Error-free and safe operation of chemical plants

Abstract

A method for identifying operation risks associated with operating one or more target chemical plant(s), the method comprising: providing a request for identifying operation risks associated with operating the one or more target chemical plant(s), wherein the request includes an indication on the one or more target chemical plant(s), providing unstructured data related to a plurality of operation risks associated with operating one or more chemical plant(s) including the one or more target chemical plant(s), providing operation risk task instructions for identifying at least one operation risk associated with the one or more target chemical plant(s) to one or more data-driven model(s), wherein the operation risk task instructions are related to the plurality of operation risks and the request for identifying the one or more operation risk(s), and wherein the one or more data-driven model(s) are configured to follow the provided task instructions, providing the at least one identified operation risk.

Description

[0001] ERROR-FREE AND SAFE OPERATION OF CHEMICAL PLANTS

[0002] TECHNICAL FIELD

[0003] The disclosure relates to a safe and error-free operation of chemical plants by operating instructions and to a method for identifying operation risks associated with operating one or more target chemical plant(s), an apparatus, an apparatus for identifying operation risks associated with operating one or more target chemical plant(s), use of a data-driven model, use of an operation risk, a method for monitoring and / or controlling one or more target chemical plant(s).

[0004] TECHNICAL BACKGROUND

[0005] Currently, operating instructions for operating chemical plants have to fulfill a variety of requirements. This is facilitated by manually deriving and applying safety measures. Consequently, establishing operating instructions takes a lot of time and resources and is prone to human-made errors. Hence, reliable and fast application of operating instructions is desired.

[0006] SUMMARY

[0007] In an aspect, this disclosure relates to a method, in particular a computer-implemented method, for identifying operation risks associated with operating one or more target chemical plant(s), the method comprising: providing a request for identifying operation risks associated with operating the one or more target chemical plant(s), wherein the request includes an indication on the one or more target chemical plant(s), providing unstructured data related to a plurality of operation risks associated with operating one or more chemical plant(s) including the one or more target chemical plant(s), providing operation risk task instructions for identifying at least one operation risk associated with the one or more target chemical plant(s) to one or more data-driven model(s), wherein the operation risk task instructions are related to the plurality of operation risks and the request for identifying the one or more operation risk(s), and wherein the one or more data-driven model(s) are configured to follow the provided task instructions, providing the at least one identified operation risk, in particular for identifying, preferably mitigating, operation risks associated with operating one or more target chemical plant(s).

[0008] In another embodiment, it relates to an apparatus comprising: a processor configured to perform any one of the methods as described herein. In another embodiment, it relates to an apparatus for identifying operation risks associated with operating one or more target chemical plant(s), the apparatus comprising: a processor configured to perform any one of the methods as described herein.

[0009] In another aspect, it relates to use of a data-driven model according to any one of the methods as described herein for mitigating, in particular eliminating, one or more operation risk(s) for operating one or more target chemical plant(s).

[0010] In another aspect, it relates to use of an operation risk as identified according to any one of the methods as described herein for mitigating, in particular eliminating, the identified operation risk.

[0011] In another aspect, it relates to a method for monitoring and / or controlling one or more target chemical plant(s), the method comprising: providing a request for identifying operation risks associated with operating the one or more target chemical plant(s), wherein the request includes an indication on the one or more target chemical plant(s), providing unstructured data related to a plurality of operation risks associated with operating one or more chemical plant(s) including the one or more target chemical plant(s), providing operation risk task instructions for identifying at least one operation risk associated with the one or more target chemical plant(s) to one or more data-driven model(s), wherein the operation risk task instructions are related to the plurality of operation risks and the request for identifying the one or more operation risk(s), and wherein the one or more data-driven model(s) are configured to follow the provided task instructions, providing the at least one identified operation risk for identifying, in particular mitigating, operation risks associated with operating one or more target chemical plant(s).

[0012] EMBODIMENTS

[0013] Any disclosure, embodiments and examples described herein relate to the methods, the systems, apparatuses, chemical products, uses and computer elements lined out above and below. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples. In the following, terminology as used herein and / or the technical field of the present disclosure will be outlined by ways of definitions and / or examples. Where examples are given, it is to be understood that the present disclosure is not limited to said examples.

[0014] These and other objects, which become apparent upon reading the following description, are solved by the subject matters of the independent claims. The dependent claims refer to embodiments of the disclosure.

[0015] Chemical plants have high safety requirements to ensure error-free production as errors in the operation provide hazardous risks to workers, equipment and environment. Followingly, operating instructions for operating chemical plants are checked thoroughly in highly complex and time consuming processes. As this task is typically performed manually by experts in the field, non-uniform application of safety measures is common depending on the experience of the experts. Hence, learnings in safe operation may be hidden in the knowledge of the experts. Such experts acquire their knowledge through years of practical work. This knowledge can be hardly shared as these safety measures come with a plurality of formats. Followingly, safety measures are applied inconsistently and may be more complex than justified.

[0016] Providing unstructured data related to a plurality of operation risks allows to share safety measures and thus makes the hidden knowledge available to a plurality of recipients. This results in uniformly and appropriate application of safety measures. Therefore, errors in operating chemical plants are reduced while the efficiency of operating chemical plants may be increased due to increasing the fit between risks and safety measures. Further, providing the operation risk task instructions related to the unstructured operation risks to the one or more data- driven model(s) allows to identify applicable operation risks. This provides the advantage of early identification of operation risks and thus, early solving of operation risks. Early identification of operation risks has the benefit of providing more opportunities for resolving the operation risks potentially decreasing the resource and / or time needed for resolving. Furthermore, damage to chemical plant(s) can be reduced leading to less resource waste as chemical plants can be maintained in an improved manner while conversion rates for producing and / or processing chemical products are increased due to the improved operation of the chemical plants.

[0017] In an embodiment, the one or more operation risk(s) may be associated, in particular related to an operation of the one or more chemical plant(s), in particular the one or more target chemical plant(s). The one or more operation risk(s) may be related to at least a part of the equipment of the one or more target chemical plant(s), at least a part of the operating instructions for controlling at least a part of the equipment of the one or more target chemical plant(s). The equipment of the one or more target chemical plant(s) may include at least a part of machinery for processing and / or producing the one or more chemical product(s). In an embodiment, the one or more data-driven model may be pretrained data-driven model(s). The pretrained data-driven model(s) may be parametrized and / or trained based on unstructured data, in particular text data and optionally numerical data such as tabular data or image data. The pretrained data-driven model(s) may be configured to perform a plurality of task. The pretrained data-driven model(s) may be configured to perform the task according to the provided task instruction. Hence, the pretrained data-driven model may be configured to be provided with a plurality of different task instructions and / or provide a plurality of different types of output data upon receiving different task instructions. This allows to handle a plurality of use cases by one data-driven model. Such data-driven models are available and can directly be operated e.g. via API calls. The pretrained data-driven model(s) may be trained and / or parametrized based on unstructured data, in particular text data and optionally numerical data such as tabular data or image data. The unstructured data may be associated with a plurality of different tasks and / or task instructions. The pretrained data-driven model(s) may be configured to perform a plurality of task. The pretrained data-driven model(s) may be configured to perform the task according to the provided task instruction. Hence, the pretrained data-driven model(s) may be configured to be provided with a plurality of different task instructions and / or provide a plurality of different types of output data upon receiving different task instructions.

[0018] In an embodiment, the one or more data-driven model(s) may be finetuned data-driven model(s). The finetuned data-driven model(s) may be obtained by training pretrained data-driven model(s) configured to perform a plurality of tasks according to a plurality of task instructions. The finetuned data-driven model(s) may trained additionally on a training data set comprising a plurality of task instructions of one type and corresponding output data. The finetuned data-driven model may be trained additionally to provide output data of a predefined type according to the training data set. The finetuned data-driven model may be configured to be provided with a plurality of different task instructions and / or provide a plurality of different types of output data upon receiving different types of task instructions. Further, the finetuned data-driven model may be configured to provide one type of output data upon receiving one type of task instruction with a higher accuracy than providing other types of output data upon receiving other types of task instructions. In an embodiment, the one or more finetuned data-driven model(s) may comprise at least one risk identification model, at least one data selecting model and / or at least one matching model. The risk identification model may be finetuned for selecting at least one operation risk associated with the request based on the operation risk task instructions. The data selecting model may be finetuned for selecting the subset of the target configuration data according to identified operation risks. The matching model may be configured to provide a subset of target configuration data upon determining a difference between the target configuration data and the current configuration data and otherwise provide an indication on a verification of the current configuration data. By doing so, the data-driven model may operate more reliable. This in turn increases the reliability of detecting and / or mitigating, in particular eliminating operation risks. Ultimately, this increases the efficiency of operating the one or more target chemical plant(s) as risks are more reliably detected and / or eliminated resulting in handling operation risks earlier.

[0019] In an embodiment, the one or more data-driven models may be configured to follow the one or more task instruction(s) provided to the one or more data-driven models. The task instruction(s) may be indicative of a task to be performed by the one or more data-driven models. The one or more task instruction(s) may comprise unstructured data.

[0020] In an embodiment, the configuration data may be associated with and / or indicative of a configuration of the one or more chemical plant(s). The configuration data may include operating data and / or equipment data. The equipment data may be associated with the equipment of the one or more chemical plant(s) and / or a relation between two or more parts of the equipment. For example, the equipment data may comprise image data indicative of the equipment and / or the relation between two or more parts of the equipment in the one or more chemical plant(s). The current configuration data may comprise configuration data associated with the one or more target chemical plant(s) as operated currently and / or within a predefined time interval. The target configuration data may comprise configuration data associated with an error free and / or target operation of the one or more chemical plant(s). The target configuration data may comprise historical configuration data of the one or more chemical plant(s).

[0021] In an embodiment, the operating data may be related to, in particular may comprise at least parts of operating instructions for controlling at least a part of one or more target chemical plant(s). The operating data may include operating procedures. The operating procedures may be configured for being executed by workers in the chemical plant. The operating instructions may trigger and / or initiate producing and / or processing of one or more chemical product(s) by at least a part of the one or more target chemical plant(s). The operating instructions may be provided to an operating engine of the one or more target chemical plant(s). The operating instructions may be configured to trigger the operating engine of the one or more target chemical plant(s). The current operating instructions may be operating instructions used for controlling at least a part of the one or more target chemical plant(s), in particular at least once preceding the providing of the request for verifying the current configuration data. The target operating instructions may be operating instructions used for controlling at least a part of the one or more chemical plant(s), in particular at least once preceding the providing of the request for verifying the current configuration data. The one or more chemical plant(s) may comprise the one or more target chemical plant(s). The target operating instructions may operate the one or more chemical plant(s) associated with an absence of one or more operation risk(s). The target operating instructions may be associated with error-free operation of at least the part of the one or more chemical plant(s). The target operating instructions may be verified. The target operating instructions may be historical operating instructions.

[0022] In an embodiment, the request for identifying operation risks may comprise an indication on the one or more target chemical plant(s). The indication on the one or more target chemical plant(s) may comprise a digital identifier. Additionally or alternatively, the indication on the one or more target chemical plant(s) may comprise a type of the one or more target chemical plant(s) and / or a location of the one or more target chemical plant(s). The type of the chemical plant(s) may indicate a chemical process, a chemical product and / or equipment associated with the chemical plant(s). The location of the chemical plant(s) may be indicative of a subarea within a predefined area such as the planet, a continent, a country or the like. The indication on the one or more target chemical plant(s) may be configured to identify the one or more target chemical plant(s), in particular for retrieving the one or more operation risk(s) associated with the one or more target chemical plant(s) and / or retrieving the current configuration data and / or at least the part of the target configuration data associated with the one or more target chemical plant(s).

[0023] The location and / or the type of the chemical plant(s)s may be decisive for operating the chemical plant(s). For example, Tarragona may have a lower humidity than Antwerpen. Changes in humidity strongly influence processes conducted in chemical plant(s). Further, the operating instructions may be type-specific as different types of plants are associated with different challenges. Followingly, taking the type of a plant into account allows for tailoring the operating instructions to the chemical plant at hand.

[0024] In an embodiment, the at least one operation risk associated with the request for identifying and / or mitigating, in particular eliminating operation risks may be at least one operation risk associated with operating the one or more target chemical plant(s) in particular based on the current configuration data including the current operating instructions.

[0025] In an embodiment, the operation risk task instructions may be related to the plurality of operation risks associated with the one or more chemical plant(s), the identification task instructions and / or the indication on the one or more target chemical plant(s). In particular, the operation risk task instructions may include the plurality of operation risks associated with the one or more chemical plant(s), the identification task instructions and / or the indication on the one or more target chemical plant(s).

