Method and system for simulating module operations of modular industrial plants

Abstract

The invention proposes a computer-implemented method for simulating the operation of a module of a modular industrial plant, the method comprising: providing an initial model for the module based on a simulation, such as a white-box simulation or a black-box simulation or a grey-box simulation; performing a knowledge-based enhancement step or a data-driven enhancement step including obtaining an enhancement model; and simulating a module operation of the module by means of the enhanced model.

Description

Technical Field

[0001] The present invention relates to a method and a system for simulating modular operation of modules of a modular industrial plant. Background Art

[0002] The present invention relates to modular automation systems and modular factories. The goal of a modular factory is to use prefabricated, well-tested and well-understood modules, also called process equipment assemblies (PEAs), which can be easily put together in different combinations so that different configurations (recipes) can be achieved.

[0003] Modules are typically defined and created individually and then assembled into an entire factory. A standardized Module Type Package (MTP) approach can create a framework for interoperability between modules and orchestration systems.

[0004] There are different applications and use case scenarios, such as process plant virtual commissioning (VIBN) or process plant validation, which, for example, can be performed virtually before the actual installation of modules and deployment of the entire plant (which can include several modules connected together by piping) for reasons of time and cost efficiency. For these cases, module simulation can bring great benefits.

[0005] Therefore, the goal is to be able to simulate modules individually and / or in the context of an entire plant, for example in order to ensure safe and efficient operation.Current simulations have drawbacks, leading to a need to improve the quality of simulations.

[0006] It is therefore an object of the present invention to provide a method and system that addresses at least some of the above challenges, in particular provides an improved simulation quality, and in particular thereby ensures safe and efficient operation. Summary of the Invention

[0007] At least some of these challenges are addressed by the subject matter of the present disclosure. The present invention provides a method, a system, a computer-readable medium and a computer program product according to the independent claims. Preferred embodiments are given in the dependent claims.

[0008] The present invention provides a computer-implemented method for simulating the modular operation of modules of a modular industrial plant, the method comprising providing an initial model for the module based on simulation, particularly white-box simulation or black-box simulation, or a combination of both including sub-parts of white-box simulation and sub-parts of black-box simulation (referred to herein as gray-box simulation), to generate an overall module simulation. The method further comprises performing a knowledge-based enhancement step, the enhancement step comprising obtaining an enhanced model by performing a knowledge-based modification of the initial model and / or a knowledge-based extension of the initial model. The knowledge-based modification and / or the knowledge-based extension each model domain knowledge and / or implement domain rules. The method further comprises simulating the modular operation of the module by enhancing the model.

[0009] The present invention provides another computer-implemented method for simulating modular operation of a module of a modular industrial plant. The method comprises providing an initial model for the module based on a simulation, particularly a white-box simulation or a black-box simulation. The method further comprises performing a data-driven enhancement step, wherein the enhancement step comprises obtaining an enhanced model by performing a data-driven modification of the initial model and / or a data-driven extension of the initial model. The data-driven modification and / or the data-driven extension are each based on measurement data obtained for the module in operation, the module being installed as part of the modular industrial plant. The method further comprises simulating modular operation of the module using the enhanced model.

[0010] The present invention provides another computer-implemented method for simulating a module of a modular industrial plant, the method comprising providing an initial model for the module based on a simulation, in particular a white-box simulation or a black-box simulation. The method further comprises performing a knowledge-based enhancement step, the knowledge-based enhancement step comprising obtaining a first enhanced model by performing a knowledge-based modification of the initial model and / or a knowledge-based extension of the initial model. The knowledge-based modification and / or the knowledge-based extension each models domain knowledge and / or implements domain rules. The method further comprises performing a data-driven enhancement step, the data-driven enhancement step comprising obtaining a second enhanced model by performing a data-driven modification of the first enhanced model. The data-driven modification and / or the data-driven extension each are based on measurement data obtained for the module in operation, the module being installed as part of the modular industrial plant. The method further comprises simulating the module operation of the module by means of the second enhanced model.

[0011] This approach combines the approaches of the first and second approaches by first applying knowledge-based augmentation and then data-driven augmentation (e.g., data assimilation). For example, the initial simulation can be augmented to perform bias correction and / or edge case correction, and then data assimilation can be performed for calibration. Data assimilation can directly or indirectly involve the first augmentation model by modifying and / or improving / adapting the rules.

[0012] In the present disclosure, module operations of an analog module may refer to operations of an island module.

[0013] Modules can be, for example, heating modules (e.g. for heating oil in refineries, or for power generation), mixing modules (for mixing fluids, taking into account sedimentation, etc.), reaction modules (for chemical reactions), separation modules (e.g. for three-phase separation of gas, water, oil), etc.

[0014] A modular industrial plant can be any industrial plant comprising a plurality of modules that may be different. The modules of the industrial plant can be part of a chain of modules or a module topology. Thus, plant operations can include the operation of one or more modules of the plurality of modules (e.g., at least partially interdependent).

