract

The present invention is entitled "Identification of Facility States and Operating Modes in Specific Event Contexts". The present invention provides a system for processing historical facility data related to facility states and operating modes. The historical facility data is clustered into groups representing facility states and operating modes. The groups are used to determine the current state and mode of the facility. When the facility is in a normal state, the system determines whether an event in the facility is an abnormality. If an abnormality is identified, the system transmits a signal indicating the abnormality.

Description

Technical Field

[0001] The present disclosure relates to a data-driven system and method for determining the status and operating mode of a facility in a specific event context. Background Art

[0002] Facilities such as industrial plants, refineries, manufacturing plants, pulp and paper mills, steel mills, and office buildings and other buildings can have many states and modes. The state of a facility refers to the overall state of a facility or a unit in a facility. For industrial plants, the four generally accepted states of the plant are maintenance, startup, shutdown, and normal operation. During the maintenance, startup, and shutdown states, the initial conditions and process parameter values ​​can be significantly different from the normal operating state. The operating mode of a plant is the combined activities of the plant units and assets over a period of time to achieve a specified production output. For example, in a petrochemical refinery, the type and quality of crude oil input to the plant will determine the quantity and quality of different products (such as kerosene, diesel, and gasoline) that can be refined from the crude oil. Therefore, the plant state and operating mode are affected by the products produced by the plant, operating constraints, process constraints, the state of non-automated parameters, plant loads, equipment maintenance, equipment failures, process anomalies, and environmental conditions.

[0003] There are times when plant stability is most at risk because these times may be the most difficult to control and anticipate. These times are generally considered maintenance, startup, and shutdown conditions or events. These events are the three most common types of events that occur, are not appropriate for normal plant operations, and directly impact plant stability and the safe operation of equipment and processes. In current plant operations, unit / asset performance monitoring and key performance indicators (KPIs) do not contain data-driven and / or easily consumable information to determine and identify the cause of the event and diagnose the plant state and operating mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Figure 1 Examples illustrating different states of an industrial plant or other facility are shown.

[0005] Figure 2 A schematic diagram of an embodiment of a system for determining an industrial plant state and an industrial plant mode is shown.

[0006] Figure 3 is a diagram illustrating the types of data used to train a system to determine the status and operating mode of an industrial plant.

[0007] Figure 4 Is a diagram used to determine the status of industrial plants.

[0008] Figure 5A and Figure 5B It is a flow chart of the implementation plan for determining the status and operating mode of the plant.

[0009] Fig. 6A and Figure 6B is a flow chart of another embodiment for determining the status and operation mode of a plant.

[0010] Fig. 7A and Figure 7B is a flow chart of another embodiment of determining the state and operation mode of an industrial plant.

[0011] Figure 8 is a block diagram of a computer system upon which one or more embodiments of the present disclosure may be executed. DETAILED DESCRIPTION

[0012] In the following description, reference is made to the accompanying drawings, which form a part of the detailed description, and in which specific embodiments that can be implemented are shown by way of illustration. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention, and it should be understood that other embodiments may be utilized and structural changes, electrical changes, and optical changes may be made without departing from the scope of the invention. Therefore, the description of the following exemplary embodiments is not restrictive, and the scope of the invention is defined by the appended claims.

[0013] Figure 1 Examples of different possible states of an industrial plant are shown. Figure 1 Industrial plants are referred to, but embodiments of the present disclosure encompass other facilities such as refineries, manufacturing plants, pulp and paper mills, steel mills, and office buildings and other buildings. In normal operation, a control system 110 manages normal plant operations 120. Alarms 130 may occur due to process changes or other events, and these alarms may require operator intervention 140. If the operator is unable to manage the alarm condition, this may result in process anomalies and / or equipment failures or trips 150. In some cases, to avoid plant damage and manage safety, an emergency shutdown (ESD) system 160 may trip the plant and / or plant units.

[0014] To address these situations, one or more embodiments are directed to systems and methods for identifying unique states and operating modes of an industrial plant or other facility during operation of the plant or facility. More specifically, one embodiment describes the implementation and validation of a self-configuring, self-updating, machine learning-based artificial intelligence algorithm to determine plant states and operating modes by detecting when relationships between manipulated parameters change. Data available in the plant's digital control system is used, and therefore no human input is required.

[0015] The ability to determine the state and operating mode of a plant in real time or near real time can be used in several different ways to improve the operation of the plant. First, the operator's response to an event or abnormal condition in the plant will vary based on the state of the plant and the operating mode of the plant at the time of the event. Therefore, understanding both the state and the mode helps the operator respond to plant events. Second, the operator has several process variables (PVs) and key performance indicators (KPIs) related to the tag (or tags) in the alarm that tend to appear on his or her control room computer screen. Understanding the current plant state and mode will help the operator respond to these trends and alarms. Third, the operator uses these trends to determine the response required to manage such alarm events. Therefore, understanding the current state and mode of the plant will help highlight the operator's behavior and / or reaction during normal operation, startup, tripping, and emergency shutdown.