[0026] In an embodiment, the request for identifying the operations risks may be a request for identifying and / or mitigating, in particular eliminating the identified operation risks. Hence, the request for identifying the operation risks may be the request including mitigating, in particular eliminating, task instructions for mitigating, in particular eliminating the identified operation risks.

[0027] In an embodiment, the request for identifying the operations risks may comprise image data related to a representation of a configuration of the one or more target chemical plant(s). The operation risk task instructions may include at least a part of the request, in particular the image data. The configuration of the one or more target chemical plant(s) may be indicative of equipment of the one or more target chemical plant(s) and a relation between one or more component(s) of the equipment. The representation of the configuration may comprise the representation of the equipment of the one or more target chemical plant(s) and a relation between one or more component(s) of the equipment. The one or more data-driven model(s) may be further configured to receive and / or process image data, in particular to at the at least one identified operation risk. In an embodiment, the request for identifying operation risks may comprise image data and data associated with a modality different from image data such as text data. The numerical representation of the operation risk task instructions may be obtained by concatentating the numerical representation of the operations risks and numerical representation of the image data. The numerical representation of the image data may be obtained mapping the image data to the numerical representation of the image data. Configurations of chemical plant(s) may be described by images. Providing image data related to a representation of a configuration of the one or more target chemical plant(s) allows to make use of already available data. Furthermore, the image data may comprise a lot of information related to the one or more target chemical plant(s). Corresponding text data may be unavailable and / or may be prone to errors. Hence, providing image data enables efficient data sharing related to the one or more target chemical plant(s) while allowing for accurate identification of the one or more target chemical plant(s) and related operation risks. Ultimately, this contributes to increasing the efficiency of operating the one or more target chemical plant(s).

[0028] In an embodiment, the operation risk task instructions, in particular the plurality of operation risks, may comprise unstructured data. The operation task instructions may comprise the plurality of operation risks. The one or more data-driven model(s) may be configured to process unstructured data. This provides the advantage of allowing for transparent operation of the data-driven model as the instructions are human-interpretable. Thereby, the operation of the data-driven model can be tailored by experts in the field of plant operation. Ultimately, this contributes to increasing the reliability of verifying the current operating instructions.

[0029] In an embodiment, providing the operation risk task instructions may comprise mapping the operation risk task instructions to a numerical representation of the operation risk task instructions. The one or more data-driven model(s) may be configured to map the numerical representation of the operation risk task instructions to at least one numerical representation of the operation risk. Providing the at least one identified operation risk, in particular by the one or more data-driven model(s), may include mapping the at least one numerical representation of the operation risk to the at least one operation risk. The numerical representation of the operation risk task instructions may be a numerical representation of the operation risk task instructions. The at least one numerical representation of the operation risk may be a numerical representation of the at least one operation risk. Hence, the numerical representation of the operation risk task instructions and / or the at least one numerical representation of the operation risk may comprise a tensor, in particular a vector. The numerical representation of the data may be a machine-processable representation of the provided data. The one or more data-driven model(s) may be configured to map the numerical representation of the operation risk task instructions to contextualized operation risk task instructions, in particular by one or more encoder and / or decoder block(s). The contextualized operation risk task instructions may be a numerical representation related to the operation risk task instructions and a relation between at least two parts of the operation risk task instructions. Further, the one or more data-driven model(s) may be configured to map the contextualized operation risk task instructions to the at least one numerical representation of the operation risk, in particular by one or more encoder and / or decoder output(s). Transforming the input data into a numerical representation of the input data may allow to process unstructured data. Furthermore, the numerical representation of the input data may be processed according to the one or more data-driven model(s) more efficiently by computational means such as GPUs as GPUs are designed for performing matrix calculations. Hence, fast computation and thus, fast identification of operation risks is enabled. Furthermore, this allows to evaluate operation risks of chemical plants more often. Hence, the efficiency of operating the chemical plants is increased.

[0030] In an embodiment, vectorized data may refer to a numerical representation of the data.

[0031] In an embodiment, the request may be related to and / or indicative of a change of a configuration of the one or more target chemical plant(s). Identifying at least one operation risk associated with the one or more target chemical plant(s) may comprise identifying at least one operation risk associated with the change of the configuration of the one or more target chemical plant(s). The change of the configuration may be indicative of one or more components of equipment associated with the one or more target chemical plant(s) and / or at least a part of operating instructions for operating the one or more target chemical plant(s). The change of the configuration may be indicative of an exchange of at least a part of the equipment associated with the one or more target chemical plant(s) and / or a change of at least a part of the operating instructions. The configuration of the one or more target chemical plant(s) may be related to equipment of the one or more target chemical plant(s) and a relation between one or more component(s) of the equipment. In an embodiment, providing the request for identifying operation risks associated with operating the one or more target chemical plant(s) may be triggered by an exchange of at least a part of the equipment associated with the one or more target chemical plant(s) and / or a change of at least a part of the operating instructions. By doing so, a fast adaption of safety measures according to changes related to the one or more target chemical plant(s) are enabled. Hence, shutdowns of chemical plant(s) are reduced. This results in increasing the operation time in relation to the shutdown time of chemical plant(s). Followingly, chemical plants can be operated more efficiently.

[0032] In an embodiment, providing the request for identifying operation risks associated with operating the one or more target chemical plant(s) may be triggered by identifying an operation risk associated with operating a chemical plant different from the one or more target chemical plant(s) and / or by changing a configuration of the one or more target chemical plant(s). The chemical plant different from the one or more target chemical plant(s) may be associated with a type and / or a location of the one or more target chemical plant(s). The type of the chemical plant(s) may indicate a chemical process, a chemical product and / or equipment associated with the chemical plant(s). The location of the chemical plant(s) may be indicative of a subarea within a predefined area such as the planet, a continent, a country or the like. The location and / or the type of the chemical plant(s)s may be decisive for operating the chemical plant(s). For example, Tarragona may have a lower humidity than Antwerpen. Changes in humidity strongly influence processes conducted in chemical plant(s). Further, the operation risks may be typespecific as different types of plants are associated with different challenges. Followingly, taking the type of a plant into account allows for reliably and efficiently identify applicable operation risks.

[0033] In an embodiment, providing the at least one identified operation risk may comprise: providing target configuration data associated with target operating instructions for operating the one or more chemical plant(s), identifying a subset of target configuration data associated with the at least one identified operation risk by providing configuration data task instructions related to the target configuration data and the at least one identified operation risk to the one or more data-driven model(s), wherein the one or more data-driven model(s) are further configured to select the subset of the target configuration data according to identified operation risks, providing the subset of target configuration data, and optionally the at least one identified operation risk.

[0034] The configuration data task instructions may comprise unstructured data. The target configuration data may be related to and / or may include target operating instructions. Providing the configuration data task instructions may include mapping the configuration data task instructions to a numerical representation of the configuration data task instructions. The numerical representation of the configuration data task instructions may be a numerical representation of the configuration data task instructions. The one or more data-driven model(s) may be configured to map the numerical representation of the configuration data task instructions to a numerical representation of the subset of target configuration data. Further, the numerical representation of the subset of target configuration data may be mapped to the subset of the target configuration data. In another embodiment, mapping input data to a numerical representation of the input data may be based on a vocabulary specifying a relation between the input data and / or the numerical representation of the input data. The vocabulary may be obtained based on a plurality of corresponding input data and a numerical representation of the input data pairs. Analogously, the numerical representation of the output data may be mapped to output data. By doing so, data- driven models can operate on unstructured data. Further, the numerical representation of the data may be associated with a resource efficient processing, e.g. by GPUs, in comparison to processing string data associated with long documents.

[0035] In an embodiment, the any one of the methods may further comprise providing configuration data associated with current operating instructions for operating the one or more target chemical plant(s), and matching the target configuration data with the current configuration data, by providing a matching task instruction related to the target configuration data and the current configuration data to the one or more data-driven model(s), wherein the one or more data-driven model(s) are further configured to provide a subset of target configuration data upon determining a difference between the target configuration data and the current configuration data and otherwise provide an indication on a verification of the current configuration data. Providing the matching task instruction may include mapping the matching task instruction to a numerical representation of the matching task instruction. The numerical representation of the matching task instruction may be a numerical representation of the matching task instruction. The one or more data-driven model(s) may be configured to map the numerical representation of the matching task instruction to a numerical representation of the subset of target configuration data or a numerical representation of the indication on the verification of the current configuration data. Further, the numerical representation of the subset of target configuration data may be mapped to the subset of the target configuration data. The numerical representation of the indication on the verification of the current configuration data may be mapped to the indication on the verification of the current configuration data. In another embodiment, mapping input data to numerical representation of the input data may be based on a vocabulary specifying a relation between the input data and / or the numerical representation of the input data. The vocabulary may be obtained based on a plurality of corresponding input data and numerical representation of the input data pairs. Analogously, numerical representation of the output data may be mapped to output data.

[0036] In an embodiment, the one or more data-driven model(s) may be further provided with an output data structure associated with the indication on the verification of the current configuration data and / or the subset of target configuration data. The data-driven model may be further configured to provide a structured indication on the verification of the current configuration data and / or a structured subset of target configuration data according to the output data structure. In an embodiment, the data-driven model may be provided with the matching task instruction to provide the indication on the verification of the current configuration data and / or the subset of target configuration data and the data-driven model may be provided with structuring task instruction related to the output data structure associated with the indication on the verification of the current configuration data and / or the subset of target configuration data and the provided indication on the verification of the current configuration data and / or the subset of target configuration data. The data-driven model may be configured to provide the structured indication on the verification of the current configuration data and / or the structured subset of target configuration data in response to receiving the structuring task instruction. The output data structure may specify a sequence of at least two parts of the indication on the verification of the current configuration data and / or the subset of target configuration data.

[0037] In an embodiment, the request for identifying and / or mitigating, in particular eliminating operation risks associated with operating one or more target chemical plant(s) may be related to risk configuration data associated with an operation risk associated with operating the one or more chemical plant(s), in particular different from the one or more target chemical plant(s). The request may be related to and / or associated with the type and / or the location of the one or more target chemical plant(s). The risk configuration data may specify one or more operation risk(s) and / or one or more countermeasures for mitigating, in particular eliminating the one or more operation risk(s). The countermeasures for mitigating, in particular eliminating the one or more operation risk(s) may be configured to mitigate, in particular eliminate the one or more operation risk(s). The countermeasures for mitigating, in particular eliminating the one or more operation risk(s) may include and / or may be related to at least a part of the target configuration data. By doing so, chemical plants are evaluated with regard to operation risks according to current events in other chemical plants. For example, operation risks caused by external influences such as weather changes of a gas treatment plant in Antwerpen may also relate to polymerization plants in the neighbourhood. Analogously, events occurred in a polymerization plants in Europe may be related to polymerization plants in Asia as polymerization plants share similar equipment. Taking events of associated chemical plants into account for mitigating, in particular eliminating operation risks allows to focus the attention of the one or more data-driven model(s) on potentially applying operation risks.

[0038] In an embodiment, the operation risk task instructions may be related to the plurality of operation risks associated with the one or more target chemical plant(s). Any one of the methods may further comprise selecting the plurality of operation risks associated with the one or more target chemical plant(s) from the plurality of provided operation risks associated with the one or more chemical plant(s) comprising the one or more target chemical plant(s). Selecting the plurality of operation risks associated with the one or more target chemical plant(s) includes providing a request for retrieving the plurality of operation risks associated with the one or more target chemical plant(s) to one or more data source(s), wherein the request for retrieving is associated with one or more target chemical plant(s) and wherein the one or more data source(s) are configured to provide the plurality of operation risks associated with the one or more target chemical plant(s) in response to receiving requests for retrieving the plurality of operation risks associated with the one or more target chemical plant(s). The request for retrieving may be related to the indication on the one or more target chemical plant(s). In particular, the request for retrieving may be indicative of the type of the one or more target chemical plant(s) and / or the location of the one or more target chemical plant(s). This allows for a preselection of applicable operating instructions. By doing so, the amount of data provided to the data-driven model is reduced and the attention of the data-driven model can be focused towards applicable operating instructions. This increases the accuracy of the output data of the data- driven model and hence, the reliability of verifying the operating instructions. Followingly, this contributes to increasing the efficiency of operating the one or more target chemical plant(s).

[0039] In an embodiment, selecting the plurality of operation risks associated with the one or more target chemical plant(s) may include providing a request for retrieving the plurality of operation risks to one or more data source(s). The request for retrieving may be associated with the one or more target chemical plant(s). The one or more data source(s) may be configured to provide the operation risks associated with the one or more target chemical plant(s) in response to receiving the request for retrieving. The request may be related to, preferably may comprise the indication on the one or more target chemical plant(s). By doing so, the amount of data provided to the data-driven model is reduced and the attention of the data-driven model can be focused towards applicable operating instructions. This increases the accuracy of the output data of the data-driven model and hence, the reliability of verifying the operating instructions. Followingly, this contributes to increasing the efficiency of operating the one or more target chemical plant(s).