[0015] In this disclosure, the term "white box simulation" may include hard-coded models and / or simulation algorithms, which may be based, inter alia, on physical and / or engineering rules.

[0016] In this disclosure, the term "black box simulation" may include machine learning-based module simulation. Black box simulation may include (real) data-driven and AI-enhanced simulation. For example, machine learning-based module simulation functions, such as neural networks that simulate the behavior of real systems and are trained with real data.

[0017] In the present disclosure, the term "grey-box simulation" may include a combination of white-box and black-box simulations, i.e. a simulation that is partly based on clearly understandable program code, explicitly and manually programmed / created by, e.g., the module or MTP manufacturer, and based on, e.g., C&E simulations, etc., in combination with a part based on (black-box) NN or ML&DA algorithms, which are not completely transparent to the user, but are trained on real application data and are therefore not completely white-box, but which are still used to approximate some (partial) functionalities or parts of the overall simulation. Example: a module simulation of a reactor, where a certain part is based on C&E simulations or has several if-else loops, which is clearly white-box, and another part is based on data-driven, ML-based function approximation, which describes a sub-part of the overall module simulation, e.g., only the heating behavior inside the reactor, but which is black-box (meaning, data-driven, learned by training, not e.g., hard-coded if-else loops, etc.).

[0018] According to the present disclosure, obtaining an enhanced model may include modifying or extending the model in a manner that improves accuracy. Modifying the model may include limiting the range of variables and / or parameters, calibrating simulation model parameters, updating and / or improving functional dependencies based on data-based correlations or patterns, etc. Extending the model may include, for example, extending functional dependencies with additional physical and / or engineering quantities, and / or considering additional engineering rules that describe or constrain certain behaviors, and / or rules that handle specific physical or engineering constraints, etc.

[0019] Performing a knowledge-based modification of the model may include modifying the model in a manner that takes the knowledge into account.Performing a knowledge-based extension of the model may include extending the model in a manner that takes the knowledge into account.

[0020] The knowledge-based modification and / or knowledge-based extension each models domain knowledge and / or implements domain rules. The domain knowledge and domain rules may include knowledge / rules related to industrial process automation and / or engineering domain knowledge. The domain knowledge and domain rules may be provided, for example, by an operator.

[0021] Domain knowledge can refer to typical engineering education-based configurations of modules, typical (physics-based) correlations between parameters, typical co-occurrences of settings or parameters, common or standard settings / configurations, best practices in engineering, established and proven appropriate ranges for parameters or variables, etc.

[0022] For example, domain knowledge may include information about nonlinear (functional) behavior and / or interrelationships and / or correlations between multiple variables and / or edge cases, etc. In particular, knowledge may include information that is not reflected in the stored and / or collected data (e.g., used to train the ML algorithm) used to build the initial model and / or data (e.g., parameters). This is typically not considered for modeling. In particular, domain knowledge may also include information that can constitute non-data-centric ML-based models, i.e., models that have been trained on, for example, unintentionally biased or poorly chosen training data.

[0023] In the present disclosure, knowledge-based extensions or modifications may require rule-based extensions or modifications. For example, rules may be provided to address potential bias due to biased collected data used for white-box simulations, and / or rules may be provided to address a lack of information about edge cases in the collected data or from database data used for white-box simulations.

[0024] Knowledge can also be module-specific and involve knowledge about nonlinear (functional) behavior, interrelationships and / or correlations between multiple variables that are not reflected in the collected data or are not typically considered for modeling because they are not in the database storing the data used to obtain the initial model.

[0025] The term "data driven" may refer to being driven by measurement data obtained for modules in operation, the modules being installed as part of a modular industrial plant.

[0026] Performing data-driven modification of the model may include modifying the model in a manner that takes into account measurement data obtained for a module in operation, the module being installed as part of the modular industrial plant.

[0027] Performing a data-driven extension of the model may include extending the model in such a way as to take into account measurement data obtained for operating modules installed as part of the modular industrial plant.

[0028] Data-driven enhancement can include data assimilation, such as optimization. For example, data assimilation can be performed by improving or calibrating an optimization engine that uses a black-box or white-box simulation setup (e.g., parameters and relationships). This can be done by assimilating the simulated behavior of the simulation scenario to real data, such as by calibrating parameters or correlations.

[0029] As can be understood from the claims and the foregoing discussion, the independent method claims each relate to a method for simulating modular operation of a module of a modular industrial plant. For each independent method claim, the method includes providing an initial model for the module based on white-box simulation or black-box simulation, performing one or more enhancement steps, each enhancement step including obtaining an enhanced model by performing a modification of the initial model and / or an extension of the initial model, and simulating modular operation of the module using the enhanced model.