[0016] In addition, to achieve the next level of improvement in industrial plant performance, one or more embodiments include automated methods for performing the following. First, an embodiment can predict guidance indications for abnormal events (e.g., process pressure). The embodiments that handle these abnormal events are robust (e.g., improved strength of protection and control systems). Second, the embodiment can estimate future values ​​of key performance indicators (KPIs) to enable corrective actions to resolve problems. Third, the embodiment can provide real-time guidance to operators to ensure optimal operation of the plant and / or ensure appropriate responses to events. Fourth, the embodiment can provide an assessment of the operator's work capabilities. Such assessments can then be used to develop and / or improve operator training courses.

[0017] These embodiments include components for estimating plant states and plant operating modes. The reason for the need to accurately and in real time determine plant states and plant operating modes is that acceptable performance in one state or operating mode may be unacceptable in another state or operating mode. Therefore, correctly determining plant states and operating modes on a real-time or near real-time basis helps predict guidance indications for abnormal events, estimate future values ​​of KPIs, provide real-time guidance to operators, and evaluate operator performance.

[0018] The conditions that cause and / or cause changes or variations in plant states and plant operating modes are as follows. The type of product produced by the plant affects the plant state and operating mode. Plant units or assets can be constructed differently for different product types. For example, if two different products are kerosene and diesel fuel, the patterns of the time series process variable data of temperature and pressure in the distillation column will be different when the plant is in different modes. In addition, these changes can be sensed when there is a conversion from one product to another. Therefore, these different product types can be reflected in the online or offline analyzer measurements of the feed input and the actual product output itself for manufacturing the product. Plant states and operating modes may also change when the plant undergoes maintenance cycles, startup cycles (cold or hot), and planned or unplanned shutdowns. Plant units or assets can be constructed differently for different plant loads, that is, different product throughputs (e.g., 50% or 90% of total output). These different plant loads are reflected in the process variable data of the feed / input stream tag value. The environmental conditions surrounding the plant can be different and can affect the plant state and the plant operating mode. For example, in a food processing plant, there is a greater likelihood that there will be different cooling requirements in the summer months than in the winter months. Such differences can be reflected in the process variable data related to the utility flow. In addition, there may be various process constraints at the plant, which may affect the plant state and / or plant mode. For example, a constraint may be that a particular unit or asset is unavailable. As another example, the catalyst may have degraded and the plant process will subsequently have to operate at a higher temperature. Finally, there are various non-automated parameters in the plant, such as manual valves and manual locks for units / assets, and the state of these non-automated parameters can affect the plant state and plant mode.

[0019] In response to the need to be able to accurately determine the plant state and plant mode of an industrial plant in real time or near real time, one or more embodiments call for a self-configuring, self-updating, machine learning-based artificial intelligence algorithm to determine the plant state and plant operating mode (when the plant is in normal operating conditions) by detecting the relationship between changes in operating parameters. One embodiment processes and integrates multiple data sources to understand the plant state and plant mode. Figure 2 The main data sources are shown, which are used to first train the model to determine the plant state and plant operation mode, and then interpret the production data in real time to determine the plant state and plant operation mode. In one embodiment, the training is performed using a clustering method. The main data sources include: sensor data (process parameter values) from the plant process history; human-machine interface (HMI) graphics, operation logs, and plant key performance indicators (KPIs); and engineering configuration data field logs (e.g., accident reports) from field operators. These data sources are normalized and integrated into a format that is clustered and grouped into different plant states and operation modes. Figure 2These data sources are shown, as well as the entire training process and execution process. Specifically, time series process values, event data, and log file data are integrated from the plant 210, the operation room / control room 220, and the field data 230. These data may include: sensor data 242; HMI data, engineering configuration data, operation logs, plant KPI data, and event and historical data (244); and field logs and accident reports (246). These data are first processed in the machine learning / training stage 250. The learning / training generates a data mining and analysis module / model 260. The production data is then analyzed at 270 to detect the plant status and plant operation mode at 280.

[0020] Figure 2 The integration of multiple data sources in the process includes the integration of process parameter values ​​and HMI data. Figure 3 As further shown in Figure 3 As shown, the values ​​of input flow tags and output flow tags are identified to help analyze the feed input and final product output flow patterns, thereby determining the plant status and plant operation mode. More specifically, Figure 3 The types of data 310, the information 330 extracted from the data, and the use of the information in determining the state and operating mode of the plant are shown. The types of data include event data 312, configuration data 314, graphical data 316, process history data 318, and operator log data 320. The process history data 318 and operator log data 320 are primarily used for training the system, but history-based configuration data 314 and graphical data 316 may also be used. Event data 312 is real-time data used as input for the training system, and the training system uses this real-time data to determine the plant state and mode. For example, when the plant is operational, a process alarm 313 may occur. The process alarm may be related to the state of a piece of rotating equipment 322, such as the revolutions per minute (RPM) of the rotating equipment 322 is lower than its normal operating value. Then, as shown at 332, these event data 312 help identify different states of the plant.

