Method for recommending operating commands for industrial plants

Abstract

The present invention relates to the field of controlling industrial plants, and more particularly to the field of controlling industrial plants for controlling an industrial plant by an operating command. The present invention discloses a computer-implemented method for recommending an operating command (32) capable of controlling at least one actuator (16) of an industrial plant (10). The method includes receiving an alarm (45) from at least one sensor and / or from an operator (20) of the plant (10), wherein the alarm (45) is related to the current state (40) of the plant (10), and obtaining the current state (40) of the plant (10), the current state (40) including at least one current process value related to the plant (10) and / or at least one current operating command (22), comparing the current state (40) of the plant (10) with a list of historical states (30) of the plant (10), each historical state (30) including a plurality of historical process values (34) related to the plant (10) and / or at least one historical operating command (32), and when the current state (40) matches at least one subset of the historical states (30), starting a simulation based on a simulation model (18) of the plant (10) and the matching historical state (30) as the starting state, performing a plurality of simulations, each simulation being based on at least one variation of the historical operating commands (32), for each simulation of the plurality of simulations, determining a quality value based on at least one quality criterion, and recommending a variation of the operating command (32) that resulted in the simulation with the highest quality value.

Description

Technical Field

[0001] The present invention relates to the field of controlling industrial plants, and more particularly to the field of controlling industrial plants for controlling an industrial plant by an operation command. The present invention further relates to program elements, computer-readable storage media, and uses.

Background Art

[0002] The operation of industrial plants is a very complex task for at least some plants. Well-trained operators often have a sufficient understanding of the behavior of the plant, which enables the operator to execute appropriate operation commands for controlling the plant in a number of situations. This may include critical and / or alarm situations of the plant. However, in at least some situations, recommendations for operation commands may be useful.

Summary of the Invention

[0003] It is an object of the present invention to provide a method for recommending an operation command. This object is achieved by the subject matter of the independent claims. Further embodiments will become apparent from the dependent claims and the following description.

[0004] One aspect relates to a computer-implemented method for recommending an operation command capable of controlling at least one actuator of an industrial plant, the method comprising receiving an alarm from at least one sensor and / or from an operator of the plant, wherein the alarm is related to the current state of the plant, acquiring the current state of the plant, the current state including at least one current process value related to the plant and / or at least one current operation command, comparing the current state of the plant with a list of the plant's historical states, where each historical state includes a plurality of historical process values related to the plant and / or at least one historical operation command, when the current state matches at least one subset of the historical states, starting a simulation based on a simulation model of the plant and the matching historical state as the starting state; performing a plurality of simulations, each simulation being based on at least one variation of the historical operation commands; for each simulation of the plurality of simulations, determining a quality value based on at least one quality criterion; and recommending a variation of the operation command that resulted in the simulation with the highest quality value.

[0005] An industrial plant may be configured to perform a process that may include several subprocesses that can be connected. The process may be a chemical process, a manufacturing process of any type of product, and / or another type of process. The plant or process may be affected and / or controlled, for example, by operation commands input by an operator. The operation commands may be capable of controlling at least one actuator of the industrial plant. The actuators of the plant may include control elements such as valves, heaters, motors, indicator elements, for example, alarm indicators (e.g., lamps and / or messages), and other elements. A sufficiently complex industrial plant can include a plurality of these elements. Controlling at least one actuator of the industrial plant can include direct control of the actuator and / or indirect control, for example, by changing a setpoint.

[0006] An alarm can be triggered and / or brought about from at least one sensor and / or from the plant operator. At least some of the sensors may be part of an actuator, for example, a sensor indicating the position of the actuator, for example, "the valve is half open". The alarm may be "related" to the current state of the plant, for example, when the alarm is triggered and / or received within the same time frame as the current state and / or immediately before this time frame.

[0007] The alarm may be or indicate any critical situation of the plant. For example, if the risk of the plant is represented between 0 and 9 with "0 = normal operation", or with "green / yellow / red" where "green = normal operation", a critical situation may be "not 0" or "not green". The alarm may be triggered or sent by at least one sensor, or may be the result of integrating and / or preprocessing process values from multiple sensors. Alternatively, the alarm may be received when the operator presses a predefined button. The alarm may be or include an alert and / or notification related to the plant and / or process.