[0030] Claim 1 proposes a knowledge-based enhancement step, claim 2 proposes a data-driven enhancement step, and claim 3 proposes two enhancement steps, namely a knowledge-based enhancement step and a data-driven enhancement step. Accordingly, the above claims 1 to 3 can also be rewritten as a single independent claim:

[0031] A computer-implemented method for simulating module operations of modules of a modular industrial plant, the method comprising:

[0032] providing an initial model for the module based on simulation, particularly white box simulation or black box simulation (S11);

[0033] performing a knowledge-based enhancement step (S12), comprising obtaining a first enhanced model by performing a knowledge-based modification of the initial model and / or a knowledge-based extension of the initial model, wherein the knowledge-based modification and / or the knowledge-based extension each model domain knowledge and / or implement domain rules; and / or

[0034] performing a data-driven enhancement step, comprising obtaining (S13) a second enhanced model by performing a data-driven modification of the initial model or the first enhanced model and / or by performing a data-driven extension of the initial model or the first enhanced model, wherein the data-driven modification and / or the data-driven extension are each based on measurement data obtained for a module in operation, the module being installed as part of the modular industrial plant; and

[0035] The module operation of the module is simulated (S14) by means of the first enhanced model or the second enhanced model.

[0036] It should be noted that claims 1 to 3 have been drafted as independent claims. As can be understood from the above, they provide alternative but unitary technical solutions. To ensure readability, particularly of the dependent claims, it is most appropriate to cover these alternative solutions not through a single claim but through multiple separate claims.

[0037] As can be seen from the above, the present disclosure provides a method for providing module simulation capabilities, for example, to a plant owner.

[0038] As an example, a base (initial) model can be used that utilizes modeling-related information derived from process and automation engineering sources (DEXPI, MTP, etc.), i.e., a white-box approach, and / or utilizes out-of-the-box, data-driven ML-based mechanisms to learn the model / simulation settings, i.e., a black-box approach. This base / initial model is enhanced with knowledge-based and / or data assimilation and / or optimization mechanisms, thereby overcoming the typical shortcomings of the base approach and thereby allowing correct / accurate and low-cost simulation of module functions.

[0039] As will be outlined in detail below, additionally, an optional simulation-based optimization component may be added, for example to facilitate in-the-loop simulations and thereby provide dataflow-driven simulation-based optimization recommendations for control parameters.

[0040] Furthermore, as also outlined in detail below, the method may rely on data acquisition, for example, performed by data acquisition hardware such as sensors / measurement devices.

[0041] As mentioned above, in order to obtain the module simulation function, for the initial model, white box simulation or black box simulation can be used.

[0042] As an example, white-box simulation functionality may be provided by dedicated white-box simulation program code that is manually programmed / created by the module or MTP manufacturer, based on, for example, C&E simulations, and may be derived from and executed in accordance with information provided in the MTP or other sources (e.g., from process engineering, such as DEXPI). As another example, black-box simulation functionality may be provided by a real-data-driven and AI-enhanced black-box simulation that is not manually programmed (e.g., by the module manufacturer) but rather simulates (i.e., approximates) the behavior of the module based on a neural network or ML algorithm that has been trained on real training data.

[0043] The present disclosure also provides for enhancing simulations by at least one of a knowledge-based approach and a data-driven (e.g., data assimilation) approach. A knowledge-based approach can be provided that models and implements domain knowledge, such as domain knowledge of industrial process automation and engineering. Thus, it can handle, for example, simulations that do not violate basic domain rules. Thus, it can also ensure the interpretability of the underlying algorithms and improve transparency and credibility. A data-driven approach can be provided, for example, in the form of a data assimilation (optimization) engine that improves or calibrates (black box or white box / hard coded) simulation parameters based on assimilating the simulated simulation scenario behavior to real scenario data.

[0044] For example, to achieve high accuracy, white-box methods may require significant manual effort and high domain expertise to correctly model module behavior. When left out of the box, black-box methods are often not fully properly calibrated, i.e., they may not cover all possible cases (as some error cases may rarely occur in real factories) and often do not represent the actual behavior in a fine-grained manner.

[0045] The enhancement steps may address these and other challenges, such as missing or incompletely correct representation / simulation of simulated behavior when compared to real behavior, or partially incomplete simulation models along with incomplete consideration of control parameters that affect the overall simulated behavior.

[0046] According to the present disclosure, as part of a knowledge-based enhancement step, rule-based bias and / or edge case correction may be performed, and subsequently, as part of a data-driven enhancement step, data assimilation may be performed for calibration of a first enhancement model and / or calibration of one or more rules for rule-based bias and / or edge case correction.

[0047] Thus, particularly accurate models can be obtained in a structured and efficient manner.

[0048] According to the present disclosure, the knowledge-based enhancement step can be based on module-specific knowledge, which module-specific knowledge includes at least one of the following: information about correlations between multiple variables, information about linear or nonlinear functional behavior, information about interrelationships, information about edge cases, information about biases in the input data, and information about potentially relevant additional simulation parameters.

[0049] As a result, particularly precise and comprehensive simulations can be obtained.

[0050] Module-specific knowledge can refer to knowledge that is specific to a module type or specific to a particular module.

[0051] Information about nonlinear (functional) behavior may include information indicating that a parameter / variable representing an operational characteristic of the module has a nonlinear dependency on other parameters / variables for at least a portion of the range of values ​​of the other parameters / variables (e.g., at values ​​above or below certain thresholds).