[0040] In an embodiment, the one or more data source(s) may comprise an embedding database. The embedding database may be configured to determine a distance between a numerical representation of the indication on the one or more target chemical plant(s) and a plurality of numerical representations of the operation risk(s) associated with the one or more chemical plant(s). The embedding database may be further configured to provide the operation risks associated with the one or more target chemical plant(s) based on determining a distance smaller than a predefined distance between the numerical representations of the operation risk(s) associated with the one or more chemical plant(s) and the numerical representation of the request for retrieving the plurality of operation risks associated with the one or more target chemical plant(s). Retrieving the plurality of operation risks associated with the one or more target chemical plant(s) from structured databases allows for more accurate retrieval of operation risks as semantic meanings are considered to determine the requested data rather than occurrence of specific data points. Thereby, the reliability of identifying and / or mitigating, in particular eliminating the operation risks is increased. The numerical representation of the operation risks associated with the one or more chemical plant(s) may be a numerical representation of the operation risks associated with the one or more chemical plant(s), such as a vector. The numerical representation of the operation risks associated with the one or more target chemical plant(s) may be a numerical representation of the operation risks associated with the one or more target chemical plant(s). The numerical representation of the operation risks associated with the one or more target chemical plant(s) may be obtained by mapping the operation risks associated with the one or more target chemical plant(s) to numerical representation of the operation risks associated with the one or more target chemical plant(s). The numerical representation of the operation risks associated with the one or more chemical plant(s) may be obtained by mapping the operation risks associated with the one or more chemical plant(s) to a numerical representation of the operation risks associated with the one or more chemical plant(s).

[0041] In an embodiment, the request for retrieving the operation risks associated with the one or more target chemical plant(s) may comprise a structured query associated with the one or more target chemical plant(s), in particular the indication on the one or more target chemical plant(s). The one or more data source(s) may comprise a structured database configured to receive structured queries and providing data as requested by the structured query. In particular, the structured databased may be configured to provide the operation risks associated with the one or more target chemical plant(s) in response to receiving a structured query for retrieving the operation risks associated with the one or more target chemical plant(s). The structured query may be related to the type of the one or more target chemical plant(s) and / or the location of the one or more target chemical plant(s). Retrieving target configuration data from structured databases allows for robust retrieval of target configuration data.

[0042] In an embodiment, the plurality of operation risks may be associated with a plurality of confidence scores. The confidence scores may be indicative of a likelihood that the operation risks may result in a decreased efficiency of operating the one or more chemical plant(s). Any one of the methods may further include determining if the confidence score associated with the identified one or more operation risk may be within a confidence range, in particular a predefined and / or user-defined confidence range. The at least one identified operation risk may be provided in response to determining that the confidence score associated with the one or more identified operation risks may be within the confidence range. The confidence scores associated with the plurality of operation risks may be provided, e.g. together with the plurality of operation risks.

[0043] In an embodiment, the target configuration data may be associated with a plurality of resources scores. The resource score may be indicative of a resource usage related to providing the target configuration data for controlling the one or more target chemical plant(s). The subset of the target configuration data may be associated with one or more resource score(s). The subset of configuration data may be provided in response to determining that the resource score may be within a resource range. Any one of the methods may further comprise determining if the resource score associated with the subset of target configuration data may be within a resource range, in particular a predefined and / or user-defined resource range. This allows to limit reengineering of chemical plants to a meaningful and necessary quantity without sacrificing on safety.

[0044] In an embodiment, the request may be related to and / or indicative of building the one or more target chemical plant(s), and wherein identifying at least one operation risk associated with the one or more target chemical plant(s) comprises identifying at least one operation risk associated with building of the one or more target chemical plant(s). This allows to identify and / or eliminate operation risks while building new chemical plants.

[0045] In an embodiment, the operation risk task instructions may be associated with the plurality of operation risks and the request for identifying the one or more operation risk(s),

[0046] BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0047] In the following, the present disclosure is further described with reference to the enclosed figures. The same reference numbers in the drawings and this disclosure are intended to refer to the same or like elements, components, and / or parts.

[0048] In the following, the present disclosure is further described with reference to the enclosed figures. The same reference numbers in the drawings and this disclosure are intended to refer to the same or like elements, components, and / or parts.

[0049] FIG. 1 illustrates an embodiment of an operating system 106 of one or more chemical production facilities 102.

[0050] FIG. 2 illustrates an embodiment of a method for verifying operating instructions.

[0051] FIG. 3 illustrates an embodiment of verifying operating instructions.

[0052] FIG. 4 illustrates an embodiment of a user interface for verifying operating instructions.

[0053] FIG. 5 illustrates an embodiment of processing input data by the one or more data-driven model(s). FIG. 6 illustrates an embodiment of a method for verifying operating instructions.

[0054] FIG. 7 illustrates an embodiment of training an embedding layer.

[0055] FIG. 8A illustrates an embodiment of a transformer encoder architecture.

[0056] FIG. 8B illustrates an embodiment of a transformer decoder architecture.

[0057] FIG. 8C illustrates an embodiment of a transformer encoder-decoder architecture.

[0058] FIG. 9 illustrates an embodiment of training and / or deploying the transformer encoder, the transformer decoder and / or the transformer encoder-decoder.

[0059] FIG. 10 illustrates an embodiment of input embedding.

[0060] FIG. 11 illustrates an embodiment of input embedding.

[0061] DETAILED DESCRIPTION

[0062] The following embodiments are mere examples for implementing the method, the system or application device disclosed herein and shall not be considered limiting.

[0063] FIG. 1 illustrates an embodiment of an operating system 106 of one or more chemical production facilities 102.

[0064] One or more chemical production facilities 102 may be configured for producing one or more chemical products 104 from raw material 124. The one or more chemical production facilities 102 may comprise a plurality of components. For example, the components may include heat exchangers, cooling unit, filter, pumps, valves, separation chambers or the like. The one or more components of the one or more chemical production facilities 102 may be controlled by a control engine 108. The control engine 108 may be configured for controlling the one or more chemical production facilities 102. The control engine 108 may receive operating instructions, in particular from the model engine 110. The model engine 1 10 may be configured for providing operating instructions, in particular verified operating instructions such as verified current configuration data and / or a subset of target configuration data associated with one or more identified operation risks. Examples for target configuration data can be found for example in the stature database (Stature Risk Management™). Further examples for retrieving target configuration data may include retrieving the target configuration data from https: / / www.csb.gov / recommendations / (20.12.2023) or https: / / www.kas-bmu.de / tras.html (20.12.2023). The verified operating instructions may be provided by the model engine 110 and / or as described within the context of FIG. 2. The subset of target configuration data may be related to and / or may comprise target operating instructions, in particular historical operating instructions obtained from previous operations of the one or more chemical plant(s) corresponding to the one or more target chemical plant(s). The current configuration data may be used for controlling the one or more target chemical plant(s) at least once prior to providing the request for identifying and / or mitigating, in particular eliminating one or more operation risk(s). The verified operating instructions may be obtained by providing a request for identifying and / or mitigating, in particular eliminating the one or more operation risk(s) associated with at least a part of the one or more target chemical plant(s). The request may be provided via an intake interface 1 12. The request may be related to an indication on the one or more target chemical plant(s). Hence, the request for identifying and / or mitigating, in particular eliminating the one or more operation risk(s) may be configured for identifying the one or more target chemical plant(s). The request may be configured for identifying one or more operation risk(s) associated with the one or more target chemical plant(s). For this purpose, the one or more operation risk(s) may be selected from a plurality of operation risks associated with the one or more chemical plants. The operation risks may be related to an operation of chemical plant(s). For example, the operation risks may include putting a container at a pressure higher than a predefined pressure or opening a valve for allowing a flammable substance to exit close to a heating element. The operation risks may related to an action for decreasing an efficiency such as an input-output ratio related to the processed and / or produced chemical products.

[0065] Further, the request may be configured for retrieving target configuration data associated with operating instructions for operating the one or more chemical plant(s) and / or equipment of the one or more chemical plant(s). For example, the target configuration data 120 may be retrieved from one or more data sources 1 14. At least a part of the target configuration data 120 may be selected by matching the indication on the one or more target chemical plant(s) with target plant data 122. The target plant data 122 may be associated with a configuration of one or more chemical plant(s), preferably comprising the one or more target chemical plant(s). The target plant data 122 may be associated with the target configuration data 120. The target configuration data and the target plant data 122 may be retrieved from the one or more data sources 114, in particular based on the request for identifying and / or mitigating, in particular eliminating the one or more operation risk(s). The retrieving of current configuration data, target configuration data and / or target plant data 122 may be described in further detail within the context of FIG. 6. Hence, matching the target plant data 122 with the indication on the one or more target chemical plant(s) may result in identifying at least a part of the target configuration data 120 associated with the one or more target chemical plant(s). The part of the target configuration data 120 associated with the one or more target chemical plant(s) may comprise target operating instructions configured for controlling the one or more chemical production facilities 102 via the control engine 108. If the current configuration data corresponds to the part of the target configuration data 120, an indication on the verification of the current configuration data 116 may be provided. The indication may trigger providing the current configuration data to the control engine 108. In an embodiment, at least a subset of target configuration data of the part of the target configuration data associated with the one or more target chemical plant(s) may differ from the current configuration data. This may result in providing the subset of target configuration data. The subset of target configuration data may be provided to the control engine 108 for controlling the one or more chemical production facilities 102. Further, parts of the current configuration data matching the part of the target configuration data associated with the one or more target chemical plant(s) may be provided to the control engine 108, preferably in addition or together with the subset of target configuration data 120. The operating instructions related to the current configuration data and / or the target configuration data may trigger the production and / or processing of one or more chemical products 104.

[0066] The indication on the verification of the current configuration data or the subset of target configuration data may be provided by the model engine 1 10. For this purpose, the model engine 1 10 may be provided with the target configuration data, target plant data 122, current plant data 118 and / or the current configuration data for determining the part of the target configuration data corresponding to the current configuration data. The model engine 110 may be configured for selecting at least the part of the target configuration data associated with the one or more target chemical plant(s) according to the target plant data 122 and the current plant data 118. The model engine 110 may be configured for providing the part of the target configuration data and the current configuration data to the one or more data-driven model(s). Hence, the model engine 1 10 may comprise an interface to the one or more data-driven model(s) for providing the part of the target configuration data and the current configuration data. Further, the one or more data-driven model(s) may provide the indication on the verification of the current configuration data and / or the subset of the target configuration data in response to receiving the current configuration data and the part of the target configuration data. Hence, the interface to the data-driven model(s) may be configured for providing output of the one or more data-driven model(s). The data- driven model may be described in more detail in the context of FIG. 2 and FIG. 5.

[0067] FIG. 2 illustrates an embodiment of a method for mitigating, in particular eliminating one or more operation risk(s) associated with operating one or more target chemical plant(s).

[0068] A request for identifying and / or mitigating, in particular eliminating one or more operation risk(s) associated with operating one or more target chemical plant(s) may be provided 208. The request may comprise unstructured data. In particular, the request may be human-interpretable, e.g. the request may comprise natural language. The request may be provided by a user and / or via a user interface. Providing the request may be triggered by detecting an operation risk in a chemical plant other than the one or more target chemical plant(s). In particular, the chemical plant other than the one or more target chemical plant(s) may of the same type and / or may be in the same location as the one or more target chemical plant(s). Detecting the operation risk may include determining one or more parameters associated with the one or more chemical plant(s) being within an operating risk range, preferably for a time interval being equal or larger than an operation risk time interval. In an embodiment, the request may be indicative of the detected operation risk associated with the chemical plant other than the one or more target chemical plant(s)

[0069] The request may comprise an indication on the one or more target chemical plant(s). The one or more target chemical plant(s) may be identified based on the indication on the one or more target chemical plant(s). For example, the request may include a digital identifier, one or more properties of the one or more target chemical plant(s), a type of the one or more target chemical plant(s), a location of the one or more target chemical plant(s) or the like.

[0070] In an embodiment, the request may be related to, in particular comprise plant configuration and / or configuration data. The configuration data may be indicative of equipment and / or operating instructions associated with the one or more target chemical plant(s). In an embodiment, the plant configuration and / or configuration data may be indicative of a change of equipment and / or operating instructions associated with the one or more target chemical plant(s).

[0071] In an embodiment, the configuration data may comprise image data related to a representation of a configuration of the one or more target chemical plant(s), in particular the equipment and a relation between one or more parts of the equipment of the one or more target chemical plant(s).