[0021] The process of using event data 312 to identify plant states and patterns is described in Figure 4 The data of the flow label 405, the rotating equipment status 406 and other information are used to identify the shutdown, startup, normal operation and maintenance status, such as Figure 4 For operators, Figure 4All conditions in should be true (TRUE) to conclude that the plant is in a particular state. For example, if the flow tag is decreasing and trending downward toward the process variable lower limit and there is a ramp-down of the feed flow (411), the rotating equipment (e.g., pumps and other equipment that moves input materials and output products) is sequentially stopped (412), the safety devices (ESD valves) are sequentially shut down (413), the continuous alarm event exceeds the threshold (414), and the operator issues a stop and shutdown command (415), then all conditions of the normal shutdown (410) are true (TRUE) and it can be concluded that the plant is undergoing a normal shutdown. Again, this is the intelligence that teaches the system so that the system can determine the state and mode of the plant.

[0022] In a similar way, Figure 4 Data used to determine whether the plant is in an emergency shutdown (420), a cold start (430), a hot start (440), a normal operating state (450), or a maintenance state (460) is shown.

[0023] As described above, one embodiment first determines the state of an industrial plant in real time or near real time, and then determines the operating mode of the plant if the plant is in a normal operating state. Knowing the plant state and operating mode allows an operator to determine whether a current event (e.g., as indicated by a process variable) is of interest. Before describing the details of the embodiment, the different possible plant states are discussed in more detail.

[0024] Normal plant operation is the plant state in which the operator maintains the operating variables / parameters within defined operating limits. Plant processes should operate under steady-state conditions. Significant deviations from normal operation lead to abnormal conditions. Normal operation of the plant has two sub-states: normal and abnormal. The normal state is detected by analyzing whether the process variable (PV) values ​​of all parameters are within the operating range without any alarm events. The normal state has two additional sub-states: optimized operation and non-optimized operation. Optimized operation is detected by analyzing whether the PV has a minimum deviation from a defined set point (SP). Non-optimized operation is detected by analyzing whether the PV deviates from the SP, but the PV does not exceed any alarm limit. Abnormal states include alarm events, which are detected by analyzing whether the PV has deviated from the alarm limit values.

[0025] Major shutdowns in industrial plants usually occur infrequently (every year or two). However, when such shutdowns occur, they usually take several days to complete. Generally speaking, these shutdowns have two purposes. First, to repair the problems identified during the previous major shutdown, and second, to check the parts of the plant that are inaccessible during operation in order to identify the problems that will be repaired during future planned shutdowns. There are usually two types of shutdowns: normal shutdowns and emergency shutdowns. Normal shutdowns are part of the plant protection system, and their purpose is to minimize the frequency and consequences of excursions outside the operating range of the facility. Possible causes of normal shutdowns include composition changes in the feed (e.g., feed raw material changes, product demand changes), process limitations (e.g., limitations of exchangers, containers, towers and other equipment to meet process requirements), maintenance activities (e.g., regeneration / activation of catalysts, management of pressure differences or heat transfer problems through heating furnaces / exchangers, misoperation of compressors, pumps and towers), and management of parameter problems that degrade products. Emergency shutdowns are part of the plant protection system, and their purpose is to keep the plant process within the design limits and prevent abnormal conditions from deteriorating into major hazardous events. Possible causes of emergency shutdowns include equipment failure, process upsets / trips, process / operating constraints (such as pipeline blockages), tower overflows / leaks, runaway conditions, overpressure, and process control problems.

[0026] A plant start is identified as the time between shutdown and normal operation. Plant starts have two subcategories: cold start and hot start. In a cold start, variables such as casing temperature, shaft temperature, and bearing temperature of the equipment are at ambient temperature, and there is zero (0%) holdup in tanks, vessels, and other similar equipment. A cold start is typically followed when equipment such as a turbine is in a stopped / standby state for more than a certain amount of time, such as 20 hours. In a hot start, variables such as casing temperature, shaft temperature, and bearing temperature are not equal to ambient temperature, and holdup in tanks, vessels, and other equipment is greater than 0%.

[0027] Typical reasons for going into unplanned maintenance include failure to replace worn parts, lack of lubrication, and other issues. Information indicating maintenance status is captured from condition monitoring alarms, maintenance override switch (MOS) activation events, red tag identification (tagged equipment that is out of service), and flows across plant equipment.

[0028] As previously described, when it is determined that the plant is in normal operation, then an embodiment can determine not only the current operating mode of the plant, but also whether an event in the plant is a cause for concern. This determination of whether attention is needed is based on a machine learning (ML) / artificial intelligence (AI) algorithm trained with historical plant data (to learn the relationship between all relevant events / data / equipment) and immediate events, and using the trained ML / AI algorithm to make a decision on the need / cause for attention based on the plant events. In one embodiment, the ML / AI algorithm involves a clustering method.

[0029] The plant operating mode varies based on several factors. These factors include the type of product being produced, the current load of the plant, the environmental conditions in which the plant is currently operating, any operating constraints, and the state of any non-automated parameters. When the value of the key variable changes to another steady state, or when the correlation between the parameters changes (as learned by the training of the ML / AI algorithm), the implementation scheme determines that the plant unit / asset is in a different operating mode. During normal plant operation, all alarm events require a response from the operator to return to the normal operating range. However, the operator may face different scenarios, such as event flooding (i.e., multiple events or alarms occurring at the same time), operating constraints (which will prevent the operator from responding to the alarm / event in the most effective way), and flow label trends. In order to optimally respond to events, it is necessary to be able to distinguish such scenarios. Operating mode recognition provides context for events, that is, the operating mode in which a given event occurs. Simply put, in order to achieve accurate fault detection, it is necessary to accurately detect the plant unit / asset operating mode, because a fault in one operating mode may not be a fault in another operating mode. In one embodiment, the method of determining the operating mode is discussed in conjunction with the following figures: Figure 5A , Figure 5B , Fig. 6A , Figure 6B , Fig. 7A and Figure 7B .