[0052] The information about the correlation between multiple variables may, for example, include information about the correlation between parameters / variables representing the operational characteristics of the module and other parameters / variables.Such information may include, for example, an indication of linear and / or nonlinear (functional) behavior.

[0053] The information about the correlation can include information about the influence of the first parameter / variable on the correlation of the second parameter / variable with the third parameter / variable. The information about the correlation can include information about, for example, nonlinear (functional) behavior of the second parameter / variable caused by the first parameter / variable. For example, the pressure dependence of the second variable can be affected by temperature, etc.

[0054] Information about edge cases may refer to information relating to rare scenarios, i.e., scenarios that are unlikely to occur. Information about edge cases may include information describing the edge cases and, optionally, information about how the edge cases are handled in the simulation / model. Such edge cases and / or information about them may be provided and identified by a user, such as an operator.

[0055] Bias in the input data can result from non-central methods used for data collection or from incomplete data collection, etc. Information regarding bias in the input data can include an indication that the data is biased and, optionally, information regarding how the data is biased and / or how the bias is handled in the simulation / model. Such bias and / or information regarding the bias can be provided and identified by a user (e.g., an operator).

[0056] The information about potentially relevant additional simulation parameters may refer to parameters that were not initially included in the simulation, for example, due to assumed low relevance. The information about potentially relevant additional simulation parameters may include an indication of the potentially relevant additional simulation parameters and, optionally, information about how the potentially relevant additional simulation parameters were considered in the simulation / model. Such parameters and / or information thereon may be provided and identified by a user (e.g., an operator).

[0057] According to the present disclosure, modeling of domain knowledge may include modeling module-specific knowledge and / or wherein enforcement of domain rules may include enforcing rules that implement the module-specific knowledge.

[0058] In this context, enforcing domain rules can mean ensuring that certain values ​​are always kept above or below certain allowed thresholds, or that specific hazard and operability (HAZOP) guidelines are followed, or that best practices are considered / accounted for and / or not ignored or forgotten, or that common mutual engineering exclusion criteria are considered and not violated, etc.

[0059] According to the present disclosure, the data-driven enhancement step may include performing data-driven modification and / or expansion including data assimilation configured to minimize the difference between simulated output data and actual measured data, such as employing a Kalman filter.

[0060] In other words, the data-driven enhancement step may include modifying and / or extending the model so that its predictions about module operation more closely match measurements representative of module operation.

[0061] According to the present disclosure, the data-driven enhancement step may include performing data-driven modifications and / or extensions, which include, for example, module-specific calibration of the initial model, in particular the initial model simulation parameters and / or initial model relationships, after the module is installed.

[0062] The present disclosure also provides a method for determining modular plant control parameters, the method comprising, for example, a method for simulating module operation of modules of a modular industrial plant of the present disclosure as outlined above. The method for determining modular plant control parameters further comprises modeling interrelationships between the modules of the modular industrial plant, and performing an optimization step based on the simulated module operation of the modules and the modeled interrelationships to provide modular plant control parameters. The plant control parameters may be any parameters suitable for controlling plant operation, such as operating parameter values ​​of the plant and / or its components.

[0063] According to the present disclosure, the optimization step may include data-driven optimization based on simulation data and / or measurement data obtained for modular industrial plant operations.

[0064] Therefore, an (optional) optimization component for in-loop simulations can be provided. This allows providing control parameter optimization suggestions based on data flow driven simulations.

[0065] The optimization can be performed on a module chain or topology comprising multiple modules, e.g. from the same or different manufacturers and / or the same or different types, in parallel and / or in series. The optimization allows modeling of interrelationships between modules that are not reflected in individual models, in particular in database data.

[0066] For example, individual modules and entire module-to-module processing chains or pipelines can be considered for optimization, such as by providing inter-module calibration. Thus, for example, bottlenecks or module downtime (e.g., due to a module being located at a certain position in the module chain) can be prevented. Furthermore, for multiple modules, such as two or more parallel modules, it can be determined how to operate the modules so that excess power from one module can be used to power other modules. Thus, the performance of the entire process plant is improved.

[0067] The present disclosure also provides a system comprising a computing system configured to execute a method for simulating module operations of modules of the modular industrial plant of the present disclosure, for example, as outlined above.

[0068] The system according to the present disclosure may also include one or more sensors / measuring devices that are configured to obtain measurement data of modules in operation installed as part of the modular industrial plant and / or measurement data for the operation of the modular industrial plant.

[0069] The system according to the invention may also comprise modules of a modular industrial plant.

[0070] The proposed overall system for module simulation may comprise: software components for virtually executing the method steps; and one or more hardware components, e.g. for sensing and / or monitoring and / or encoding and / or processing, to facilitate the data acquisition, processing and calculation parts.

[0071] For example, computer-implemented methods may utilize complementary data acquisition hardware such as sensors / measurement devices.

[0072] A computing system may include one or more computing devices and, in particular, may be a distributed computing system.