[0072] A plurality of operation risks associated with operating one or more chemical plant(s) may be provided 210. The plurality of operation risks may be provided by one or more data sources 1 14. The plurality of operation risks may be detected in the one or more target chemical plant(s) before and / or during operating the one or more chemical plant(s).

[0073] One or more operation risk(s) may be selected from the plurality of operation risks 212. The one or more operation risk(s) may be associated with the one or more target chemical plant(s). This may include retrieving the one or more operation risk(s) from the one or more data sources 114. Hence, a subset of available operation risks may be provided. The one or more operation risk(s) associated with the one or more target chemical plant(s) may comprise one or more operation risk(s) applicable to the one or more target chemical plant(s). The one or more operation risk(s) may be detected in one or more chemical plant(s) of the type and / or at the location of the one or more target chemical plant(s). Selecting the one or more operation risk(s) may include providing a request for retrieving the one or more operation risk(s). The request may comprise a structured query related to the request, in particular the type and / or the location of the one or more target chemical plant(s). The one or more data sources 1 14 may be configured for providing the one or more operation risk(s) in response to receiving the structured query. Additionally or alternatively, the one or more data sources 1 14 may be configured for providing current the one or more operation risk(s) by determining that a distance between the numerical representation of the request for retrieving the one or more operation risk(s) associated with the one or more target chemical plant(s) and a numerical representation of the plurality of operation risks associated with the one or more chemical plant(s) may be within a predefined range. The numerical representations of the operation risk(s) may be obtained by mapping the operation risks to numerical representations of the operation risk(s), preferably via passing the operation risks through one or more embedding layers as described within the context of FIG. 7, FIG. 10 and FIG. 11 . Preselecting potentially applicable operation risks helps to focus the attention of the data-driven model for identifying applicable operation risks among the operation risks with the highest potential.

[0074] At least one operation risk associated with at least part of the one or more target chemical plant(s) may be identified by providing operation risk task instructions to one or more data-driven model(s) 214. The operation risk task instructions may be related to, preferably may comprise, the one or more operation risk associated with the one or more target chemical plant(s) and the request. The one or more data-driven model(s) may be configured for selecting at least one operation risk associated with the request, in particular with the change and / or part of the one or more target chemical plant(s) indicated by the request, based on the operation risk task instructions. The operation risk task instructions, in particular the operation risks and / or the request may comprise unstructured data.

[0075] Providing the operation risk task instructions to the data-driven model may comprise mapping the operation risk task instructions to a numerical representation of the operation risk task instructions. The data-driven model may be configured for mapping the numerical representation of the operation risk task instructions to numerical representation of the selected operation risks associated with at least the part of the one or more target chemical plant(s), in particular associated with the change indicated by the request for identifying and / or mitigating, in particular eliminating the one or more operation risk(s). Preferably, the data-driven model may be configured for mapping the numerical representation of the operation risk task instructions to contextualized operation risk task instructions and / or mapping the contextualized operation risk task instructions to a numerical representation of the selected operation risks. The contextualized operation risk task instructions may be a context tensor related to the operation risk task instructions. The numerical representation of the selected operation risk(s) and / or the numerical representation of the operation risk task instructions may be a numerical representation of the selected operation risk(s) and / or the operation risk task instructions. The numerical representation may comprise structured data. The numerical representation of the operation risk task instructions may be obtained by embedding the operation risk task instructions. The operation risk task instructions may comprise human-interpretable data, in particular text data, numerical data and / or image data. The image data may be embedded via one or more embedding layers as described within the context of FIG. 11 . The text data may be embedded as described within the context of FIG. 7. The numerical, in particular tabular, data may be embedded as described within the context of FIG. 10. Where the operation risk task instructions may comprise at least two of numerical data, in particular tabular data, text data or image data. The numerical representation of the operation risk task instructions may be obtained by mapping the text data to a numerical representation of the text data according to FIG. 7, mapping the numerical, preferably tabular, data to a numerical representation of the numerical, preferably tabular, data according to FIG. 10 and / or mapping the image data to a numerical representation of the image data according to FIG. 1 1. The numerical representation of the text, image, numerical and / or tabular data may be concatenated to a numerical representation of the operation risk task instructions.

[0076] Target configuration data associated with target operating instructions for operating the one or more chemical plant(s) may be provided 216. The target configuration data may be provided by the one or more data sources 114. The data source for providing the target configuration data may be an configuration data source and / or may be different from the data source for providing the operation risk(s), i.e. a risk providing data source. In particular, target configuration data associated with the one or more target chemical plant(s) may be provided, e.g. by providing a request for retrieving the target configuration data associated with the one or more target chemical plant(s). The one or more data sources 114 may be configured for providing the target configuration data in response to receiving a request for retrieving the target configuration data. The request for retrieving the target configuration data may be indicative of the type and / or the location of the one or more target chemical plant(s). Additionally or alternatively, the request for retrieving the target configuration data may be related to the indication on the one or more target chemical plant(s). Target configuration data associated with the one or more target chemical plant(s) may comprise target configuration data associated with target operating instructions for operating the one or more chemical plant(s) with the type and / or the location of the one or more target chemical plant(s).

[0077] At least a part of the target configuration data associated with the at least one identified operation risk may be identified by providing configuration data task instructions related to the target configuration data and the at least one identified operation risk to the data-driven model 218. The one or more data-driven model(s) may be further configured for selecting at least a part of the target configuration data according to identified operation risks. The data-driven model may be configured for matching the target configuration data and the identified operation risk(s). The target configuration data, in particular the target operating instructions may be associated with the operation risk(s). The target operating instructions may be configured for mitigating, in particular eliminating the at least a part of the plurality of operation risk(s). In particular, the identified part of the target configuration data may be configured for mitigating, in particular eliminating the at least one operation risk. The configuration data task instructions may comprise unstructured data, in particular text data and / or numerical, in particular tabular, data. The unstructured data may instruct the data-driven model to select the part of the target configuration data associated with the at least one identified operation risk.

[0078] Current configuration data associated with current operating instructions for operating the one or more target chemical plant(s) may be provided 220. The current configuration data associated with the one or more target chemical plant(s) may be provided by the one or more data sources 1 14. The one or more data sources 114 may store current configuration data associated with the one or more chemical plant(s). The one or more data sources 114 may be configured for providing the current configuration data associated with the target configuration data in response to receiving a request for retrieving the current configuration data associated with the one or more target chemical plant(s). Providing the current configuration data may comprise providing the request for retrieving the current configuration data. The request for retrieving the current configuration data associated with the one or more target chemical plant(s) may be indicative of the one or more target chemical plant(s), in particular may comprise the indication on the one or more target chemical plant(s).

[0079] A subset of the identified part of the target configuration data deviating from the current configuration data may be determined 222. Hence, a difference between the current configuration data and the target configuration data may be determined. This may include providing the current configuration data and at least the selected part of the target configuration data to a data-driven model, in particular matching task instruction related to the subset of target configuration data and the current configuration data associated with the one or more target chemical plant(s). The data-driven model may be configured for selecting the subset of target configuration data not yet applied for mitigating, in particular eliminating the identified operation risks. Matching the current configuration data with the target configuration data for mitigating, in particular eliminating the identified operation risks may provide allow to check whether measures for mitigating, in particular eliminating the identified operation risks may already be implemented. For example, where the configuration of the one or more target chemical plant(s) may allow for operation risks, specific operating instructions may prevent the operation of the one or more target chemical plant(s) from the operation risks. Hence, by doing so, already solved issues will not be propagated and for example, unnecessary stoppages may be prevented. Followingly, this contributes to increasing the efficiency of operating target chemical plant(s).

[0080] The identified subset of target configuration data may be provided for controlling the one or more target chemical plant(s) 224. Additionally or alternatively, the identified at least one operation risk and / or at least the part of the target configuration data for mitigating, in particular eliminating the identified operation risk(s) may be provided.

[0081] Additionally or alternatively, is the part of the target configuration data may match the current configuration data, in particular at least partially and / or fully, an indication on a verification of the current configuration data may be provided. The indication may be provided to a control engine 108 for controlling the one or more target chemical plant(s) according to the current configuration data.

[0082] FIG. 3 illustrates an embodiment of verifying operating instructions.

[0083] In an embodiment, the request may be related and / or may comprise an operation risk(s) associated with a chemical plant where the operation risk may be detected. The request, in particular the operation risk, may be provided via a user interface. The operation risk may be provided together with an indication on the one or more target chemical plant(s). The data-driven model may be configured for identifying one or more operation risk(s) associated with the one or more target chemical plant(s). In response to determining one or more operation risk(s) associated with the one or more target chemical plant(s), the data-driven model may be further configured for identifying and / or providing the subset of target configuration data as described in the context of FIG. 2. The subset of the target configuration data may be provided in response to determining a difference between the target configuration data and the current configuration data.

[0084] Taking operation risks of associated chemical plant(s) into account for mitigating, in particular eliminating operation risk associated with the one or more target chemical plant(s) allows to identify applicable operation risks and adapt the operating instructions in real time according to current events in other plants. Thereby, the error rate of plants is reduced and potential malfunction of chemical plants can be eliminated. Ultimately, this contributes to increase the plant safety and the efficiency of operating chemical plants.

[0085] FIG. 4 illustrates an embodiment of a user interface for mitigating, in particular eliminating operation risks.

[0086] The user interface 402 may be configured for providing the indication on the one or more target chemical plant(s).

[0087] The user interface 402 may allow to enter a plant ID. Hence, the request may be indicative of a digital indentifier associated with the one or more target chemical plant(s). The current configuration data, the target configuration data and / or the target plant data may be retrieved based on a digital indentifier associated with the one or more target chemical plant(s). In particular, the request may indicate the type and / or the location of the one or more target chemical plant(s). Based on the request, the part of the target configuration data associated with the one or more target chemical plant(s) may be selected and / or retrieved as described in the context of FIG. 2. The request may identify the one or more target chemical plant(s). Based on the request, one or more operation risk(s) as described in the context of FIG. 1 may be identified and / or eliminated, e.g. by introducing a subset of target configuration data for controlling the one or more target chemical plant(s).

[0088] FIG. 5 illustrates an embodiment of processing input data by the one or more data-driven model(s).

[0089] The data-driven model may be configured for receiving a representation of the unstructured data, i.e. the numerical representation of the input data, and relate the numerical representation of the input data to a numerical representation of the output data. The data-driven model may comprise one or encoder block(s) and / or one or more decoder block(s) configured for mapping the numerical representation of the input data to contextualized input data and / or one or more encoder output(s) and / or one or more decoder output(s) for mapping the contextualized data to a numerical representation of the data. The encoder block(s), the decoder block(s), the decoder output(s) and / or the encoder output(s) may be described in further detail in FIG. 8A - FIG. 8C. The input data may be for example the operation risk task instructions, the configuration data task instructions and / or the matching task instruction. The output data may be the one or more selected operation risk(s) associated with at least the part of the one or more target chemical plant(s), in particular the change of equipment and / or the part associated with the one or more target chemical plant(s), the identified part of the target configuration data and / or the subset of the identified part of the target configuration data different from the current configuration data.

[0090] Providing input data to the one or more data-driven model(s) may comprise generating vectorized input data. The vectorized input data may be generated by providing the input data to an embedding engine 506. The vectorized input data may be a numerical representation of the input data. The embedding engine may be configured for embedding the input data, i.e. providing vectorized input data in response to receiving input data. The embedding engine 506 may comprise one or more embedding layer(s) as described within the context of FIG. 7. The one or more data-driven model(s) may be configured for mapping the vectorized input data to contextualized input data, in particular by one or more encoder and / or decoder blocks as described in the context of FIG. 8A - FIG. 8C. The contextualized input data may be a numerical representation of vectorized input data and / or the input data and a relation between parts of the vectorized input data and / or the input data. Further, the one or more data-driven model(s) may be configured for mapping the contextualized input data to vectorized output data, in particular by one or more encoder output(s) and / or decoder output(s). The vectorized output data may be mapped to output data by a decoding engine 508. The decoding engine may be configured for mapping a vectorized representation to natural language, numbers and / or symbols, i.e. human-interpretable output data. The output data may comprise the at least one operation risk, the indication on the verification of the current configuration data, the subset of target configuration data or a combination thereof. The data-driven model may be parametrized and / or trained based on unstructured data, in particular data comprising natural language, numbers and / or symbols. The data- driven model may be parametrized and / or trained to generate a sequence of elements associated with the output data. In particular, the output data may comprise the sequence of the elements. A possible implementation of the data-driven model may be seen in FIG. 8A - FIG. 8C. Where the input data may comprise natural language data, the word embedding as described in the context of FIG. 7 may be deployed. Where the input data may comprise numerical data, in particular tabular data, the embedding as described in the context of FIG. 10 may be deployed. Where the input data may comprise image data, embedding according to FIG. 11 may be deployed. Where the input data may be associated with two or more types of input data, i.e. two or more modalities of input data, the vectorized representation of the two or more modalities of the input data may be concatentated prior to providing the vectorized representation to the data-driven model.