[0008] When obtaining the current state of the plant, often not all current process values from all sensors can be obtained, rather, current process values from a predefined subset of the plant's sensors can be obtained. The predefined subset of sensors can include sensors located outside the plant and thus "related" to the plant, for example, a temperature sensor configured to measure the temperature of the external environment of the plant.

[0009] To compare the current state of a plant with a list of historical states of the plant, the historical states may have been previously acquired and / or may be available in a journal of the plant, such as a database. The list of historical states may have the form of a chronological sequence of historical states. Each historical state can include a plurality of historical process values associated with the plant and / or at least one historical operation command. The historical states can have a different number of process values and / or operation commands than an equivalent current state of the plant, for example, more process values and / or operation commands. This can be the case, in particular, for large or complex plants that can include thousands of sensors supplying process values. In one or more test runs, more powerful computers and / or more memory can be used when acquiring the historical states than during the production runs of the plant. The number of process current values and / or operation commands can be different for each historical state. The comparing step may, for example, compare fewer current process values and / or operation commands with a greater number of historical process values and / or operation commands. In at least some cases, the comparing step may consider only a subset of the historical process values and / or operation commands. Thus, the matching for at least one subset of the historical states can be a matching for all historical process values and / or operation commands, or a matching for an actual subset (i.e., fewer values). Thus, the matching can be a complete match or a partial match. The matching can include, for example, an exact matching for sensors having binary values and / or discrete values, or a kind of "similar" matching for ranges of sensor values.

[0010] The matching can result in zero hits or one or more hits. In the case of zero hits, the simulation is not started. In the case of one hit, the simulation can be started with all the historical process values and / or operation commands included in the matching historical state. In the case of two or more hits, the simulation may be started for each of the matching historical states. This can be a solution strategy for the operator's real problems. Since various parameters may affect the operator's decision, there may remain some uncertainty as to whether the current situation is actually equivalent to the situation on which the selected strategy was based. In this method, the strategy can include considering all the situations that may be related to the current state, thus the historical process values and / or operation commands, and recommending operation commands based on this broad criterion. The simulation method can include any simulation method suitable for the process of the plant.

[0011] Additionally or alternatively, the historical data of the plant control system can be used to extract operation strategies. The historical data can include the sensor information of the process and the events generated by the control system components, which enables capturing the plant state at different moments. Further, the historical data can include the so-called audit trails, which are events representing the operator's interaction with the plant. The audit trails of operator commands and plant state information may be used as a data basis for strategy extraction.

[0012] Based on the matching historical state, multiple simulations may be executed as the starting state. As one option, the plant simulator may be set to a state comparable to the situation of the historical interaction. As one option, the plant simulator may be set to a state comparable to the situation of the operator within the plant. This initialization using the starting state provides a valid representation of the current state and thus enables simulations to test some strategies that may be candidates for recommending operation commands. Each strategy can be created for one specific problem.

[0013] Multiple simulations may reflect or be based on a strategy for dealing with the current state, that is, for testing possible reactions to an alarm. One of the multiple simulations may use the historical operation commands as data inputs. This may be used to test the effectiveness and starting state of the simulator. Other simulations may include, for example, modifications of the operation commands, such as removing steps of the historical commands, changing the order, and / or additional ones, such as at least one modification of the historical operation commands. The changes may include, for example, changes in the duration of the operation commands and / or changes in the time, intensity, etc. of the actuator control, various set values for set value changes, and / or random removal of steps. For example, the module "Workflow Pertubator" may automatically create a modification (perturbation) of the workflow, such as removing an operator input / changing the input value. The perturbed workflow may be applied to a process simulation instance. Each of the simulations may end based on an end criterion, such as "reached a steady state" or "exceeded a predetermined number of simulation steps".

[0014] For each of the multiple simulations, a quality value may be determined based on at least one quality criterion (or "success criterion"). The quality criteria can include "reached a steady state", "did the situation of the plant at the end of the simulation deteriorate or improve?", "can the alarm be stopped?", comparison of important values to the values as a result of the historical operation commands, the duration of the time period to stop the alarm, the number of process steps to stop the alarm, returning to the normal state, alarm off, the process variable range reached for ideal production, and / or additional criteria. The criteria may be weighted and mapped to a quality value, and the quality value may be a numerical value for easier comparison.