[0073] The present disclosure also provides a computer program product comprising instructions, which, when executed by a computing system, cause the computing system to perform any method of the present disclosure.

[0074] The present invention also provides a computer-readable medium comprising instructions that, when executed by a computing system, cause the computing system to perform any of the methods of the present disclosure.

[0075] The features and advantages outlined above in the context of the method similarly apply to the system, computer program product, and computer-readable medium herein.

[0076] Other features, examples, and advantages will become apparent from the detailed description which proceeds with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In the accompanying drawings,

[0078] Figure 1 is a flow chart schematically illustrating a first method according to the present disclosure;

[0079] Figure 2 is a flow chart schematically illustrating a second method according to the present disclosure;

[0080] Figure 3 is a flow chart schematically illustrating a third method according to the present disclosure;

[0081] Figure 4 is a schematic diagram of a system according to the present invention; and

[0082] Figures 5a to 5c Systems and methods according to the present disclosure are shown. DETAILED DESCRIPTION

[0083] Figure 1 is a flow chart schematically illustrating a computer-implemented method for simulating module operations of a module of a modular industrial plant. The method comprises a step S11 of providing an initial model for the module based on white-box simulation or black-box simulation.

[0084] The method includes a step S12 of performing a knowledge-based enhancement step, including obtaining an enhanced model by performing a knowledge-based modification of an initial model and / or a knowledge-based extension of the initial model. The knowledge-based modification and / or the knowledge-based extension each model domain knowledge and / or implement domain rules. The method also includes a step S14 of simulating module operations of the module using the enhanced model.

[0085] Figure 2 is a flow chart schematically illustrating a computer-implemented method for simulating module operations of a module of a modular industrial plant. The method comprises step S11 of providing an initial model for the module, for example based on white box simulation, black box simulation or grey box simulation.

[0086] The method includes a step S13 of performing a data-driven enhancement step, including obtaining an enhanced model by performing a data-driven modification of an initial model and / or a data-driven extension of the initial model. The data-driven modification and / or the data-driven extension are each based on measurement data obtained for an operating module, the module being installed as part of a modular industrial plant. The method also includes a step S14 of simulating module operation of the module using the enhanced model.

[0087] Figure 3 is a flow chart schematically illustrating a computer-implemented method for simulating module operations of a module of a modular industrial plant. The method comprises a step S11 of providing an initial model for the module based on white-box simulation or black-box simulation.

[0088] The method comprises a step S12 of performing a knowledge-based enhancement step, comprising obtaining a first enhanced model by performing a knowledge-based modification of the initial model and / or a knowledge-based extension of the initial model. The knowledge-based modification and / or the knowledge-based extension each model domain knowledge and / or implement domain rules.

[0089] The method comprises a step S13 of performing a data-driven enhancement step, comprising obtaining a second enhanced model by performing a data-driven modification of the first enhanced model and / or a data-driven extension of the first enhanced model, wherein the data-driven modification and / or the data-driven extension are each based on measurement data obtained for an operating module installed as part of the modular industrial plant.

[0090] As an example, as part of a knowledge-based enhancement step, rule-based bias and / or edge case correction is performed, and subsequently, as part of a data-driven enhancement step, data assimilation is performed for calibration of the first enhancement model and / or calibration of one or more rules for rule-based bias and / or edge case correction.

[0091] The method further comprises step S14 of simulating module operations of the module through a second enhanced model.

[0092] In all of the above methods comprising a knowledge-based enhancement step, the knowledge-based enhancement step may be based on module-specific knowledge, the module-specific knowledge comprising at least one of: information about correlations between multiple variables, information about nonlinear (functional) behavior, information about interrelationships, information about edge cases, information about biases in input data, information about potentially relevant additional simulation parameters. Modeling of domain knowledge may comprise modeling the module-specific knowledge and / or wherein enforcing domain rules comprises enforcing rules that implement the module-specific knowledge.

[0093] In all of the above methods comprising a data-driven enhancement step, the data-driven enhancement step may comprise performing a data-driven modification comprising data assimilation configured to minimize the difference between the simulated output data and the actual measured data, for example using a Kalman filter. Alternatively or additionally, the data-driven enhancement step comprises performing a data-driven modification comprising a module-specific calibration of the initial model, in particular the initial model simulation parameters and / or the initial model relationships, after installation of the module.

[0094] Figures 1 to 3 A computer-implemented method for determining modular plant control parameters according to the present disclosure is also schematically illustrated by optional steps S15 and S16, the method comprising a method for module operation of modules for simulating a modular industrial plant according to the present disclosure, including, for example, steps S11, S12 and / or S13 and S14 as described above.

[0095] The method for determining modular plant control parameters includes: in step S15, modeling the interrelationships between modules of the modular industrial plant; and in step S16, performing an optimization step to provide modular plant control parameters based on the simulated module operations and the modeled interrelationships of the modules. The optimization step may include data-driven optimization based on simulation data and / or measurement data obtained for the modular industrial plant operation.

[0096] It should be understood that at least some of the above steps are optional and may be omitted.