[0091] FIG. 6 illustrates an embodiment of a method for verifying operating instructions.

[0092] A request for verifying current configuration data associated with current operating instructions configured for controlling one or more component(s) of one or more target chemical plant(s) 608 may be provided. The request may comprise unstructured data. In particular, the request may be human-interpretable, e.g. the request may comprise natural language. The request may be provided by a user and / or via a user interface. Providing the request may be triggered by detecting a malfunction in the one or more chemical plant(s). The one or more chemical plant(s) may comprise the one or more target chemical plant(s). In an embodiment, the malfunction may be detected in a chemical plant of a type equal to a type of the one or more target chemical plant(s). Detecting a malfunction may include determining one or more parameters associated with the one or more chemical plant(s) being within a malfunction range, preferably for a time interval being equal or larger than a malfunctioning time interval.

[0093] The request may comprise current configuration data. The current configuration data may be indicative of current operating instructions. Current configuration data may include current operating instructions configured for triggering one or more target chemical plant(s) to process and / or produce one or more chemical product(s). The current operating instructions may be used for operating the one or more target chemical plant(s). The current operating instructions may be provided to one or more chemical production facilities for operating one or more component(s) of the one or more chemical production facilities. In particular, the current operating instructions may be provided to a control engine 108 associated with the one or more chemical production facilities 102 for controlling one or more component(s) of the one or more chemical production facilities 102 as described within the context of FIG. 1 .

[0094] Further, the request may comprise current plant data associated with a configuration of the one or more target chemical plant(s). The configuration of chemical plant(s) may be indicative of one or more components of the chemical plant(s) and / or a relation between two or more components of the chemical plant(s). The current plant data may be indicative of a location, a type, one or more chemical product(s) associated with the one or more target chemical plant(s). The location associated with the one or more target chemical plant(s) may indicate a region where the one or more target chemical plant(s) may be located and / or planned to be located. The region may be a continent, a country, a state or the like. Taking the location of chemical plant(s) into account for determining operating instructions is advantageous as this allows to operate chemical plant(s) according to region-specific conditions. For example, temperature and humidity highly depend on the location. Different levels of humidity and temperature may allow for different safe operating instructions. Dry and high temperature regions may require additional safety measures in the form of adapted operating instructions to reduce the risk of forest fires. The type associated with the one or more target chemical plant(s) may be indicative of a type of chemical reaction or physical processing carried out by the one or more target chemical plant(s). For example, the one or more target chemical plant(s) may be gas treatment plant or a polimerization plant. Both types of chemical plant(s) may require different operating instructions for a safe operation as the potential risks differ due the type of chemical reaction carried out by the chemical plant(s). Hence, taking the type associated with the one or more target chemical plant(s) into account for verifying the current configuration data allows to tailor the current configuration data to the specifics of the process conducted. This is desired as chemical products come with a wide variety of properties such as flammability, vapor pressure, melting point, toxicity or the like. These properties are decisive for how to process and / or produce the chemical products. Current plant data indicative of one or more chemical products associated with the one or more target chemical plant(s) may include an indication on the chemical structure of the chemical products, hazard symbols, a safety data sheet or the like. Hence, taking the one or more chemical product(s) into account for verifying the current configuration data allows to tailor the current configuration data to the specifics of the chemical products processed and / or produced.

[0095] In an embodiment, the request may be indicative of the one or more target chemical plant(s). The current configuration data may be provided in response to providing the request. Preferably, the current configuration data may be retrieved by providing a request for providing the current configuration data. The request for providing the current configuration data may be based on the request for verifying the current configuration data. For example, the request for retrieving the current configuration data may comprise a vectorized request for verifying the current configuration data, in particular a vectorized indication on the one or more target chemical plant(s). The vectorized indication on the one or more target chemical plant(s) may be provided to a data source configured for providing current configuration data in response to receiving the vectorized indication on the one or more target chemical plant(s). The data source may be configured for providing current configuration data by determining a distance between the vectorized request for verifying the current configuration data and vectorized current configuration data to be within a predefined range. Additionally or alternatively, the current configuration data may be retrieved by providing a structured query related to the request for verifying the current configuration data, in particular the indication on the one or more target chemical plant(s). The structured query may be configured for triggering the data source to provide the current configuration data. By retrieving the current configuration data according to the request for verifying the current configuration data, the current configuration data can be retrieved in real time. Further, errors are reduced as the current configuration data does not need to be provided by a user and the necessary data for verifying the current configuration data can be retrieved reliably allowing for a robust verification based on the current plant data.

[0096] The current plant data may be provided in response to providing the current configuration data. The current configuration data may be indicative of the one or more target chemical plant(s). The current plant data may be retrieved according to the current configuration data. Hence, providing the current plant data may comprise providing a request for retrieving the current plant data. The request for retrieving the current plant data may be related to the current configuration data. For example, the request for retrieving the current plant data may comprise vectorized current configuration data. The vectorized current configuration data may be provided to a data source configured for providing current plant data in response to receiving current configuration data. The data source may be configured for providing current plant data by determining that a distance between the vectorized current configuration data and vectorized current plant data may be within a predefined range. Additionally or alternatively, the current plant data may be retrieved by providing a structured query related to the current configuration data, in particular indicative of the one or more target chemical plant(s). The structured query may be configured for triggering the data source to provide the current plant data. By retrieving the current plant data according to the current configuration data, the current plant data can be retrieved in real time. Further, errors are reduced as the current plant data does not need to be provided by a user and the necessary data for verifying the current configuration data can be retrieved reliably allowing for a robust verification based on the current plant data

[0097] Target configuration data associated with target operating instructions configured for triggering one or more chemical plant(s) to process and / or produce one or more chemical product(s) may be provided 610. The one or more chemical plant(s) may include the one or more target chemical plant(s). Where the one or more chemical plant(s) differ from the one or more target chemical plant(s), an indication that the current configuration data is unverified or can not be verified may be provided. The target configuration data may be obtained from previous operations of the one or more chemical plant(s). Hence, the target configuration data may comprise historical configuration data used for operating the one or more chemical plant(s). In particular, the historical configuration data may be associated with an error-free operating of the one or more chemical plant(s). This allows to uniformly use past experience with operating chemical plant(s) to ensure future error-free operation of the one or more target chemical plant(s). The target configuration data may be provided by a data source. The data source may be provided by a data source. Providing the target configuration data may include storing the target configuration data in an available data source. In an embodiment, providing the target configuration data may include retrieving the target configuration data from the data source. The retrieved target configuration data may be provided to a data-driven model for selecting at least a part of the target configuration data. Further, the target plant data may be provided to the data-driven model for selecting at least a part of the target configuration data. The data-driven model may be configured for selecting at least the part of the target configuration data by matching at least the plant data associated with at least the part of the target configuration data with the current plant data. Hence, the data-driven model may be provided with the request for selecting at least the part of the target configuration data.

[0098] Further, target plant data associated with a configuration of the one or more chemical plant(s) may be provided 612. The target plant data may be indicative of a location, a type, one or more chemical product(s) associated with the one or more chemical plant(s), preferably analogous to the above-described current plant data. The target plant data may be provided by a further data source or the data source for providing the target configuration data. In an embodiment, the target plant data and the target configuration data may be provided together or separate.

[0099] At least a part of the target configuration data associated with one or more target chemical plant(s) may be selected by matching the current plant data and the target plant data 614. This may include selecting at least the part of the target configuration data where the target plant data associated with at least the part of the target configuration data corresponds to the current plant data. Hence, at least the part of the target configuration data may be selected according to a type, a location and one or more chemical products associated with the one or more chemical plant(s). The type, the location and the one or more chemical products associated with at least the part of the target configuration data may correspond to the type, the location and the one or more chemical products associated with the current configuration data.

[0100] In an embodiment, providing the target configuration data and selecting at least a part of the configuration data may include retrieving at least a part of the target configuration data. Retrieving the target configuration data may include providing a request for retrieving at least a part of the target configuration data. At least the part of the target configuration data may be provided by the data source in response to receiving the request for retrieving at least a part of the target configuration data. The request for retrieving the target configuration data may be related to the request, in particular to a vectorized request. The vectorized request may be a numerical, i.e. machine- interpretable, representation associated with the request. For example, the request may be provided to one or more embedding layers as described within the context of FIG. 7. The one or more embeddings layers may be configured for generating and / or providing the vectorized request. The vectorized request may be provided to the data source. The data source may be configured for providing at least a part of the target configuration data in response to receiving the vectorized request. In particular, the data source may be configured for determining a distance between the vectorized request and vectorized target configuration data. The vectorized target configuration data may be a numerical representation of the target configuration data. The data source may be configured for providing at least a part of target configuration data where the distance between the vectorized request and at least the part of the vectorized target configuration data may be within a predefined range. Additionally or alternatively, the request for retrieving at least the part of the target configuration data may comprise providing a structured query configured for triggering the data source to provide at least the target configuration data. The structured query may be indicative of the current plant data. Providing the structured query to the data source may result in matching the target plant data associated with the target configuration data and the current plant data.

[0101] It may be determined if the current configuration data and at least the selected part of the target configuration data match 616. Hence, a difference between the current configuration data and the target configuration data may be determined. This may include providing the current configuration data and at least the selected part of the target configuration data to a data-driven model. The data-driven model may be configured for matching configuration data, in particular for determining a difference between two sets of configuration data. The data-driven model may be provided with vectorized target configuration data and vectorized current configuration data. Vectorized configuration data may be obtained by providing the configuration data to the one or more embedding layers as described within the context of FIG. 7. In particular, the data-driven model may be provided by matching task instruction related to the target configuration data and the current configuration data. The matching task instruction may be obtained by merging the target configuration data and the current configuration data. The data- driven model may be configured for mapping the vectorized matching task instruction to contextualized matching task instruction and / or mapping the contextualized matching task instruction to a vectorized indication on a verification of the current configuration data where at least the selected part of the target configuration data may match the current configuration data or a vectorized subset of target configuration data associated with the one or more chemical plant(s), wherein the subset of target configuration data may deviate from the current configuration data where at least the part of the target configuration data deviates from the current configuration data. The contextualized matching task instruction may be a numerical representation of the matching task instruction and a relation between two or more parts of the matching task instruction. Hence, the contextualized matching task instruction may be a dense representation of data points and their relation to each other. Followingly, the mapping provides the advantage to take the relation between single datapoints associated with the matching task instruction into account. The contextualized matching task instruction may be obtained by providing the vectorized matching task instruction to one or more encoder block(s) and / or one or more decoder block(s) as described in detail within the context of FIG. 8A - FIG. 8C. The one or more encoder block(s) and / or decoder block(s), optionally further the one or more encoder output(s) and the one or more decoder output(s) may be configured for mapping the contextualized matching task instruction to the vectorized indication or the vectorized subset.

[0102] Further, the vectorized indication and / or the vectorized subset may be mapped to the indication and / or the subset. For this purpose, the vectorized indication and / or the vectorized subset may be provided to a decoding engine. The decoding engine may be configured for mapping the vectorized subset and / or the vectorized indication to the subset and / or the indication. The decoding engine may be configured for mapping a vectorized representation associated with data to the data. The decoding engine may have obtained a relation between vectorized representations associated with data and the data. The decoding engine may be provided with a lookup table indicative of a relation between the vectorized representations associated with the data and the data.

[0103] The indication on the verification of the current configuration data and / or the subset of target configuration data deviating from the current configuration data may be provided 620 and / or 618, e.g. via a user interface. Additionally or alternatively, the indication on the verification of the current configuration data and / or the subset of target configuration data different from the current configuration data may be provided to one or more chemical production facilities 102, e.g. for operating the one or more chemical production facilities 102 according to the current configuration data where the indication may be provided, or according to the subset of target configuration data where the subset may be provided.

[0104] In an embodiment, providing the indication and / or the subset may include providing the request to the data-driven model for generating a template indicative of a structure associated with the indication and / or the subset. This may include mapping the request to a vectorized request and providing the vectorized request to the data-driven model. The vectorized request may be a numerical representation of the request. The data-driven model may be configured for mapping the vectorized request to contextualized request. The contextualized representation may be a numerical representation of the request and a relation between two or more parts associated with the request. The data-driven model may be configured for mapping the contextualized request to the vectorized template. The vectorized template may be mapped to the template by the decoding engine. Further, providing the indication and / or the subset may include merging the subset and / or the indication with the template. Hence, providing the subset and / or the indication may include providing the subset and / or the indication together with the template, in particular included in the template according to the structure indicated by the template.