[0030] One or more embodiments of the present disclosure have several unique and novel features. First, one embodiment can automatically identify key plant tags that affect plant status. That is, given an event on a single or a group of process variables (PVs), it is necessary to check specific relevant PVs to determine the plant status and operating mode when the event occurred. These relevant PVs are learned from ML / AI algorithms trained on plant historical data. The embodiment can automatically identify and process the main data components related to candidate tags using configuration data and graphic data (HMI).

[0031] Second, an embodiment can identify the operating mode of the plant in real time. As previously described, the embodiment uses historical plant data to perform machine learning on the plant operating mode, and then uses the machine learning model to process the real-time data from the plant and determine the operating mode of the plant. In addition, when new operating modes are identified through the learning process and the operator confirms these new modes, the database containing the learned operating modes (models) is updated. The features of this embodiment can assist and guide the operator in the real-time operation of the plant.

[0032] Third, one embodiment achieves real-time optimization of plant processes by identifying and evaluating deviations from steady-state values ​​to determine the duration when the plant is not in an optimal operating mode, and further identifying the factors that cause the operating deviations. This information will enable operators to adjust parameters to further optimize the process. Again, this is further assisted by a training model that identifies plant states and plant operating modes in real time.

[0033] Fourth, as described above, one embodiment implements an additional system for industrial plant monitoring that facilitates the next level of improvement in plant performance. For example, the embodiment can predict guidance indications for abnormal events and improve the robustness of handling these abnormal events. Future values ​​of key performance indicators (KPIs) can be estimated to enable corrective actions to be taken. Operator performance can be assessed to enable training interventions. Real-time guidance can be provided to operators to ensure optimal operation.

[0034] Figure 5A , Figure 5B , Fig. 6A , Figure 6B , Fig. 7A and Figure 7B is a block diagram illustrating the operation and features of a system and method for determining the state and operating mode of a plant. Figure 5A , Figure 5B , Fig. 6A , Figure 6B , Fig. 7A and Figure 7B The plurality of blocks 505 to 576, 605 to 674 and 705 to 761 are included respectively. Figure 5A , Figure 5B , Fig. 6A , Figure 6B , Fig. 7A and Figure 7BThe blocks may be arranged substantially continuously in the flowchart, but other examples may use multiple processors or a single processor organized into two or more virtual machines or sub-processors to reorder the blocks, omit one or more blocks, and / or execute two or more blocks in parallel. In addition, other examples may implement the blocks as one or more specific interconnected hardware or integrated circuit modules, with related control signals and data signals transmitted between and through the modules. Therefore, any process flow is applicable to software, firmware, hardware, and hybrid implementations.

[0035] See first Figure 5A and Figure 5B , operations 505, 506, and 507 perform data engineering tasks to prepare the plant data for analysis to determine plant status and patterns. The preparation includes segmenting the data into subsets. At 505, by collecting data and preparing it for analysis Figure 3 At 506, the environmental conditions under which the plant will operate are prepared by subsetting the data for different environmental conditions (e.g., summer / winter). This is accomplished by analyzing environmental parameters from the process data (e.g., ambient temperature, fixed-point weather reports) so that, for example, data from summer months are placed in one set and data from winter months are placed in another set. At 507, the data is subsetted as a function of plant load. For example, in one embodiment, the data is subsetted based on plant throughput loads greater than 50% and plant throughput loads less than 50%. In operation 507, the configuration data ( Figure 3 , No. 314) and HMI data ( Figure 2 , No. 244) identifies tags for plant KPI PV data (e.g., controller PV, SP data) and throughput data. The values ​​of these tags are used to subset the data for plant throughput loads greater than 50% and plant throughput loads less than 50%. For example, at 50% load, only one of the three pumps may be required. However, at 100% load, two of the three pumps may be required. Operations 505, 506, and 507 produce several subsets, each representing a different operating cycle. Each different operating cycle is processed separately by the following process.

[0036] Additionally, during the data preparation phase of runs 505, 506, and 507, historical data for each tag is analyzed to determine the distribution and density of data values. The percentage of data in each range is calculated. It is assumed that the range where most data exists is the steady-state range. Note that these may or may not be continuous ranges.

[0037] At 510, data from operations 505, 506, and 507 are used to determine the plant state and plant operating mode of the industrial plant. Operation 510 involves determining whether the plant is in a state of normal operation 520, planned maintenance 530, planned startup 540, or shutdown mode 550. The process for determining the normal operating mode begins at operation 520 and includes determining the context 521 of an event (such as a process alarm 313). At 522, the trend of the event and its relationship to the relevant process value flow tags are analyzed. As indicated at 522A, these flow tags are assigned to specific plant assets. At 522B, flow tags that may be related to the flow tags at 522A are considered. At 523, persistent pre-alarms are considered. For example, if the upstream alarm is persistent (i.e., it has not been resolved and reset), the final resolution of the problem that caused the persistent alarm may also resolve the current alarm 521 in question. At 524, the steady-state value of the process variable associated with the alarm is determined. The degree of deviation from the steady-state process value can help determine the severity of the alarm. At 525, the relevant tags are analyzed again to help determine the event context. At 526, the status of other operator tags may be considered. Finally, at 527, other operating constraints are considered. For example, a particular plant unit may be currently out of service. This may involve analyzing historical data for all tags and calculating a percentage range (527A).