[0097] Figure 4 A system 1 is shown comprising a computing system 2 configured to perform a method according to the present disclosure, e.g. Figures 1 to 3 The method is outlined in the context of the invention. The system in this example optionally further comprises a sensor 3, and may optionally comprise a plurality of sensors, the / each sensor being configured to obtain measurement data of an operating module installed as part of a modular industrial plant and / or measurement data of the operating modular industrial plant. The system in this example optionally further comprises a module 4 of a modular industrial plant 5.

[0098] The following will refer to Figures 5a to 5c Some other aspects and examples of the invention are outlined.

[0099] The systems and methods of the present disclosure, and in particular the systems and methods outlined below, enable and facilitate better (i.e., realistically modeled / simulated, accurate, perceptible, well-calibrated, easy to set up) module simulation capabilities for different scenarios in the fields of modular industrial automation and process engineering.

[0100] In this example, a method and system are provided that provide a first step of a coarse module simulation ( FIG. 5A ), where a coarse (also called initial) simulation is one that is not accurate enough for, for example, tuning parameters of a PID controller, but is suitable for, for example, testing an interlock.

[0101] In the first step, at least one of two main basic approaches can be adopted, which can provide a rough module simulation (out of the box):

[0102] 1. Dedicated white-box module simulations manually programmed / created by the module or MTP manufacturer, which are based on, for example, C&E simulations and can be derived and executed from information provided in the MTP or other sources (e.g., from process engineering, e.g., DEXPI).

[0103] 2. Real data-driven and AI-enhanced black-box module simulations, which are not manually programmed by the module manufacturer, but are based on what a neural network or ML algorithm learns based on provided real training data (e.g., from existing installations of the corresponding module / PEA) to simulate the module’s behavior.

[0104] The first approach requires significant manual effort and high domain expertise to adequately model the module behavior, for example, by deriving the basic structure / setup of the model and simulation from the content of MTP or other typical sources for the process industry (DEXPI, FM, etc.). This approach can produce (if significant effort is put in by domain and simulation experts) good models based on physical / chemical equations with default parameters, however, these parameters may not be appropriately calibrated, for example.

[0105] The second approach, when used out of the box, may not have been properly / completely calibrated and, in particular, may not necessarily cover / simulate all possible cases (as some error cases may rarely occur in a real plant) and typically does not represent the actual behavior at a fine-grained level. This is mainly due to the lack of underlying physical / chemical behavior models and due to possible suboptimal training data (e.g., it may not represent the complete picture and the complete behavior, but only a part of the actual overall behavior or only the behavior during a limited sub-period in time).

[0106] Therefore, both approaches require great efforts and costs, or bring difficulties, challenges and obstacles.

[0107] Difficulties and obstacles lie, for example, in the absence or incompletely correct representation / simulation of the simulated behavior when compared to the real behavior, or in partially incomplete simulation models together with incomplete consideration of control parameters affecting the overall simulated behavior.

[0108] To address the above issues, in the next step, the initial / coarse module simulation is enhanced with the help of one or more of the following:

[0109] Knowledge-based approaches (marked “a” in Figure 5C ), which model and implement domain knowledge (industrial process automation and engineering) and thus handle, for example, simulations that do not break the underlying domain rules (thus also providing interpretability of the underlying algorithms and increasing transparency and trustworthiness);

[0110] A data assimilation (optimization) engine that improves or calibrates (black-box or hard-coded) simulation parameters based on assimilating real-world scenario data into the simulated simulation scenario behavior (labeled "b" in FIG5C).

[0111] Therefore, the proposed overall system for module simulation may include software components for virtually performing the above tasks, and optionally (in the case of data assimilation) hardware components for at least one of sensing, monitoring, recording, processing, to facilitate the data acquisition, processing and calculation parts.

[0112] The software component of part (a) in particular overcomes the shortcomings and deficiencies of real data driven and AI enhanced black box module simulation components which learn to simulate the behavior of the module based on what is provided by a neural network (NN) or ML algorithm based on the real training data provided. That is, it is checked whether the NN representation is consistent with the expected target representation of the module / processing functionality. For example, in a reaction module with a capacitive level sensor and a regulation for measuring and balancing the fill level, the NN may misinterpret long (data logging) periods of "no fill" as an "always true" default value, but this should not be the case. Therefore, the knowledge representation component remains responsible for integrating this parameter into the model and for not allowing the user to set the value here to be consistent with the underlying modeling knowledge (further data acquisition is required, for example by taking into account another period where the capacitive level sensor does not measure the "no fill" value).

[0113] The implementation here can be based on a penalty mechanism that penalizes data-driven NN-based simulation behaviors that contradict engineering / physical knowledge / rules. Alternatively, the implementation can be built on a physics-based machine learning algorithm (or model-based machine learning algorithm), where, for example, forward and backward propagation are restricted to allow only physically plausible behaviors, or utilize a reinforcement learning algorithm, where a virtual agent evaluates the NN-based simulation behavior of a given module relative to the physical / chemical / engineering potential behavior and penalizes the resulting errors, thereby improving the weights in the NN representation and, therefore, iteratively improving the actual simulation itself.