[0105] FIG. 7 illustrates an embodiment of obtaining an embedding layer. The embedding layer may be obtained by training for example a continuous bag of words model (CBOW) or a skip-gram model. The embedding layer may be suitable for generating embedded input data based on input data. Generating embedded input data may refer to embedding input data. Embedding input data may result in a representation associated with the input data. Thus, the embedded input 714 may be the representation associated with the input data. The input data may comprise one or more elements. The one or more elements may be represented by the input vector 706. In particular, the embedded input 714 and / or the input vector 706 may be machine- readable and / or processable by a processor. For this purpose, the embedded input 714 and / or the input vector 706 may be a tensor, in particular a first-rank tensor. Specifically, the input vector 706 may be a one-hot vector or a summation of a plurality of one- hot vectors. A one-hot vector may be a vector with one entry unequal to zero. Examples for one-hot vectors may be 708, 710 and 712. The entries unequal to zero in the one-hot vector and / or in the input vector 706 may indicate the element. For example, a lookup table may define the relation between the position of the entries unequal to zero and the element indicated by the one-hot vector. The lookup table may specify a plurality of different elements. The number of different elements may be equal to the number of entries in the one-hot vector. The number of different elements may be referred to as vocabulary size. I n an example, the elements may be represented by tokens and a sequence of elements may refer to at least a part of a sentence. The at least a part of the sentence may be represented by a plurality of tokens. A token may represent at least a part of the element and / or word. For example, where one element would be associated with only one word, words such as “embeddings", “embedding” or “embed” would constitute different elements. A first token may represent the stem “embed” and the endings, typically appearing in a plurality of word, may be represented by a second token, a third token and a fourth token. The second token, the third token and the fourth token may be used for representing other words such as “look”, “looking” or the like, preferably together with a fifth token representing the stem “look”. Ultimately, this tokenization of elements associated with a plurality of stems and a plurality of endings results in less tokens to be used for representing a plurality of elements and thus, uses less computational resources. A lookup table specifying a subset of the vocabulary size e.g. of the English language may comprise 10,000 words or more. The embedded input 714 may be a lower-dimensional representation than the input vector 706. For example, typical embedded inputs 714 may comprise some hundreds of different entries. Followingly, the embedded inputs 714 constitute a densified representation of one or more elements using less computational resources. More than that, the embedded input 714 may represent a relation between two or more elements. For example, the words “Italy” and “Germany” may be similar or may be more closely related since they both define European countries, whereas the word “embodiment” may be very different from the two respective words. The smaller the dot product between two embedded inputs 714 may be the more similar the two elements associated with the embedded inputs 714 may be. Hence, the embedded inputs 714 may represent one or more elements accurately and lead to accurate results based on processing the embedded inputs 714.

[0106] For transforming the input vector 706 into the embedded input 714, the embedding layer may comprise a number of neurons equal to the number of entries in the embedded input 714. Based on the embedded inputs 714, the output layer may generate the output vector 716. The output vector may be a vector and / or may indicate one or more elements. The output vector 716 may indicate one or more elements different from the input vector 706 and / or the one-hot vectors associated with the input vector 706. For this purpose, the output layer may comprise a number of neurons equal to the number of entries of the input vector 706 and / or the output vector 716. The output layer may apply a softmax function to the embedded inputs 714. By doing so, the output vector may comprise the probabilities associated with the elements associated with the entries of the output vector 716 unequal to zero. Hence, from the output vector 716 one or more elements may be obtained with a corresponding probability. Where the input vector 706 may specify one or more sequence(s) of elements, the output vector 716 may specify one or more elements corresponding to the sequence(s) of elements specified by the input vector 706. In the example of FIG. 7, the element associated with vector 718 may correspond to the input vector with a probability of 71 %. Additional or alternative elements may correspond to the input vector as indicated by the output vector with lower probability. By defining a threshold to which the probability may be compared, the selection of the corresponding elements may be tailored to the needs of the user. The elements generated by the model comprising the embedding layer 702 and the output layer 704 may refer to the most probable elements indicated by the output vector 716. Hence, the model depicted in FIG. 7 may generate the element associated with the vector 718 with a confidence score of 71 %.

[0107] The model of FIG. 7 may be continuous bag of words (CBOW) model. The CBOW model may be trained based on a training data set comprising a plurality of input vectors and corresponding output vectors. As the training data set may not be labeled, the training of the CBOW model may be referred to as self-supervised. Before training of the CBOW model, the CBOW model may be initialized with random values assigned to the weights of the neurons. During the training of the CBOW model, the input vectors may be passed through the initialized embedding layer and the output layer and a loss may be determined by comparing the output vector obtained by passing the input vector 706 through the model to the output vector corresponding to the input vector 706 as specified by the training data set. Based on the determined loss, backpropagation may be applied to determine the gradients associated with the neurons of the embedding layer 702 and the output layer 704 to lower the loss. According to the determined gradients, the weights of the neurons may be updated by using a gradient descent algorithm. If a predetermined loss may be achieved by the CBOW model, the training may be terminated and a trained CBOW model may be obtained. From the trained CBOW model, the embedding layer 702 may be suitable for embedding input data comprising one or more elements. This embedding layer 702 may be used in other machine-learning architectures requiring an embedding layer 702 such as a transformer encoder, transformer decoder or transformer encoder decoder architecture as described within the context of FIG. 8A, FIG. 8B and FIG. 8C. For training these architectures, a trained embedding layer 702 may be required. Hence, a model such as a CBOW model may be trained prior to training the transformer encoder, transformer decoder or transformer encoder decoder architecture.

[0108] FIG. 8A illustrates an embodiment of a transformer encoder architecture. The transformer encoder comprises an encoder input 878, one or more encoder blocks 874, 814 and an encoder output. The transformer encoder architecture may be derived from the transformer encoder-decoder architecture as known in the art and shown in FIG. 8C. In particular, the transformer encoder may be referred to as X-former. The transformer encoder architecture may correspond to the encoder architecture associated with the transformer encoder-decoder architecture with an additional encoder output instead of connecting the encoder block directly to the decoder of the transformer encoder-decoder architecture. A plurality of transformer encoder architectures are available in the art such as the bi-directional encoder representations from transformers (BERT).

[0109] The input data may be received at the encoder input 878. The encoder input 878 may apply an input embedding 802. Applying the input embedding 802 may refer to passing the input data through an embedding layer e.g. as described within the context of FIG. 7. Further, the encoder input 878 may apply positional encoding 804. Applying positional encoding 804 may refer to adding a positional factor to the embedded input obtained via input embedding. Preferably, the input data may specify a sequence of elements.

[0110] The positional factor Pp° may be indicative of the position of the elements within the sequence. For example, the positional factor may be obtained based on the following equation: where pos may refer to the position of the element within the sequence, / may refer to the dimension associated with the input embedding and d may refer to the dimension of the model, e.g. transformer decoder, transformer encoder or transformer encoder-decoder. This may be referred to as absolute positional embeddings. Alternatively, the positional encoding may be based on rotary positional embeddings (RoPE). Positional encoding is beneficial since it enables the processing of sequential data without requiring further dimensions indicating the position of each element. Followingly, the positional encoding 804 reduces the computational resources needed for embedding the input data. By passing the input data through the encoder input, the input data may be transformed into a second-rank tensor representing the sequence of elements. This second-rank tensor may be referred to as embedded input data. The embedded input data may be processed by the encoder block. The embedded input data may be provided to the layer normalization 808 by a residual connection. Multi-head self-attention 806 may be applied to the embedded input data. Multi-head self-attention 806 may comprise the two components multi-head and self-attention. Self-attention may be understood as being a filter applied to the embedded input data. By applying the filter to the embedded input data, the elements associated with the embedded input data contributing to the to be generated output data may be identified for generating the output data. Hence, the filter may represent the degree of contributing to the to be generated output data by the elements associated with the embedded input data. Applying the filter may be referred to as weighting the elements associated with the embedded input data. This is advantageous specifically regarding long sequences of elements. The filter may be learned and improved during the training by learning to identify the contribution of elements associated with the embedded input data. For example, in the partial sentence “I went to the bakery to buy a” the last word may be generated by the data-driven model such as the transformer encoder. The self-attention may focus the transformer encoder to attend to the word “bakery” and “buy” mostly to generate the word “bread”. Self-attention may refer to attention generated based on the input data. Hence, the filter may be determined based on the input data, preferably the embedded input data. The embedded input data may serve as query Q, key K and value V with respect to the self-attention operation. The self-attention may refer to attention based on the received input data. Hence, the filter may be calculated based on the following formula by inserting the respective tensors based on the embedded input data: where dk corresponds to the dimension of the key.

[0111] For improving the efficiency of the transformer encoder further, the multiple heads are used to apply the filter resulting in the multi-head self-attention 806. Multi-head self-attention 806 may comprise applying the filter to two or more parts of the embedded input data. Hence, the tensor may be split into two or more parts and the filter may be applied to the two or more parts separately by two or more heads according to the following equation:head i= Attention QWiQ, KW , VWtv) with parameter matrices where i may refer to the number of heads, may refer to the dimensions of the value, key and query.

[0112] The result of the two or more head may be concatenated according to the following equation: MultiHead(Q, K, V) = Concat(head 1, . . . , headh) W° whereeand h may refer to the number of heads.

[0113] The embedded input data may be transformed via the multi-head self-attention 806 into a context tensor. The context tensor may represent the sequence of elements and the relation between two or more elements of the input data. The context tensor may be a second rank tensor and / or may comprise one or more first rank tensor(s). After the multi-head self-attention 806 layer normalization 808 may be applied based on the context tensor and / or the embedded input data from the residual connection. Applying layer normalization 808 may refer to normalizing the context tensor. Normalizing the context tensor may lower the values of the entries of the context tensor. This reduces the computational cost associated with processing the context tensor. Further, it improves the training by contributing the loss to converge and preventing instabilities.

[0114] Layer normalization 808 may be followed by passing the context tensor to a feed-forward layer 810 again followed by layer normalization 812 based on the residual connection to the context tensor and / or the output of the feedforward layer 810. The feed-forward layer 810 may be a feed-forward neural network. The feed-forward neural network may comprise of a plurality of fully connected neurons. Passing the context tensor through the feedforward neural network may result in transforming the context tensor linearly. Additionally or alternatively, the neural network may comprise one or more activation functions such as a rectified linear unit (ReLU). Hence, the neural network may be configured for performing one or more non-linear operations to the context tensor and / or transforming the context tensor non-l inearly . After the context tensor has been transformed and / or normalized by the feed-forward layer 810 and the layer normalization 812, the context tensor may be provided to one or more further encoder blocks 814. Having passed the context tensor through the feed-forward layer 810 may adapt the context tensor for the processing by a further attention layer of the one or more further encoder blocks 814 for applying a self-attention filter, preferably multi-head self-attention 806. The context vector after being transformed by the layer normalization 812 and the feed-forward layer 810 may be referred to as hidden state.

[0115] The encoder output 876 comprises of a linear layer 816 and a softmax layer 818. The linear layer 816 may transform the context vector into a logits vector. The linear layer may be fully-connected. The logits vector obtained by passing the context tensor through the linear layer 816 may be passed through the softmax layer 818. Passing the logits vector through the softmax layer 818 may refer to applying the softmax function to the logits vector. Applying the softmax function to the logits vector may result in a probability distribution of one or more elements corresponding to the sequence of elements in the input data. From the probability distribution based on predefined selection criteria, one or more elements may be chosen. The one or more chosen elements may be referred to as the one or more elements generated by the transformer encoder. The one or more generated elements may be provided to the encoder input for generating further one or more elements corresponding to the sequence of the input data and the one or more elements generated by the transformer encoder as described within the context of FIG. 9.

[0116] FIG. 8B illustrates an embodiment of a transformer decoder architecture.

[0117] The transformer decoder comprises a decoder input 884, one or more decoder blocks 880, 832 and a decoder output 892. The transformer decoder architecture may be derived from the transformer encoder-decoder architecture as known in the art and shown in FIG. 8C. The transformer decoder may be referred to as X-former. The transformer decoder architecture may correspond to the decoder architecture associated with the transformer encoder-decoder architecture independent of receiving one or more hidden states from the encoder of the transformer encoder-decoder. A plurality of transformer decoder architectures are available in the art such as the generative pretrained transformers (GPT).

[0118] The decoder input 884 may apply input embedding 820 and positional encoding 822 analogous to analogous to the input embedding 802 and the positional encoding 804 as described within the context of FIG. 8A.