[0038] At 530, the system is trained to recognize that the plant is in a planned maintenance state. The plant can be determined to be in a planned maintenance state at 531 by notifying that a maintenance override switch (MOS) is activated, and the plant can be determined to be in a planned maintenance state at 532 by determining that a particular piece of equipment or plant unit has been red-tagged (i.e., a red tag has been attached to the equipment indicating that it is not in proper working order).

[0039] Operation 540 describes in detail the determination of the plant being in the planned startup state. Figure 5B As shown in , the planned start can be a cold start at 541 or a hot start at 543. Regardless of whether it is a cold start or a hot start, at 542 and 544, all operating levels and temperature tags are analyzed to determine and verify whether it is a cold start or a hot start. This is Fig. 6A and Figure 6B Detailed description is given further in .

[0040] Run 550 and its related runs determine that the state of the plant is a shutdown state. At 560, it can be determined that the shutdown is a planned shutdown, and at 570, it can be determined that the shutdown is an emergency shutdown. At 561, all process value stream tags are analyzed by first considering processing the planned shutdown 560. At 562, all persistent alarms during the duration of the planned shutdown are analyzed. The persistent alarms can provide information about the cause of the planned shutdown. Similarly, at 563, rotating equipment alarms are analyzed during the duration of the planned shutdown. Finally, at 564, the controller operation is analyzed in conjunction with the planned shutdown.

[0041] As shown at 570, the emergency shutdown is automatically handled by the plant system. Historically, such emergency shutdowns may be handled automatically by the plant system or manually by an operator (e.g., when the operator observes that the process value is outside the safe operating range). Similarly, in one embodiment, the emergency shutdown may be handled automatically or by operator intervention. However, one advantage of the embodiment is that the emergency shutdown decision is more robust due to the training and learning of the system. As indicated at 571, the emergency shutdown may be triggered by a process anomaly and / or a process trip, or as indicated at 572, by detecting an equipment failure. During the emergency shutdown, all process value stream tags are recorded and analyzed at 573. Similar to the case of planned shutdown 560, all continuous alarms during the duration of the emergency shutdown are analyzed at 574, and any rotating equipment alarms are analyzed during the duration of the emergency shutdown at 575. Finally, at 576, the status of all safety solenoid values ​​is analyzed.

[0042] exist Figure 5A and Figure 5B The process of determining the plant status is shown in Fig. 6A and Figure 6B. At 605, similar to operation 506 and operation 507, plant unit / asset data is subsetted by environmental conditions and plant loads. At 610, historical data related to each plant unit / asset is retrieved and the time window of the asset is checked. Then, at 615, the key process indicator (KPI) variable values ​​are checked to determine whether they are within the correct operating range and not less than the process variable lower limit. At 620, if not all KPI variable values ​​and all KPI variable values ​​are within the correct operating range and not all KPI variable values ​​are not less than the process variable lower limit, then at 622 the system determines whether the plant is in the startup state 624. In order to check at 622 and determine at 624 that the plant is in the startup state, all KPI values ​​must be in a transient state at 625 (i.e., all KPI process variable values ​​are greater than the process value lower limit and less than the operating range), all rotating equipment states must be open at 626, and at 627, the number of continuous alarms must be less than a threshold (e.g., less than the ASM standard).

[0043] If at 620, all KPI variable values ​​are within the correct operating range and all KPI variable values ​​are not less than the process variable lower limit, the process checks whether the plant is in normal operating state at 630. If the plant's controller operations are all within normal ranges (631), and if the status of all rotating equipment is open (632), and all safety solenoid tags are open (633), and the number of continuous alarms is less than a threshold (e.g., less than the ASM standard) (634), then at 630, it is determined that the plant is in normal operating state 635.

[0044] If the plant is not in normal operation at 630, 635, the system checks at 640 to determine if the plant is in a shutdown state 645. This is performed by determining whether MOS is enabled (641); determining whether a particular plant unit has been red-tagged (642); and determining whether the flow label value across plant equipment is equal to zero (643). If one or more of operations 641, 642, and 643 are not true, then at 645 it is determined that the plant is in a shutdown state. If all operations 641, 642, and 643 are true, then starting at 650, the system determines whether the plant is in a shutdown state 674 or a maintenance state 672. Specifically, at 655, a query is made to determine whether the amount of time a particular piece of equipment has been red-tagged. If this determination cannot be made, then at 656, the equipment switching time is identified in order to derive the maintenance time period (i.e., the red tag time period). If this determination can be made at 655, then at 657, the equipment that was switched during the red tag period is identified (i.e., the amount of time the equipment was red tagged). In either event, after running 656 and running 657, at 660, the maintenance override switches (MOS) that were enabled during the duration of the red tag event (i.e., the time the tag was in the red zone) are identified. At 665, the MOS enabled switches are connected to the equipment that was switched, and at 670, if there are a large number of plant assets under maintenance in the plant fleet, then at 674, the plant fleet is in a shutdown state. If not, then at 672, the plant fleet is in maintenance mode.