[0114] On the other hand, the software components of part (b) will in particular overcome the drawbacks and shortcomings of dedicated white-box module simulation components that are manually programmed / created by the module or MTP manufacturer, based on, for example, C&E simulations, or that can be derived from and executed in accordance with information provided in the MTP or other sources (e.g., from process engineering, such as DEXPI). As mentioned above, these drawbacks are mainly due to the fact that only the basic structure / settings of the model and simulation can be derived from the content of the MTP or from other process industry typical sources (DEXPI, FM, etc.), and therefore this approach can produce a good model based on physical / chemical equations with default parameters, but the parameters may not, for example, have been properly calibrated.

[0115] Therefore, part (b) can overcome this problem using mathematical methods for data assimilation, i.e., mathematical optimization in the sense of minimizing the error between the real behavior measured in the recorded data and the simulated behavior based on the model that has not yet been calibrated. Thus, the data assimilation engine "improves" or calibrates the (possibly black-box or hard-coded) simulation parameters based on assimilating the real scene data into the simulated simulated scene behavior.

[0116] Although Figure 5a and 5b involves processing one type of module and therefore involves calibration / tuning of a simulation of one module, but Figure 5c A processing module chain / topology is shown, which may even comprise modules of different types.

[0117] When simulating modules and entire module-to-module process chains / pipelines, a “cheap” additional feature is the optimization of inter-module calibration, e.g. to prevent bottlenecks or module downtimes and thus improve the performance of the entire process plant.

[0118] Thus, for example, in order to provide data flow driven simulation-based control parameter optimization recommendations to the plant owner, an additional (optional) simulation-based optimization component may be provided with in-the-loop simulation to be added on top of the above-described system.

[0119] Thus, a chain or topology of (possibly different) modules is provided. A chain can be a sequence of modules one after another. A topology can also include parallel modules or chains of modules that feed or consume to / from a next / previous module, including multiple inputs / outputs, etc.

[0120] Back-and-forth / iterative optimization is facilitated through loop simulation with realistically installed module chains / topologies.

[0121] For example, strict parameter requirements and actual lifetime data from a plant can be fed to the simulation and optimization engines for integrating these data into their calculations and, based on this, again proposing an overall better optimized combination set of parameters for all modules used in the respective chain / topology.

[0122] In another direction, simulations and optimizations can be performed based on different hypothetical scenarios including different module setups and parameter combinations, and the remaining tunable parameters calculated in such a way that the actual module chain / topology is optimal with respect to a given criterion.

[0123] Possible criteria could be minimizing module downtime, or avoiding bottlenecks, or reducing energy costs, etc.

[0124] From the above description of the different methods and systems of the present disclosure, it can be seen that the following effects can be provided:

[0125] a) Systems with simulation capabilities that allow reasonably applicable and correct / accurate simulations to be obtained at a relatively low cost: (a priori) simulations themselves can provide a cheaper alternative (or precursor) to actual commissioning and post-commissioning failures or for process plant validation.

[0126] b) Out-of-the-box simulations typically require significant effort and cost, and present difficulties, challenges, problems, and obstacles. However, the proposed enhanced solution approach helps automate many components, further reducing cost / effort, and helps improve the problems and shortcomings associated with simulations (lack of accuracy, poorly calibrated out-of-the-box, difficult to set up, difficult to understand, etc.) through additional capabilities (knowledge-based methods and / or data assimilation engines, plus optional optimization components).

[0127] c) Real data-driven, engineering knowledge-based calibration and fine-tuning of simulation model parameters and control characteristics in order to obtain useful module simulation capabilities, which allows the utilization of virtual module simulation in application scenarios such as VIBN, factory verification, etc.

[0128] d) Interpretable, transparent, resilient and trustworthy simulation model calibration thanks to the combination of (black-box) ML algorithms (working based on evidence together with the provided plant data) and (white-box) knowledge representation algorithms (representing explicit, expert-based engineering knowledge and knowledge models).

[0129] e) Module-by-module and module-to-module optimization, when used with the additional (optional) optimization component and in-the-loop simulation, can provide plant owners with control parameter optimization recommendations based on data flow driven simulation to further improve plant performance.

[0130] f) Simplifies the pre-deployment process, therefore allowing time and cost savings during deployment, and also enables the detection of suboptimal parameter settings by the optimization component of the system and suggests improvements to the control engineer.

[0131] In summary, the present disclosure provides a system that allows obtaining reasonably applicable, correct / accurate and relatively inexpensive module simulation capabilities. It can combine different approaches for building module simulations and simulation models, on the one hand leveraging modeling-related information that can be derived from automation and process engineering sources (DEXPI, MTP, etc.), and on the other hand using data-driven ML-based mechanisms to learn the model / simulation settings.