[0015] Based on the quality value, recommendations can be given as to which changes led to the best results in the simulation run, more specifically: which operator actions, actuator activations, etc. lead to the optimal response to a given alarm. Furthermore, for example, "antirecommendations" can be given based on simulations that lead to bad behavior of the plant. The method can advantageously be used for training for the actual situation on the plant and / or as a basis for training an artificial neural network, ANN, for those situations.

[0016] In a further embodiment, the method further comprises the step of obtaining a list of historical states of the plant, each historical state including a plurality of historical process values and / or at least one historical operation command related to the plant, and the step of storing the list of historical states.

[0017] These steps may be performed as preparation for the steps described above. Additionally or alternatively, the steps may be performed continuously, for example, during operation of the plant. The list of historical states of the plant may be stored in a repository, such as a database and / or the cloud.

[0018] In various embodiments, at least one subset of the historical states is selected based on algorithms, heuristics, and / or ANN-based sensitivity analysis of a plurality of historical process values, and / or random selection from a plurality of historical process values and / or at least one current process value.

[0019] Algorithm-based sensitivity analysis can be based on, for example, variance decomposition, partial derivatives, or elementary effects. Heuristic sensitivity analysis can be based on, for example, the expertise of experienced service personnel, which can be collected in discussions and / or workshops. ANN-based sensitivity analysis can be based on multiple runs, and the criterion or "reward" can be, for example, an alarm. Generally, sensitivity analysis can be based on questions such as "which process values contribute most to an alarm or another event that is considered important for this plant?" Sensitivity analysis and / or random selection can advantageously contribute to reducing the number of values considered in a simulation. This can contribute to reducing the complexity of the system so that useful statements can be made within a reasonable time frame.

[0020] In various embodiments, comparing includes using a similarity function. The similarity function can be advantageously useful for comparing similar values. For example, for at least some sensors, a "similarity" range may be defined. This "similar" range can be, for example, a kind of "value tolerance range" that interprets historical process values within a 5% tolerance range, so that, as a result, "90 °C" may be interpreted as "between 85 °C and 95 °C".

[0021] In various embodiments, the matching of at least one subset of the current state and the historical state is based on a similarity analysis of the elements of the current state compared to the elements of the historical state. The similarity analysis can use, for example, the so-called Mueen's algorithm for similarity search as described at www.cs.unm.edu / ~mueen / FastestSimilaritySearch.html. Using similarity analysis can contribute to further improving the importance, validity, and / or relevant spectrum of the recommendations for variations of the operating commands.

[0022] In various embodiments, an alarm is at least one of a critical situation characterized by exceeding a predetermined threshold of at least one current process value and / or a current operating command, an alarm of the plant's alarm system, tripping of a predetermined actuator, and / or startup of a subsystem. Even before exceeding the threshold or before an alarm is generated, a process variable may exhibit abnormal and, in the long term, problematic behavior. Tripping can include stopping of at least one device and / or interruption of an operating process.

[0023] In some embodiments, comparing the current state of a plant to a list of historical states includes comparing one state or a sequence of states. "One state" can be interpreted as comparing one current state to each of the historical states. "Sequence of states" can be interpreted as comparing a sequence of current states, i.e., an ordered list of temporally consecutive states, to a sequence of historical states. The sequence can include two, three, or more states, i.e., the same number of current and historical states.

[0024] In various embodiments, each of a plurality of simulations is based on variations in waiting times, variations in setpoint ranges, and / or changes in order, random removal of steps. This can include, for example, providing the simulator with operator action strategies and / or operator workflows in a previously activated state, taking into account waiting times, setpoint ranges, etc., which are part of the strategy and recording the effects. Additionally or alternatively, the operator action strategy / operator workflow can be associated with goals, e.g., alarm off, efficiency represented by a process variable, etc., providing the simulation results to the simulator in each state, and evaluating whether the effects shown from the simulation conform to the expected values.

[0025] In some embodiments, each of the simulations ends when it reaches one of the following criteria: when the simulated process value falls below a predetermined threshold, when a configured alarm of the plant's alarm system is stopped, and / or when a composite signal for normal plant operation is set.

[0026] The predetermined threshold may characterize a "normal" or "non-critical state", i.e., the values of a subset of the relevant process variables do not exceed the threshold and / or do not return to the normal state. The configured alarm may be an alarm that is set and reset or stopped by the plant's alarm system, for example, by aggregating simulated, current, or historical process values. The configured signal for normal plant operation may include the values of a subset of the relevant process variables suitable for optimal plant operation.