[0119] The decoder block 880 may comprise the layer normalizations 826, the masked multi-head self-attention 824, the feed-forward layers 828 and / or the layer normalization 830. The embedded input data resulting from passing the input data through the decoder input 884 may be provided to the layer normalization 826 via a residual connection. Further, masked multi-head self-attention 824 may be applied to the embedded input data. Masked multi-head self-attention 824 corresponds to the multi-head self-attention 806 as described within the context of FIG. 8A with additionally masking a part of the embedded input data associated with elements later in the sequence than the element to be generated. Additionally or alternatively, the part of the input data associated with elements later in the sequence than the element to be generated may not be received and / or transformed into the embedded input data. Thus, the transformer decoder may be suitable for generating a subsequent element to a sequence, whereas the transformer encoder may be suitable for generating a missing element in within one sequence and / or between two or more sequences. Therefore, the transformer encoder may be configured for classification tasks. The transformer decoder may be configured for text generation. Similar to the transformer encoder as described within the context of FIG. 8A, a context tensor may be generated by applying the masked multi-head self-attention 824 and the layer normalization 826. The context tensor may be provided to the layer normalization 830 via a residual connection. Further, the feed-forward layer 828 and the layer normalization 830 may be analogous to the feed-forward layer 810 and the layer normalization 812 as described within the context of FIG. 8A. The context tensor may be provided to one or more further decoder blocks 832.

[0120] The decoder output 892 may comprise of a linear layer 834 and a softmax layer 836. The linear layer 834 and the softmax layer 836 may be analogous to the linear layer 816 and the softmax layer 818 as described within the context of FIG. 8A.

[0121] FIG. 8C illustrates an embodiment of a transformer encoder-decoder architecture. The transformer encoderdecoder may comprise the encoder input 888, the one or more encoder blocks 886, 864, the decoder input 894, the decoder block 890 and the decoder output 892. The encoder input 888 may correspond to the encoder input 878 of FIG. 8A. The one or more encoder block 886, 864 may correspond to the one or more encoder blocks 874, 814 of FIG. 8A. The decoder input 894 may correspond to the decoder input 884 of FIG. 8B.

[0122] The decoder block 890 may comprise a masked multi-head self-attention 870, a layer normalization 872, a feedforward layer 838 and a layer normalization 840 analogous to the masked multi-head self-attention 824, the layer normalization 826, the feed-forward layer 828 and the layer normalization 830 as described within the context of FIG. 8B. The decoder block 890 may further comprise a multi-head self-attention 850 and a layer normalization 848. Analogous to the description of FIG. 8B, the context tensor may be obtained from the masked multi-head self-attention 870 and the layer normalization 872. Multi-head self-attention 850 analogous to the multi-head selfattention 806 of FIG. 8A may be applied to the context vector obtained from the layer normalization 872 and the hidden states of the one or more encoder blocks 886, 864. Layer normalization 848 may be applied to the context vector obtained from the multi-head self-attention 850 and the context vector obtained from the layer normalization 872 provided via a residual connection. The context vector resulting from the layer normalization 848 may be processed via the feed-forward layer 838 and the layer normalization 840 analogous to the description of FIG. 8B. The context vector resulting from the layer normalization 840 may be provided to further decoder blocks 842 analogous to the decoder block 890. The context vector obtained from the one or more decoder blocks 890, 842 may be provided to the decoder output 892. The decoder output 892 may correspond to the decoder output 882 of FIG. 8B. With the above-described architecture, the transformer encoder-decoder may receive and process input data at the encoder input 888 and the one or more encoder blocks 886, 864 and the decoder block 890 and the decoder output 892. Based on the input data, the transformer encoder-decoder may generate output data part by part or sequentially. The sequentially generated output data may be provided to and / or may be processed by the decoder input 894, the one or more decoder blocks 890, 842 and the decoder output 892. Preferably, a sequence may be provided to the encoder input 888 and after having generated at least a part of the output data, the decoder input 894 may be provided with at least the part of the elements of the output data already generated. By doing so, the next elements of the output data may be generated with a higher accuracy by taking the input data and the generated output data into account since more data is received by the transformer encoder-decoder may be received over time.

[0123] Because of the transformer encoder-decoder architecture, the transformer encoder-decoder may be configured for transforming a sequence into another representation of the sequence. An example for transforming one sequence into another representation may be translation of one sentence into another language. A plurality of transformer encoder-decoders are available in the art such as BART, T5 or the like.

[0124] In an embodiment, the layer normalization 808, 812 may be applied prior to the masked multi-head self-attention 824, multi-head self-attention 806 and / or the feed-forward layer 810 in the transformer decoder, the transformer encoder and / or the transformer encoder-decoder. By doing so, the computational resources for applying the multihead self-attention 806 and / or the feed-forward layer 810 to the embedded input data and / or the context tensor may be decreased as the entries of the respective tensors may be lower after normalization.

[0125] In an embodiment, the decoder output 892 may comprise of a classification neural network, further feedforward layers, convolutional layers, fully connected layers or the like. For example, the transformer encoder-decoder may be configured for choosing between a plurality of options. For this purpose, the transformer encoder-decoder may be provided with three different input data sets and may classify the context vectors obtained from the one or more decoder blocks 890 via one or more linear layers. Followingly, the architecture may be extended depending on the use case to be solved. [1]

[0126] FIG. 9 illustrates an embodiment of training and / or deploying the transformer encoder, the transformer decoder and / or the transformer encoder-decoder.

[0127] The encoder / decoder / encoder-decoder architecture 902 may correspond to the transformer decoder, the transformer encoder and / or the transformer encoder-decoder as describe within the context of FIG. 8A- FIG. 8C. The output data generated by the encoder / decoder / encoder-decoder architecture 902 may comprise of one or more elements, in particular a sequence of elements. The previously generated elements of the output data may be provided as input for generating the next element in the sequence of the output data.

[0128] In the example of FIG. 9, the input data may comprise of N elements, in particular input tokens. An input token may be a token dedicated to be inputted into a data-driven model such as the transformer decoder, the transformer encoder or the transformer encoder-decoder. The output data to be generated may comprise of M elements. The encoder / decoder / encoder-decoder architecture 902 may generate one element of the output data based on receiving the input data and optionally previously generated elements of the output data at a timestep. Hence, for generating M elements M time steps are required. A time step comprises of providing input 910, 912, 914 to the encoder / decoder / encoder-decoder architecture 902 and receiving output data 904, 908, 906 from the encoder / decoder / encoder-decoder architecture 902. In a first timestep, the input 910 may comprise of N input tokens. The N input tokens may be associated e.g. with N words, stems or endings. Preferably, the N input tokens may specify a question. One or more input tokens may specify the beginning of the sequence of tokens and / or the end of the sequence of tokens. The input 910 may be processed by the encoder / decoder / encoder-decoder architecture 902. Based on the input 910 at least a part of the output data 904 may be generated. The at least a part of the output data may comprise a first output token. In the next timestep, the generated first output token may be provided together with the input 912. Specifically, where the input 912 may be received by a transformer encoder-decoder the input tokens may be received at the encoder input 888 and the first output token may be received at the decoder input 894. Where the input 912 may be received by the transformer encoder, the input 912 may be received by the encoder input 878 and analogously regarding the transformer decoder and the decoder input 884. Based on the input 912, the output data 908 comprising the first output token and a second output token may be generated. Generating the output data 908 based on the input 912 may refer to generating the second token based on the first token and the N input tokens, wherein the first token may have been generated based on the N input tokens. This process may be repeated until the last token in the sequence of the output data 906 may be generated. Preferably, the last token may be an end token. The end token may terminate the generation of a further output token.

[0129] Similarly, to the data processing during deployment of the encoder / decoder / encoder-decoder architecture 902, the encoder / decoder / encoder-decoder architecture 902 may be trained. The training data set may comprise a plurality of sequences comprising a plurality of elements. The sequences may be associated with the input data and / or the output data. Additionally or alternatively, the sequences may be independent of the input data and / or the output data. For example, where the input data and the output data may refer to chemical compositions represented via text, the training data set may comprise sequential text data independent of chemical compositions. In this example, the training data set may comprise sequences of words originating from a conversation. In an embodiment, the training data set may comprise at least partially input data sets and / or output data sets. The training may be initialized by initializing the encoder / decoder / encoder-decoder architecture 902. In an embodiment, the parameters associated with the encoder / decoder / encoder-decoder architecture 902 may be initialized randomly. Additionally or alternatively, the input embedding of the encoder / decoder / encoder-decoder architecture 902 may be obtained by training a CBOW model or a skip gram model as described within the context of FIG. 7. The trained embedding layer may be used during training. The parameters associated with the embedding layer may be kept constant and / or may be updated after a predefined number of training epochs. By doing so, the number of parameters to be updated is lower enabling a faster and less computational resources- consuming training. Further, the accuracy associated with the embedding layer may be constant and / or may be increased by avoiding error compensation in relation to the just initialized encoder / decoder / encoder-decoder architecture 902.

[0130] During the training of the encoder / decoder / encoder-decoder architecture 902, at least a part of the sequences of the training data set may be provided to the encoder / decoder / encoder-decoder architecture 902 one by another and one or more elements may be generated based on the sequences of the training data set one by another. The elements generated based on the sequences may follow the elements of the parts of sequences the encoder / decoder / encoder-decoder architecture 902 may have been provided with. The generated one or more elements may be compared to the one or more elements following the at least a part of the sequences provided to the encoder / decoder / encoder-decoder architecture 902 as specified by the training data set. Hence, during the training the encoder / decoder / encoder-decoder architecture 902 may generate a guess on the next element and the guess on the next element in a sequence may be compared to the ground truth specifying the actual next element according to the training data set. Based on the guess on the next element and the ground truth a loss may be determined. The loss may define the similarity between the guess on the next element and the ground truth. The loss may be determined by forming a vector dot product between the token associated with the one or more elements and the token associated with the ground truth. A loss unequal to zero may result in updating the parameters associated with encoder / decoder / encoder-decoder architecture 902. Preferably the parameters associated with the encoder / decoder / encoder-decoder architecture 902 may be independent of the embedding layer. For example, the parameters associated with the encoder / decoder / encoder-decoder architecture 902 may be weights of the neurons of the encoder / decoder / encoder-decoder architecture 902.

[0131] Based on the determined loss, backpropagation may be applied to determine the gradients associated with the parameters of the parameters associated with encoder / decoder / encoder-decoder architecture 902 to lower the loss. According to the determined gradients, the parameters associated with the encoder / decoder / encoder- decoder architecture 902, preferably the weights of the neurons associated with the encoder / decoder / encoder- decoder architecture 902, may be updated by using a gradient descent algorithm.

[0132] The training data set may be unlabeled. The sequences of elements within the training data set may inherently comprise the ground truth for determining the loss with respect to the one or more elements generated during the training of the encoder / decoder / encoder-decoder architecture 902. Hence, the encoder / decoder / encoder-decoder architecture 902 may be trained self-supervised. This is advantageous since time and resources for creating a labeled training data set may be saved. Furthermore, this enables the usage of large training data sets associated with a size of several tera bytes. Consequently, the data-driven model may be accurate in generating elements of a sequence. In addition, the large training data set enables few shot predictions or even zero shot predictions. Hence, the data-driven models trained as described above are versatile contributing to saving resources needed for training and / or hosting a plurality of purpose-driven models such as convolutional neural networks. The training described above may be referred to as pretraining. The data-driven model may be configured for performing few shot or even zero shot predictions with respect to a plurality of use cases after pretraining. The performance of the data-driven model may be increased further by additional training referred to as finetuning.

[0133] FIG. 10 illustrates an embodiment of input embedding. Where the sequence of elements associated with the input data, preferably comprised in the input data, may be of one type, the input embedding 802, 820, 852, 866 as described within the context of FIG. 8A - 2C may be used. For example, a type of input data may be text where the elements may be associated with at least a part of a word, a punctuation character, a start token specifying the beginning of one or more sequences associated with the input data and / or the end token. In another example, the input data may be at least partially numerical. Hence, the input data may comprise a plurality of numbers. Numerical input data may be for example tabular data. Tabular data may specify one or more rows and / or one or more columns. Hence, the tabular data may comprise one or more cells, wherein the cells may be associated with one or more numerical values.

[0134] Numerical input data may require a different embedding than text input data. Input embeddings for numerical input data may comprise a token embedding, a positional embedding, a column embedding, a row embedding or a combination thereof.