[0132] It is possible to enhance these basic settings with knowledge-based and / or data assimilation-based enhancement and optimization mechanisms, thus overcoming the typical shortcomings of the basic approaches, thereby allowing reasonably applicable, correct / accurate and relatively inexpensive modular simulation capabilities.

[0133] Additionally, an optional simulation-based optimization component can be added to facilitate simulation in the loop and, consequently, provide dataflow-driven, simulation-based optimization recommendations for control parameters to the plant owner.

[0134] Although the present invention has been described in detail in the drawings and the foregoing description, such description and illustration are to be considered illustrative rather than restrictive. The present invention is not limited to the disclosed embodiments. In view of the foregoing description and the drawings, it will be apparent to one skilled in the art that various modifications may be made within the scope of the present invention as defined by the claims.

Claims

1. A computer-implemented method for simulating module operations of a module of a modular industrial plant, the method comprising: Providing an initial model for the module based on simulation, in particular white box simulation, black box simulation or grey box simulation (S11); Performing a knowledge-based enhancement step (S12), comprising: obtaining an enhanced model by performing a knowledge-based modification on the initial model and / or a knowledge-based extension on the initial model, wherein the knowledge-based modification and / or the knowledge-based extension each model domain knowledge and / or implement domain rules; The module operation of the module is simulated (S14) by means of the enhanced model.

2. A computer-implemented method for simulating module operations of a module of a modular industrial plant, the method comprising: providing an initial model for the module based on simulation, in particular white box simulation or black box simulation (S11); performing a data-driven enhancement step (S13), comprising: obtaining an enhanced model by performing a data-driven modification of the initial model and / or a data-driven extension of the initial model, wherein the data-driven modification and / or the data-driven extension are each based on measurement data obtained for the module in operation, the module being installed as part of the modular industrial plant; The module operation of the module is simulated (S14) by means of the enhanced model.

3. A computer-implemented method for simulating modules of a modular industrial plant, the method comprising: providing an initial model for the module based on simulation, in particular white box simulation or black box simulation (S11); Performing a knowledge-based enhancement step (S12), comprising: obtaining a first enhanced model by performing a knowledge-based modification on the initial model and / or a knowledge-based extension on the initial model, wherein the knowledge-based modification and / or the knowledge-based extension each model domain knowledge and / or implement domain rules; performing a data-driven enhancement step (S13), comprising: obtaining a second enhanced model by performing a data-driven modification of the first enhanced model, wherein the data-driven modification and / or the data-driven extension are each based on measurement data obtained for the module in operation, the module being installed as part of the modular industrial plant; and The module operation of the module is simulated (S14) by means of the second enhanced model.

4. The method of claim 3 , wherein as part of the knowledge-based enhancement step, rule-based bias and / or edge case correction is performed, and subsequently, as part of the data-driven enhancement step, data assimilation is performed for calibration of the first enhancement model and / or calibration of one or more rules for the rule-based bias and / or edge case correction.

5. A method according to claim 1, 3 or 4, wherein the knowledge-based enhancement step is based on module-specific knowledge, the module-specific knowledge comprising at least one of the following: information about correlations between multiple variables, information about linear or nonlinear functional behavior, information about interrelationships, information about edge cases, information about biases in the input data, information about potentially relevant additional simulation parameters.

6. The method of claim 5, wherein said modeling of domain knowledge comprises modeling said module-specific knowledge and / or wherein enforcing domain rules comprises enforcing rules that implement said module-specific knowledge.

7. A method according to any one of claims 2 to 6, wherein the data-driven enhancement step comprises performing data-driven modification and / or extension including data assimilation, wherein the data assimilation is configured to minimize the difference between the simulated output data and the actual measured data, for example using a Kalman filter.

8. The method according to any one of claims 2 to 7, wherein the data-driven enhancement step comprises performing data-driven modification and / or extension, comprising: After the installation of the module, a module-specific calibration of the initial model, in particular of the initial model simulation parameters and / or initial model relationships, is performed.

9. A computer-implemented method for modular plant control parameter determination, the method comprising the method for simulating module operation of modules of a modular industrial plant according to any one of claims 1 to 8, and further comprising: Modeling the relationships between modules of the modular industrial plant (S15); Based on the simulated module operations and the modeled interrelationships of the modules, an optimization step is performed (S16) to provide modular plant control parameters.

10. The method for modular plant control parameter determination according to claim 9, wherein the optimization step comprises data-driven optimization based on simulation data and / or measurement data obtained for the modular industrial plant operation.

11. A system (1) comprising a computing system (2) configured to perform the method according to any one of claims 1 to 10.

12. The system according to claim 11 further comprises one or more sensors (3), which are configured to obtain measurement data for the module in operation and / or measurement data for the modular industrial plant in operation, as part of which the module is installed.

13. The system according to claim 11 or 12, further comprising the modules (4) of the modular industrial plant (5).

14. A computer program product comprising instructions which, when said program is executed by a computing system, cause said computing system to perform the method according to any one of claims 1 to 10.

15. A computer-readable medium comprising instructions which, when executed by a computing system, cause the computing system to perform the method according to any one of claims 1 to 10.

Images