Synthetic representations of production and / or processing facilities for monitoring and / or controlling the production and / or processing facilities

By employing synthetic data and advanced data-driven models, the monitoring and control of production and processing facilities are enhanced, addressing data limitations and improving operational efficiency and accuracy.

WO2026233906A1PCT designated stage Publication Date: 2026-11-12BASF SE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/EP2026/062756
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-05
Filing Date
2026-05-04
Publication Date
2026-11-12

AI Technical Summary

Technical Problem

Existing production and processing facilities face challenges in monitoring and controlling operations due to limited availability of high-quality training data, leading to inaccurate and inefficient monitoring and control systems.

Method used

Utilization of synthetic data and classification data-driven models, such as convolutional neural networks and generative adversarial networks, to enhance the accuracy and precision of monitoring and control systems by providing detailed representations of facility equipment and connections, using generative and classification models trained on both historical and synthetic data.

Benefits of technology

Improves the reliability and efficiency of monitoring and controlling production and processing facilities by enabling precise classification and adaptation on an equipment part-level, facilitating better root cause analysis and predictive maintenance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure EP2026062756_12112026_PF_FP_ABST
    Figure EP2026062756_12112026_PF_FP_ABST
Patent Text Reader

Abstract

A method for monitoring and / or controlling a production and / or processing facility, the method comprising: • obtaining, in particular receiving, a representation of the production and / or processing facility, • providing the representation to a classification data-driven model for obtaining an indication of the production and / or processing facility, wherein the classification data-driven model is obtained according to any one of claims 2 to 5, • providing the indication of the production and / or processing facility, in particular for monitoring and / or controlling a production and / or processing facility.
Need to check novelty before this filing date? Find Prior Art