[0135] Applying a token embedding to one or more elements, in particular tokens associated with the input data may result in a machine-processable representation associated with the one or more elements, in particular tokens. Applying the token embedding to one or more elements may refer to passing the one or more elements through the embedding layer, e.g. as described within the context of FIG. 7. Hence, token embeddings may specify the one or more elements, in particular tokens in a machine-processable representation. For example, the token embedding may transform a numerical value into a vector. This is advantageous since this representation can be enriched by further information such as the position of the token within the sequence and / or within a table associated with the sequence of tokens. The positional embedding may be analogous to the positional embedding as described within the context of FIG. 7, FIG. 8A-2C. Where the input data may be tabular data, column embedding may be applied. Applying a column embedding to one or more elements, in particular tokens associated with the input data may result in a machine-processable representation specifying the location of the one or more elements within a table 1002, preferably within the columns of the table 1002. Applying the column embedding may refer to adding a column factor to the input data embedded via token embeddings, in particular the embedded input data. The column factor may be the same for elements associated with the same column and / or may differ between two or more elements associated with different columns. Analogous, row embeddings may be applied where the input data may be tabular data. Applying a row embedding to one or more elements, in particular tokens associated with the input data may result in a machine-processable representation specifying the location of the one or more elements within a table 1002, preferably within the rows of the table 1002. Applying the row embedding may refer to adding a column factor to the input data embedded via token embeddings, in particular the embedded input data. The row factor may be the same for elements associated with the same row and / or may differ between two or more elements associated with different rows.

[0136] In an embodiment, input data may be at least partially numerical and at least partially text. Hence, the input data may comprise two or more types of data. A type of data may refer to a modality. Followingly, different embeddings may be applied to the input data. To parts of the input data comprising text the input embedding referred to in FIG. 7, FIG. 8A-2C may be applied. To parts of the input data being numerical token embeddings, positional embeddings, column embeddings and row embeddings may be applied. Further, segment embeddings may be applied to the input data independent of the type of input data. The segment embedding may specify the type of input data one or more elements may be associated to. For example, if the input data comprises of text and numbers, the input data may comprise of two types of input data. Applying the segment embedding to the input data may refer to adding a segment factor to the input data, preferably the embedded input data and / or the input data after having applied the token embedding. The segment factor may specify the type of data associated with the one or more elements. The segment factor may be the same for one or more elements associated with the same type of input data and / or may differ between two or more elements associated with different types of input data.

[0137] Applying the token embedding, the positional embedding, the segment embedding, the column embedding, the row embedding or a combination thereof may result in embedded input data and / or may be the output of any one of the encoder input 878, 884, 888 or decoder input 884, 894. The data obtained by applying the token embedding, the positional embedding, the segment embedding, the column embedding, the row embedding or a combination thereof may be processed by the encoder block 874, 886, decoder block 880, 890, encoder output 876, decoder output 892, 882.

[0138] FIG. 1 1 illustrates an embodiment of input embedding.

[0139] Input data to the data-driven model, in particular to the encoder input and / or the decoder input as described in the context of FIG. 8A-C, may comprise image data. The data-driven model may be parametrized to receive image data. For processing image data as input data, the data-driven model may comprise one or more encoder blocks and / or one or more decoder blocks and / or one or more encoder outputs and / or one or more decoder outputs as described within the context of FIG. 8A-C. FIG. 11 may show an embodiment of an encoder input and / or a decoder input. When processing image data, the encoder input and / or the decoder input of the data-driven model may be as described within the context of FIG. 1 1 . The encoder input and / or decoder input may comprise one or more linear projection layers 11 14 for a linear projection of one or more images, preferably one or more partial images, more preferably a sequence of two or more partial images. The one or more linear projection layers 11 14 may be suitable for changing the dimension of the one or more received images, preferably one or more partial images, preferably passing the one or more images, preferably partial images, through the one or more linear projection layers 1 114 may result in applying image embedding, preferably partial image embedding to the one or more images and / or partial images.

[0140] Furthermore, when a sequence of two or more images and / or partial images may be received, positional embedding may be applied to the sequence, preferably by passing the sequence of one or more images and / or partial images through the one or more linear projection layers 1114. Applying positional embedding may refer to adding a positional factor. The positional factor may be different depending on the position of the image and / or the partial image within the sequence. In particular, the positional factor added to a first element of the sequence may be different to the positional factor added to a second element of the sequence. The first element of the sequence may be a first image and / or first partial image. The second element of the sequence may be a second image and / or a second partial image.

[0141] The representation of the one or more images, preferably one or more partial images, may be obtained based on the following following equation: wherexciass is the image class embedding,x^ is the n-th image, in particular partial image in the sequence,ZQ is the representation of the one or more images, preferably one or more partial images, (H,W) are the resolution of the image, in particular the image the partial images are generated on, C is the number of channels associated with the one or more image, in particular the one or more partial images and D is the dimension of the representation of the one or more images, preferably one or more partial images. Applying the partial image embedding may refer to forming the product ofx^ with E above-described equation. Applying the positional embedding may refer to adding the factor pos according to the above-described equation.

[0142] By doing so, text-based data, numerical data, tabular data, image data or the like may be processed by one data- driven model.

[0143] The present disclosure has been described in conjunction with preferred embodiments and examples as well. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed subject-matter, from the studies of the drawings, this disclosure and the claims. Notably, in particular, the any steps presented can be performed in any order, i.e. the present disclosure is not limited to a specific order of these steps. Moreover, it is also not required that the different steps are performed at a certain place or at one node of a distributed system, i.e. each of the steps may be performed at different nodes using different equipment / data processing.

[0144] As used herein ..determining" also includes ..initiating or causing to determine", “generating" also includes ..initiating and / or causing to generate" and “providing” also includes “initiating or causing to determine, generate, select, send and / or receive”. “Initiating or causing to perform an action” includes any processing signal that triggers a computing node or device to perform the respective action.

[0145] In the claims as well as in the description the word “comprising” or “including” or similar wording does not exclude other elements or steps and shall not be construed limiting to the elements or steps lined out. The indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation or further elements may be included.

[0146] Providing in the scope of this disclosure may include any interface configured to provide data. This may include an application programming interface, a human-machine interface such as a display and / or a software module interface. Providing may include communication of data or submission of data to the interface, in particular display to a user or use of the data by the receiving entity. Various units, circuits, entities, nodes or other computing components may be described as “configured to” perform a task or tasks. Configured to shall recite structure meaning “having circuitry that” performs the task or tasks on operation. The units, circuits, entities, nodes or other computing components can be configured to perform the task even when the unit / circuit / component is not operating. The units, circuits, entities, nodes or other computing components that form the structure corresponding to “configured to” may include hardware circuits and / or memory storing program instructions executable to implement the operation. The units, circuits, entities, nodes or other computing components may be described as performing a task or tasks, for convenience in the description. Such descriptions shall be interpreted as including the phrase “configured to.” Any recitation of “configured to” is expressly intended not to invoke 35 U.S.C. § 112(f) interpretation.

[0147] In general, the methods, apparatuses, systems, computer elements, nodes or other computing components described herein may include memory, software components and hardware components. The memory can include volatile memory such as static or dynamic random-access memory and / or nonvolatile memory such as optical or magnetic disk storage, flash memory, programmable read-only memories, etc. The hardware components may include any combination of combinatorial logic circuitry, clocked storage devices such as flops, registers, latches, etc., finite state machines, memory such as static random-access memory or embedded dynamic random-access memory, custom designed circuitry, programmable logic arrays, etc.

[0148] Any disclosure and embodiments described herein relate to the methods, the systems, devices, the computer program element lined out above and vice versa. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples and vice versa.

[0149] All terms and definitions used herein are understood broadly and have their general meaning.

[0150] The present disclosure has been described in conjunction with preferred embodiments and examples as well. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed invention, from the studies of the drawings, this disclosure and the claims. Notably, in particular, the any steps presented can be performed in any order, i.e. the present invention is not limited to a specific order of these steps. Moreover, it is also not required that the different steps are performed at a certain place or at one node of a distributed system, i.e. each of the steps may be performed at different nodes using different equipment / data processing. As used herein ..determining" also includes ..initiating or causing to determine", “generating" also includes ..initiating and / or causing to generate" and “providing” also includes “initiating or causing to determine, generate, select, send and / or receive”. “Initiating or causing to perform an action” includes any processing signal that triggers a computing node or device to perform the respective action.

[0151] In the claims as well as in the description the word “comprising” does not exclude other elements or steps. The indefinite article “a” or “an” and the definite article “the” does not exclude a plurality. In particular, indefinite article “a” or “an” may be replaced with one or more and the definite article “the” may be replaced with the one or more. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation.

[0152] Any disclosure and embodiments described herein relate to the methods, the systems, devices, the computer program element lined out above and vice versa. Advantageously, the benefits provided by any of the embodiments and examples equally apply to all other embodiments and examples and vice versa.

Claims

CLAIMSWhat is claimed is:1 . A method for identifying operation risks associated with operating one or more target chemical plant(s), the method comprising: providing a request for identifying operation risks associated with operating the one or more target chemical plant(s), wherein the request includes an indication on the one or more target chemical plant(s), providing unstructured data related to a plurality of operation risks associated with operating one or more chemical plant(s) including the one or more target chemical plant(s), providing operation risk task instructions for identifying at least one operation risk associated with the one or more target chemical plant(s) to one or more data-driven model(s), wherein the operation risk task instructions are related to the plurality of operation risks and the request for identifying the one or more operation risk(s), and wherein the one or more data-driven model(s) are configured to follow the provided task instructions, providing the at least one identified operation risk.

2. The method of claim 1 , wherein the indication on the one or more target chemical plant(s) includes image data related to a representation of a configuration of the one or more target chemical plant(s).

3. The method of claim 1 or 2, wherein the operation risk task instructions, in particular the plurality of operation risks, comprise unstructured data and wherein the one or more data-driven model(s) are configured to process unstructured data.

4. The method of any one of claims 1 to 3, wherein providing the operation risk task instructions comprises mapping the operation risk task instructions to a numerical representation of the operation risk task instructions, and wherein the one or more data-driven model(s) are configured to map the numerical representation of the operation risk task instructions to at least one numerical representation of the operation risk, and wherein providing the at least one identified operation risk, in particular by the one or more data-driven model(s), includes mapping the at least one numerical representation of the operation risk to the at least one operation risk.

5. The method of any one of claims 1 to 4, wherein the request is related to and / or indicative of a change of a configuration of the one or more target chemical plant(s), and wherein identifying at least one operation risk associated with the one or more target chemical plant(s) comprises identifying at least one operation risk associated with the change of the configuration of the one or more target chemical plant(s).

6. The method of any one of claims 1 to 5, wherein providing the request for identifying operation risks associated with operating the one or more target chemical plant(s) is triggered by identifying an operation risk associated with operating a chemical plant different from the one or more target chemical plant(s) and / or by changing a configuration of the one or more target chemical plant(s).

7. The method of any one of claims 1 to 6, wherein providing the at least one identified operation risk comprises: providing target configuration data associated with target configuration of the one or more chemical plant(s), identifying a subset of target configuration data associated with the at least one identified operation risk by providing configuration data task instructions related to the target configuration data and the at least one identified operation risk to the one or more data-driven model(s), wherein the one or more data-driven model(s) are further configured to select the subset of the target configuration data according to identified operation risks, providing the subset of target configuration data, and optionally the at least one identified operation risk.

8. The method of claim 7, further comprising providing current configuration data associated with current operating instructions for operating the one or more target chemical plant(s), and matching the target configuration data with the current configuration data, by providing a matching task instruction related to the target configuration data and the current configuration data to the one or more data-driven model(s), wherein the one or more data-driven model(s) are further configured to provide a subset of target configuration data upon determining a deviation between the target configuration data and the current configuration data and otherwise provide an indication on a verification of the current configuration data.

9. The method of claim 8, wherein the one or more data-driven model(s) are further provided with an output data structure associated with the indication on the verification of the current configuration data and / or thesubset of target configuration data , and wherein the data-driven model is further configured to provide a structured indication on the verification of the current configuration data and / or a structured subset of target configuration data according to the output data structure.

10. The method of any one of claims 1 to 9, wherein the request for identifying and / or mitigating, in particular eliminating, operation risks associated with operating one or more target chemical plant(s) is related to risk configuration data associated with an operation risk associated with the configuration of the one or more chemical plant(s), in particular deviating from the one or more target chemical plant(s).1 1 . The method of any one of claims 1 to 10, wherein the one or more data-driven model(s) are pretrained data-driven model(s), wherein the pretrained data-driven model(s) is configured to perform a plurality of different tasks according to a plurality of different task instructions.

12. The method of any one of claims 1 to 10, wherein the one or more data-driven model(s) are finetuned data- driven model(s), wherein the finetuned data-driven model(s) are obtained by further training pretrained data-driven model(s), wherein the pretrained data-driven model(s) are configured to perform a plurality of different tasks according to a plurality of different task instructions.

13. An apparatus, in particular for monitoring and / or controlling one or more target chemical plant(s), the apparatus comprising: a processor configured to perform any one of the methods according to any one of claims 1 to 12.

14. Use of a data-driven model according to any one of claims 1 to 12 for mitigating, in particular eliminating, one or more operation risk(s) for operating one or more target chemical plant(s).

15. Use of an operation risk as identified according to any one of claims 1 to 12 for mitigating, in particular eliminating, the identified operation risk.

Images