[0045] Fig. 7A and Figure 7B is a flow chart of another exemplary embodiment that uses programmed expert intelligence and machine learning relationships between facility process variables to determine the current state and operating mode of a facility. The current state and operating mode of the facility can then be used to determine whether a plant event is truly abnormal, to correctly estimate future values ​​of key performance indicators (KPIs), to provide real-time guidance to facility operators, and to evaluate work actions of facility operators. Now specifically refer to Fig. 7A and Figure 7B At 705, facility data related to the status and operating mode of the facility is received into the processor-based system. These facilities may include industrial plants, such as refineries, chemical plants, pulp and paper mills, and steel mills. These facilities may also include buildings, such as office buildings and schools. The facility data may include process data, such as event (e.g., alarms from process variable sensors) data, configuration data, graphical data, historical process data, human machine interface (HMI) graphical data, key performance indicator (KPI) data, and operator log data. See Figure 2The facility data may also include building status data. In one embodiment, the facility status may include states such as a normal state, a startup state, a shutdown state, and a maintenance state. The facility mode relates to a specific configuration of a facility, such as whether a refinery is making diesel fuel or kerosene, or whether an office building is using its cooling system or heating system.

[0046] In one embodiment, the data is structured prior to further processing. For example, as shown at 706A, the facility data is structured by dividing the facility data into subsets. These subsets may be based on certain environmental conditions, certain facility load conditions (e.g., the facility is running at 80% of capacity), and certain time windows. Additionally, as shown at 706B, the facility data may be normalized, cleaned (e.g., by removing outliers), and missing data may be filled (e.g., by interpolation).

[0047] At 710, the facility data is grouped into clusters. Although clustering is used in one embodiment, other forms of machine learning may also be used. Each cluster in the cluster relates to a specific facility state and a specific facility operating mode. In addition, each cluster in the cluster includes a descriptor associated with a specific facility state and a specific facility operating mode. For example, the descriptor may include a process variable value, a process variable range, an HVAC set point, and a temperature range. The result of the clustering is a plurality of clusters associated with different facility states and facility operating modes. At 715, the cluster is used to determine the current state and current operating mode of the facility. For example, as shown at 716A, the current facility data is accessed, and at 716B, the current facility data is compared with the cluster. At 716C, a cluster similar to the current facility data is identified. This identification may be performed, for example, by identifying a cluster whose data is within a specific percentage (e.g., within 10%) of the current facility data. In other embodiments, other numerical analysis techniques may be used to identify similar clusters. As shown at 716D, this identification of similar clusters generates a subset of clusters. Then, at 716E, the state and operating mode of the facility can be determined from the identified similar clusters. As described at 717, the state and mode of the facility can be based on factors such as: the type of products manufactured by the industrial plant, planned maintenance of the plant, operating load of the plant, environmental conditions, process constraints, and the state of non-automated parameters.

[0048] As described above, in one embodiment, the state of the facility can be a normal state, a startup state, a shutdown state, and a maintenance state. When the facility is in a normal operating state, then at 720, it is determined whether an event such as an alarm or a change in the facility process is abnormal. At 725, if the event is abnormal, a signal reporting the abnormality is transmitted to an operator or a computer display device. In addition to determining whether the event is abnormal at 720, the system can further perform other operations when the facility is in a normal operating state. For example, as shown at 721, the system can use clusters to estimate the future value of a key performance indicator (KPI). As related to the determination of anomalies, determining the future value of a KPI can provide real-time guidance to the operator of the facility (e.g., suggesting that the operator shut down the plant, reduce the flow rate of a component, etc.), and can be further used to evaluate the operator's work actions (e.g., determining whether the operator follows the advice and / or the degree to which the advice is followed, and / or the result of the operator following the advice).

[0049] When it is determined that the facility is in normal operating state (720), then at 722A, the system may use the cluster to determine whether the facility is in steady state and / or optimal operating mode. For example, the system may have learned from the cluster that when certain KPIs and / or process variables are within a certain percentage of their set points, the facility is in steady state and / or optimal operating mode. Conversely, at 722B, if the facility is not in steady state or optimal operating mode, analysis using the cluster may identify one or more factors that may cause the facility to not be in steady state or operating in optimal operating mode.

[0050] In addition to determining whether an alarm is a cause for concern based on the current state and operating mode of the facility (as learned from machine learning historical data), the system can perform several other functions based on machine learning. For example, at 730, the system can use the cluster to determine whether the facility is experiencing a change in the state or mode of the facility. In one embodiment, this can be performed by examining the values ​​of the key process indicators (KPIs) in the cluster. In addition, the correlation changes between the facility data can be examined. At 731, the system can transmit a signal to an operator or computer display device reporting the change in state or mode.