Description

250120SYNTHETIC REPRESENTATIONS OF PRODUCTION AND / OR PROCESSING FACILITIES FOR MONITORING AND / OR CONTROLLING THE PRODUCTION AND / OR PROCESSING FACILITIESTECHNICAL FIELDThe disclosure relates to synthetic data and a method for monitoring and / or controlling a production and / or processing facility, a method for obtaining a classification data-driven model, a method for obtaining synthetic representations of a production and / or processing facility, method for training a generative data-driven model to provide synthetic representations of a production and / or processing facility, an apparatus and / or a system comprising at least one processor and at least one memory, use of a synthetic representation of a production and / or processing facility, use of synthetic image data.TECHNICAL BACKGROUNDWhen a production or processing facility malfunctions or undergoes a process change, it is crucial to have a comprehensive understanding of the facility's structure, as even minor adjustments in one area can significantly impact the functionality of another. Therefore, effectively monitoring and controlling the production and / or processing facility is desired.SUMMARYIn one aspect, this disclosure relates to a method for monitoring and / or controlling a production and / or processing facility, the method comprising:obtaining, in particular receiving, a representation of the production and / or processing facility, providing the representation to a classification data-driven model for obtaining an indication of the production and / or processing facility, wherein the classification data-driven model is obtained according to as described herein,providing the indication of the production and / or processing facility, in particular for monitoring and / or controlling a production and / or processing facility.In another aspect, it relates to a method for obtaining a classification data-driven model for monitoring and / or controlling a production and / or processing facility, the method comprising:obtaining, in particular receiving, synthetic representations of the production and / or processing facility as generated as described herein,obtaining, in particular receiving, indications of the production and / or processing facility associated with the synthetic representations,2501202training the classification data-driven model based on the obtained synthetic representations and the obtained indications to classify representations according to the associated production and / or processing facility,optionally providing the classification data-driven model, in particular for monitoring and / or controlling a production and / or processing facility.In another aspect, it relates to use of synthetic image data as described herein for training a classification data-driven model to provide an indication of a production and / or processing facility upon receiving representations of the production and / or processing facility for monitoring and / or controlling an operation of the production and / or processing facility.In another aspect, it relates to use of a synthetic representation as described herein for monitoring and / or controlling an operation of a production and / or processing facility.In another aspect, it relates to an apparatus and / or system comprising at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to carry out the steps of any one of the methods as described herein.In another aspect, it relates to a method for obtaining synthetic representations of a production and / or processing facility, in particular for monitoring and / or controlling the production and / or processing facility, the method comprising:obtaining, in particular receiving, a trigger for triggering a generative data-driven model to generate the synthetic representation of the production and / or processing facility, wherein the generative data-driven model is trained as described herein and / or according to historical representations of the production and / or processing facility to generate synthetic representations of the production and / or processing facility, providing the trigger to the generative data-driven model for generating the synthetic representation, providing the generated synthetic representation.In another aspect, it relates to a method for training a generative data-driven model to provide synthetic representations of a production and / or processing facility, in particular for monitoring and / or controlling the production and / or processing facility, the method comprising:obtaining, in particular receiving, historical representations of the production and / or processing facility, training a generative model according to the historical representations for generating synthetic representations, in particular synthetic representations corresponding to the historical representations,2501203optionally, providing the trained generative data-driven model, in particular for monitoring and / or controlling the production and / or processing facility.EMBODIMENTSAny disclosure, embodiments and examples described herein relate to the methods, the systems, apparatuses, 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 usedherein and / or the technical field of the present disclosure will be outlined by ways of embodiments and / or examples. It is to be understood that the present disclosure is not limited to said embodiments and / or examples. In case of any malfunction of a production and / or processing facility and / or change of a process operated by a production and / or processing facility, adapting the production and / or processing facility requires an in-depth overview of the structure of the production and / or processing facility. Even small adaptions in a part of the production and / or processing facility can highly affect functioning of another part of the production and / or processing facility. Hence, for monitoring and / or controlling operation of the production and / or processing facility 102, determining the equipment items present in the production and / or processing facility, their properties and their connections is of high importance.For this purpose, classification data-driven models 106 may be utilized. The classification data-driven model 106 may be configured, in particular trained, to provide an indication of a production and / or processing facility, in particular an indication of at least a part of the equipment of the production and / or processing facility, in response to receiving representation associated with the production and / or processing facility 104. The classification data-driven model 106 may be a data-driven model, preferably a classification data-driven model such as a convolutional neural network or a vision transformer. Additionally or alternatively, the classification data-driven model 106 may comprise a clustering algorithm and / or a k-nearest neighbors algorithm. Other algorithms and / or model architectures are known in the art and can be used accordingly. In order to obtain accurate and precise classification data-driven models 106, a high amount of qualitative training data, such as image