[0027] In some embodiments, the quality criteria for determining the quality value comprise at least one of the duration of the time period for stopping the alarm, the number of process steps for stopping the alarm, the number of actuators involved, and / or a predetermined operating range of the process variable.

[0028] One aspect relates to a program element for performing the above and / or the following methods when executing the program element on a processing unit and / or a control unit of an industrial plant.

[0029] One aspect relates to a processing unit and / or a control unit of an industrial plant configured to perform the methods described above and / or below.

[0030] One aspect relates to a computer-readable storage medium storing the computer program or computer program product described above.

[0031] One aspect relates to a control unit configured to control an industrial plant by using the above and / or the following methods and / or to recommend operating commands so as to execute the methods and / or to apply the recommended operating commands.

[0032] For the sake of clarity, the present invention is illustrated by the embodiments shown in the figures. These embodiments should be considered by way of example only and not by way of limitation.

Brief Description of the Drawings

[0033] The drawings are as follows.

Figure 1

Figure 2

Mode for Carrying Out the Invention

[0034] FIG. 1 schematically illustrates an arrangement according to an embodiment. The arrangement comprises an industrial plant 10 that can be configured to execute a process. The process can include several sub-processes that can be connected. The process can be a chemical process, a manufacturing process of any type of goods, and / or another type of process. The plant 10 or the process can be affected and / or controlled, for example, by an operation command 22 input by an operator 20. The operation command can be capable of controlling at least one actuator 16 of the industrial plant. The actuator 16 of the plant 10 can comprise control elements such as valves, heaters, motors, indicator elements, for example, alarm indicators (e.g., lamps and / or messages), and other elements. A simulation model 18 is available for the plant 10. The method used in the simulation 18 can include any simulation method suitable for the process of the plant. Process values related to the plant 10, for example, values from internal or external sensors 14 of the plant 10 that can affect the behavior of the plant, may be monitored and stored as historical process values 34. Also, the operation command 22 may be monitored and stored as a historical operation command 32. A plurality of historical process values 34 and / or historical operation commands 32 may be stored as a list of historical states 30, and each entry in this list includes historical process values 34 and / or historical operation commands 32 at different instants. The list of historical states 30 can have the form of a time sequence of historical states. The list of historical states 30 can be stored in a repository such as a database located in the cloud.

[0035] Furthermore, the current state 40 of the plant 10 is monitored, in particular continuously. The current state 40 includes at least one current process value and / or at least one current operating command 22 related to the plant 10. When a critical situation occurs, for example, a combination of the current process value and / or the current operating command 22, the processing unit and / or the control unit 12 of the plant 10 may receive an alarm 45. The alarm 45 may be related to the current state 40 of the plant 10 and may indicate, for example, a critical situation. Next, the current state 40 of the plant 10 is acquired. The current state 40 can include at least one current process value and / or at least one current operating command 22 related to the plant 10. The current state 40 may be compared with a list of historical states 30. Depending on the result of the comparison, one or more simulations may be executed in the simulator 18. As a result of the simulation, the processing unit and / or the control unit 12 can output a recommendation for an operating command 50. Additionally or alternatively, the processing unit and / or the control unit 12 may output an operating command (not shown) based on the recommendation.

[0036] Figure 2 shows a flow diagram 100 according to one embodiment. In an optional step 102, a list of historical states 30 of the plant 10 is obtained. Each entry in this list, i.e., each historical state 30, can include a plurality of historical process values 34 and / or at least one historical operation command 32 related to the plant 10. In an optional step 104, the list of historical states 30 is stored. In step 106, an alarm 45 is received from at least one sensor 14 of the plant 10 and / or from the operator 20, and the alarm 45 is related to the current state 40 of the plant 10. In step 108, the current state 40 of the plant 10 is obtained, and the current state 40 includes at least one current process value and / or at least one current operation command 22 related to the plant 10. In step 110, the current state 40 of the plant 10 is compared with the list of historical states 30 of the plant 10, and each historical state 30 includes a plurality of historical process values 34 and / or at least one historical operation command 32 related to the plant 10. If the current state 40 does not match at least one subset of the historical states 30, in step 112, the method ends.