[0051] The system can also determine the relationship between facility data based on machine learning. For example, facility data directly related to the event and facility data indirectly related to the event are checked. Specifically, at 735, facility data directly related to the event in the cluster is checked. An example of directly related data is pressure data in a storage tank that causes a high pressure alarm. Then, at 736, facility data indirectly related to the event in the cluster is checked. An example of indirectly related data is the flow rate in a pipeline that supplies material to a storage tank. At 737, the system determines an action course in response to the event based on the inspection of direct facility data and indirect facility data. Then at 738, the system transmits a signal indicating the action course to an operator or a computer display device.

[0052] While the system may be initially trained with historical facility data, the system may be retrained as more current data becomes available. Specifically, as shown at 740, the cluster may be regenerated upon identification of a new operating mode of the facility. At 741, the regenerated cluster is used (again, when the facility is in a normal state) to determine whether an event in the facility includes an anomaly. Finally, at 742, a signal indicating the anomaly is transmitted to an operator or computer display device.

[0053] In addition to determining that the facility is in a normal state, the system can also determine whether the facility is currently in another state and / or when another state should be initiated. For example, the system can use the cluster to determine when a normal shutdown is initiated (750), when an emergency shutdown is initiated (755), and whether the detected start is a cold start or a hot start (760). After each of these, a signal is transmitted to an operator or computer display device to convey the information (751, 756, 761).

[0054] Figure 8 It is a block diagram of a machine in the form of a computer system, in which an instruction set for making a machine perform any one or more of the methods discussed herein can be executed. In an alternative embodiment, the machine is operated as a stand-alone device or can be connected (for example, networked) to other machines. In a networked deployment, the machine can be operated as a server or client machine in a client-server network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. In a preferred embodiment, the machine will be a server computer, but in an alternative embodiment, the machine can be a personal computer (PC), a tablet PC, a set-top box (STB), a personal digital assistant (PDA), a mobile phone, a network device, a network router, a switch or a bridge, or any machine that can (sequentially or otherwise) perform the instruction of the action to be taken by the specified machine. In addition, although a single machine is shown, the term "machine" should also be considered to include any set of machines that perform one (or more) instruction sets to perform one or more methods discussed herein individually or jointly.

[0055] The exemplary computer system 800 includes a processor 802 (e.g., a central processing unit (CPU), a graphics processing unit (GPU), or both), a main memory 801, and a static memory 806 that communicate with each other via a bus 808. The computer system 800 may further include a display unit 810. The computer system 800 may additionally include a storage device 816 (e.g., a drive unit), a signal generating device 818 (e.g., a speaker), a network interface device 820, and one or more sensors 828, such as a global positioning system sensor, a compass, an accelerometer, or other sensors.

[0056] The driver unit 816 includes a machine-readable medium 822 on which is stored one or more sets of instructions and data structures (e.g., software 824) embodying or utilized by any one or more of the methods or functions described herein. The software 824 may also reside completely or at least partially in the main memory 801 and / or the processor 802 during its execution by the computer system 800, the main memory 801 and the processor 802 also constituting machine-readable media.

[0057] Although the machine-readable medium 822 is shown as a single medium in the exemplary embodiment, the term "machine-readable medium" may include a single medium or multiple media (e.g., a centralized or distributed database and / or associated caches and servers) that store one or more instructions. The term "machine-readable medium" should also be deemed to include any tangible medium that can store, encode, or carry instructions for execution by a machine and cause the machine to perform any one or more of the methods of the present invention, or any tangible medium that can store, encode, or carry data structures utilized by or associated with such instructions. Therefore, the term "machine-readable medium" should be deemed to include, but is not limited to, solid-state memory and optical and magnetic media. Specific examples of machine-readable media include: non-volatile memory, including, for example, semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.

[0058] The software 824 may further be transmitted or received via the network interface device 820 using a transmission medium over a communication network 826 using any of a number of well-known transmission protocols (e.g., HTTP). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the Internet, a mobile telephone network, a plain old telephone (POTS) network, and a wireless data network (e.g., and The term "transmission medium" shall be deemed to include any intangible medium that is capable of storing, encoding or carrying instructions for execution by the machine, and includes digital or analog communications signals or other intangible medium to facilitate communication of such software.

[0059] It should be understood that there are other variations and modifications to the present invention and its various aspects, which may be apparent to, for example, those of ordinary skill in the art, and the present invention is not limited to the specific embodiments described herein. The above features and embodiments may be combined with each other in different combinations. Therefore, it is contemplated that any and all modifications, variations, combinations or equivalents falling within the scope of the present invention are encompassed.

[0060] The Abstract is provided to comply with 37 CFR § 1.72(b) and will allow the reader to quickly ascertain the nature and gist of the technical disclosure. It is presented with the understanding that it will not be used to interpret or limit the scope or meaning of the claims.

[0061] In the foregoing description of the embodiments, various features are grouped together in a single embodiment for the purpose of simplifying the disclosure. This approach to the disclosure should not be understood to reflect that the claimed embodiments have more features than those expressly stated in each claim. On the contrary, as reflected in the following claims, the subject matter of the invention lies in less than all the features of a single disclosed embodiment. Therefore, the following claims are hereby incorporated into the detailed description, with each claim standing alone as a separate exemplary embodiment.