data and / or graph data structures, may be required. Typically, availability of such training data, associated with the production and / or processing facility 104 is limited due to resources required to build production and / or processing facilities.This can be improved by utilizing synthetic training data as described herein for obtaining classification data-driven models. The so-obtained synthetic training data is highly qualitive and provides uniform use and representation of the production and / or processing facilities. This allows for comparability of different sets oftraining data. As a result, the classification data-driven model 106 trained based on this training data is more accurate and / or precise. The reason for this is that the classification data-driven model can be focused on the relevant parts for analyzing the images and is not distracted by changes in representation due to non-uniform training data. Consequently, the features obtained by the classification data-driven model are more relevant for a classification task. Because the output of the classification data-driven models can be used for monitoring and / or controlling production and / or processing facilities, more reliable and meaningful result from the classification data-driven model result in a more efficient, precise and / or accurate monitoring and / or controlling of the production and / or processing facility. Ultimately, this disclosure contributes to improving operation of production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.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.A processor may be a processor of any suitable type and is preferably a processor configured for parallel processing of at least a hundred or at least a thousand threads in parallel, e.g. a graphical processing unit (GPU). For instance, the processor comprises at least a hundred or a at least a thousand parallel processing cores. In particular, the processor may comprise at least one (preferably at least a thousand) compute unified device architecture (CUDA) core(s), which may allow for using a graphical processing unit as the processor, which may increase computational efficiency. For instance, the processor may comprise at least one (e.g. at least a hundred) streaming multiprocessor cores, which may allow for increasing the data throughput. As a further example, the processor may comprise one or more (e.g. at least a hundred) tensor core(s) and / or (e.g. at least a hundred) tensor processing units (TPUs). A tensor core may be specifically adapted to perform matrix operations and may allow to accelerate large matrix operations. A tensor core may be configured to perform mixed-precision matrix multiply and accumulate calculations in a single operation. For instance, a tensor core may perform mixed-precision floating-point matrix arithmetic, specifically utilizing FP16 (half-precision) inputs to produce either fullprecision (FP32) or half-precision (FP16) outputs. In the case of FP16 output, a tensor core may provide a performance boost by storing the intermediate accumulation results in FP32 format, thereby maintaining the precision necessary for accurate results. A tensor processing unit may be an application-specific integrated circuit (ASIC). It may comprise a matrix multiplication unit (MXU), which may be specifically adapted or configured for dense linear algebra operations. TPUs may be configured to handle large-scale matrix operations efficiently, which may provide high computational throughput for Al tasks. A TPU may be equipped with on-chip high-bandwidth memory (HBM), which may enhance the capability for the use of larger models and batch sizes. TPUs may be connected in groups called Pods, which may scale up workloads with minimal code changes. An MXU may be specifically configured for performing matrix multiplications. A TPU may comprise a tensor core.5For example, a processor may comprise several thousand tensor cores, each capable of performing 64 floating point FMA (Fused Multiply-Add) operations per clock cycle or (e.g. at least several hundred) tensor processing units (TPUs) being specifically configured for accelerating machine learning (ML) workloads, particularly for cloudbased applications. Additionally, Field-Programmable Gate Arrays (FPGAs) and Application-Specific Integrated Circuits (ASICs) may provide flexibility and performance benefits for specific Al tasks. With these capabilities, such a GPU may allow for hundreds of TFLOPs (Tera Floating-Point Operations per Second) of performance in mixed-precision computations. Furthermore, a tensor core may support a variety of numerical formats, including IEEE standard half-precision, single-precision, and double-precision floating-point formats, as well as a range of integer formats.A processor may be a central processing units (CPU) configured with an advanced architecture, such as Intel’s Xeon Scalable processors or AMD’s EPYC series. A CPU may be configured for sequential processing and general-purpose computing. These CPUs may incorporate vector instruction sets, such as AVX-512, to accelerate mathematical computations that may e.g. enhance Al model training and inference. Furthermore, CPUs may integrate Al accelerators i.e. a CPU may be specifically configured for deep learning workloads.The processor may be coupled to memory having a memory bandwidth of at least a hundred gigabytes per second, which may allow efficient handling of extensive data sets and may allow faster reading, processing, and writing compared to a general-purpose processor such as a computational processing unit.The memory may be a high-capacity memory configured to manage the data-intensive nature of Al applications, providing necessary bandwidth and storage capacity for complex datasets. The memory may for instance be DDR4, DDR5, High Bandwidth Memory (HBM) and / or GDDR6X memory, which may improve data transfer rates and reduce latency. Such memory may enhance e.g. modeling and real-time sensor data for monitoring and control. Further, the memory may be operated with memory optimization techniques, such as caching and prefetching, which may enhance the execution speed of Al algorithms. Non-volatile Memory (NVM) technologies, including NAND Flash and 3D XPoint, may provide persistent storage solutions with high-speed access, which may enhance rapid data storage and retrieval for Al applications.In an example embodiment, the graph data structure and / or the image data may be associated with at least a part of equipment of a production and / or processing facility. The graph data structure may be a representation of the production and / or processing facility. The graph data structure may represent at least a part of equipment associated with the production and / or processing facility and / or a connection between two or more equipment items. The part of the equipment may comprise the two or more equipment items. The image data may comprise an image of at least a part of equipment of a production and / or processing facility and / or a representation, in particular a digital representation, of at least a part of equipment of a