[0037] If the current state 40 matches at least one subset of the historical states 30, in step 114, a simulation is started based on the simulation model 18 of the plant 10 and the matching historical state 30 as the starting state. In step 116, a plurality of simulations are executed, and each simulation is based on at least one variation of the historical operation commands 32. In step 118, for each simulation of the plurality of simulations, a quality value is determined based on at least one quality criterion. In step 120, a change in the operation command 32 is recommended so that a simulation with the highest quality value is obtained.

[0038] List of reference signs 10 Industrial plant 12 Processing unit and / or control unit 14 Sensor 16 Actuator 18 Simulation Model, Simulator 20 Operator 22 Operation Command 30 History State 32 History Operation Command 34 History Process Value 40 Current State 45 Alarm 50 Operation Command and / or Recommendation 100 Flow Chart 102 - 120 Steps

Claims

1. A computer-implemented method for recommending an operating command (32) capable of controlling at least one actuator (16) of an industrial plant (10), comprising: receiving an alarm (45) from at least one sensor (14) and / or an operator (20) of the plant (10), wherein the alarm (45) is related to the current state (40) of the plant (10); obtaining the current state (40) of the plant (10), the current state (40) including at least one current process value and / or at least one current operating command (22) related to the plant (10); comparing the current state (40) of the plant (10) with a list of historical states (30) of the plant (10), each historical state (30) including a plurality of historical process values (34) and / or at least one historical operating command (32) related to the plant (10); when the current state (40) matches at least one subset of the historical states (30), starting a simulation based on the simulation model (18) of the plant (10) and the matching historical state (30) as the starting state; executing a plurality of simulations, each simulation being based on at least one variation of the historical operating commands (32); for each simulation of the plurality of simulations, determining a quality value based on at least one quality criterion; and recommending a variation of the operating command (32) that resulted in the simulation with the highest quality value.

2. The method according to claim 1, further comprising obtaining the list of historical states (30) of the plant (10), each historical state (30) comprising a plurality of historical process values (34) and / or at least one historical operating command (32) related to the plant (10), and storing the list of historical states (30).

3. ​ At least one of the subsets of the history state (30) is selected based on algorithmic, heuristic, and / or ANN-based sensitivity analysis of the plurality of historical process values (34), and / or random selection from the plurality of historical process values (34) and / or the at least one current process value, according to the method of claim 1 or 2.

4. The method according to any one of claims 1 to 3, wherein the comparing includes using a similarity function.

5. The method according to any one of claims 1 to 4, wherein matching the current state (40) with at least one subset of the history states (30) is based on a similarity analysis of the elements of the current state (40) compared to the elements of the history state (30).

6. The alarm (45) is exceeding a predetermined threshold of at least one current process value and / or a critical situation characterized by the current operating command (22), an alarm of the alarm system of the plant, tripping of a predetermined actuator, starting of a subsystem, and / or stopping of a subsystem, according to the method of any one of claims 1 to 5.

7. The method according to any one of claims 1 to 6, wherein comparing the current state (40) of the plant (10) with a list of history states (30) includes comparing one state or a series of states.

8. Each of the plurality of simulations is based on variations in waiting time, variations in setpoint ranges, and / or random removal of steps, according to the method of any one of claims 1 to 7.

9. Each of the simulations ends when one of the following criteria is met: when the simulated process value falls below a predetermined threshold; when a configured alarm of the alarm system of the plant is stopped and / or when a composite signal for normal plant operation is set, according to the method of any one of claims 1 to 8.

10. The quality criteria for determining the quality value are the duration of the time period for stopping the alarm (45), the number of processing steps for stopping the alarm (45), the number of actuators involved, and / or A predefined operating range for a process variable The method according to any one of claims 1 to 9, comprising at least one of the above.

11. A processing unit and / or a control unit (12) of an industrial plant (10) configured to execute the method according to any one of claims 1 to 10.

12. A program element for executing the method according to any one of claims 1 to 10 when executing a program element on a processing unit and / or a control unit (12) of an industrial plant (10).

13. A computer-readable storage medium storing the computer program element according to claim 12.

14. Use of the method according to any one of claims 1 to 10, and / or for recommending an operating command (32) and / or for controlling an industrial plant (10) by applying the recommended operating command (32), a control unit configured to execute the method.

Images