Claims

1. A method, comprising: receiving, by a computer processor, facility data (705) relating to a state of a facility and an operating mode of the facility, wherein the state of the facility comprises: a normal operating state, a startup state, a shutdown state, and a maintenance state, and wherein the facility operating mode corresponds to a configuration of the facility for performing a particular activity; clustering, by the computer processor, the facility data into groups representing facility states and facility operating modes, thereby generating a plurality of clusters associated with the facility states and facility operating modes (710); determining, by the computer processor, a state and mode of the facility using the plurality of clusters (715); When the facility is in a normal operating state, determining by the computer processor whether an event occurring in the facility includes an anomaly (720); regenerating, by the computer processor, a plurality of clusters when one or more new patterns of the facility are identified (740); When the facility is in normal operation, determining by the computer processor whether an event in the facility includes an anomaly using the regenerated plurality of clusters (741); and transmitting, by the computer processor, a signal indicating the anomaly to a computer display device (725); wherein each cluster of the plurality of clusters relates to a specific facility state and a specific facility operating mode (715, 280); and Each cluster of the plurality of clusters includes a descriptor associated with the specific facility status and the specific facility operating mode (717).

2. The method according to claim 1, further comprising: When the facility is in the normal operating state, estimating, by the computer processor, a future value of a key performance indicator (KPI) using the cluster, the future value of the key performance indicator (KPI) providing real-time guidance to an operator of the facility; and evaluating, by the computer processor, the operator's work actions (721).

3. The method according to any one of claims 1 to 2, further comprising: When the facility is in the normal operating state, determining by the computer processor whether the facility is in a steady state and optimal operating mode using the plurality of clusters (722A); as well as Based on determining that the facility is not in a steady state and an optimal operating mode, one or more factors that cause the facility to not be in the steady state and the optimal operating mode are determined by the computer processor (722B).

4. The method of claim 1 , wherein determining the state and the mode of the facility further comprises: receiving, by the computer processor, current facility data (716A); processing, by the computer processor, the current facility data by comparing the current facility data to the cluster (716B); identifying, by the computer processor, one or more clusters of the plurality of clusters that are similar to the current facility data (716C); generating, by the computer processor, a subset of the plurality of clusters, the subset consisting of the identified similar clusters (716D); as well as The state and mode of the facility is determined, by the computer processor, based on the identified similar clusters (716E).

5. The method according to claim 1, further comprising: determining, by the computer processor, whether the facility is experiencing a change in the state or the mode of the facility by examining a change in a correlation between a value of a key process indicator (KPI) and the facility data (730); as well as A signal reporting the change in state or mode is transmitted by the computer processor to the computer display device (731).

6. The method according to claim 1, further comprising: examining, by the computer processor, the facility data in the plurality of clusters that is directly related to the event (735); examining, by the computer processor, facility data in the plurality of clusters that is indirectly related to the event (736); Determining, by the computer processor, a course of action to respond to the event based on facility data directly related to the event and facility data indirectly related to the event (737); as well as A signal indicative of the course of action is transmitted by the computer processor to an operator or to a computer display device (738).

7. A device, comprising: one or more processors (802); Memory (801); as well as One or more programs stored in a memory (801), the one or more programs comprising instructions for executing a method comprising: receiving, by a computer processor, facility data (705) relating to a state of a facility and an operating mode of the facility, wherein the state of the facility comprises: a normal operating state, a startup state, a shutdown state, and a maintenance state, and wherein the facility operating mode corresponds to a configuration of the facility to perform a particular activity; clustering, by the computer processor, the facility data into groups representing facility states and facility operating modes, thereby generating a plurality of clusters associated with the facility states and facility operating modes (710); determining, by the computer processor, a state and mode of the facility using the plurality of clusters (715); When the facility is in a normal operating state, determining by the computer processor whether an event occurring in the facility includes an anomaly (720); and transmitting, by the computer processor, a signal indicating the anomaly to an operator or to a computer display device (725); wherein each cluster of the plurality of clusters relates to a specific facility state and a specific facility operating mode (715, 280); and Each cluster of the plurality of clusters includes a descriptor associated with the specific facility status and the specific facility operating mode (717).

8. The device according to claim 7, further comprising: When the facility is in the normal operating state, the computer processor uses the cluster to estimate future values ​​of key performance indicators KPIs, the future values ​​of the key performance indicators KPIs provide real-time guidance to operators of the facility; and evaluate the operator's work actions (721).

9. A non-transitory computer-readable storage medium (801), the non-transitory computer-readable storage medium comprising one or more programs for execution by one or more processors (802) of a device, the one or more programs comprising instructions that, when executed by the one or more processors, cause the device to perform the following method: The one or more processors evaluate facility data related to a status of a facility and an operating mode of the facility (705), wherein the status of the facility includes: a normal operating state, a startup state, a shutdown state, and a maintenance state, and wherein a facility operating mode corresponds to a configuration of said facility to perform a particular activity; processing, by the one or more processors, the facility data using a trained machine learning model to determine a state and mode of the facility, the trained machine learning model being trained using historical facility data including a plurality of facility states and facility operating modes (250); and When the facility is in normal operation, the one or more processors use the trained machine learning model to determine whether an event in the facility includes an anomaly (270); wherein the facility data is grouped into clusters, each of the clusters relating to a specific facility state and a specific facility operating mode (715, 280); and Each of the groups includes descriptors associated with the specific facility status and the specific facility operating mode (717).

Images