production and / or processing facility. For6example, the representation of at least a part of equipment of a production and / or processing facility may comprise a piping and instrumentation diagram.In an example embodiment, the image data may be associated with at least a part of equipment of a production and / or processing facility. The image data may comprise an image of at least a part of equipment of a production and / or processing facility and / or a visual representation, of at least a part of equipment of a production and / or processing facility. For example, the representation of at least a part of equipment of a production and / or processing facility may comprise a piping and instrumentation diagram.In an example embodiment, the representation of the production and / or processing facility may be suitable for being processed by a processor and / or a classification data-driven model. The representation of the production and / or processing facility may include image data associated with the production and / or processing facility and / or a graph data structure associated with the production and / or processing facility. The representation of the production and / or processing facility may be associated with at least a part of equipment of a production and / or processing facility. In particular, the representation of the production and / or processing facility may be a representation of at least the part of the equipment and / or a connection between two or more equipment items of the production and / or processing facility. Thereby, an adaption on an equipment part-level is enabled. Ultimately, this contributes to improving operation of production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.In an example embodiment, the one or more instruction(s) for generating the graph data structure, in particular the synthetic graph data structure and / or the historical graph data structure may comprise text and / or string data. The one or more instruction(s) for generating the graph data structure may specify a property of the graph data structure and / or at least the part of the equipment associated with the production and / or processing facility. The property of the graph data structure may comprise a property of at least the part of the equipment associated with the production and / or processing facility, a property of an equipment item of the production and / or processing facility and / or a property of a connection between two or more equipment items. The property of the graph data structure may specify a property of at least the part of the equipment associated with the production and / or processing facility, a property of an equipment item of the production and / or processing facility and / or a property of a connection between two or more equipment items.In an example embodiment, the indication of the production and / or processing facility may comprise an indication of at least a part of equipment associated with the production and / or processing facility. The indication of the production and / or processing facility may comprise and / or may be indicative of a property of the production and / or processing facility, in particular at least a part of the equipment associated with the production and / or processing2501207facility. A property of the production and / or processing facility, in particular at least the part of the equipment, may characterize the production and / or processing facility, in particular the equipment. Preferably, property of the production and / or processing facility, in particular at least the part of the equipment, may characterize an appearance of the production and / or processing facility, in particular at least the part of the equipment and / or a function of the production and / or processing facility, in particular at least the part of the equipment. The indication of the production and / or processing facility may comprise an indication of at least a part of the equipment and / or a connection between two or more equipment items associated with the production and / or processing facility.In an example embodiment, the classification data-driven model may be trained to classify representations of the production and / or processing facility according to at least a part of the equipment associated with the production and / or processing facility. The classification data-driven model may be trained based on the historical and / or synthetic representations to provide an indication of a production and / or processing facility upon receiving the historical and / or synthetic representations. The classification data-driven model may be configured to classify the historical and / or synthetic representations, in particular according to at least a part of the production and / or processing facility, preferably according to at least a part of the equipment associated with the production and / or processing facility. The indication of the production and / or processing facility provided by the classification data-driven model may comprise a numerical value associated with a property of the production and / or processing facility, in particular at least a part of the equipment of the production and / or processing facility. The classification data-driven model may be configured, in particular trained, to provide an indication of a production and / or processing facility, in particular an indication of at least a part of the equipment of the production and / or processing facility, in response to receiving the historical and / or synthetic representations of the production and / or processing facility.In an example embodiment, obtaining, in particular receiving, the historical graph data structure may include obtaining, in particular receiving, via a user interface.In an example embodiment, providing the indication of the production and / or processing facility may comprise displaying the indication of the production and / or processing facility to a worker associated with the production and / or processing facility.In an example embodiment, the classification data-driven model may be trained to classify the synthetic representations according to at least a part of the equipment and / or a connection between two or more equipment items associated with the production and / or processing facility. Thereby, an adaption on an equipment part-level is enabled. Ultimately, this contributes to improving operation of production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.2501208In an example embodiment, the obtained synthetic representation of the production and / or processing facility may comprise synthetic image data associated with the production and / or processing facility. The synthetic image data may be obtained from a synthetic graph data structure associated with the production and / or processing facility. The synthetic graph data structure may be generated by:• obtaining, in particular receiving, a trigger for triggering a generative data-driven model to generate the synthetic graph data structure of a production and / or processing facility, wherein the generative data-driven model is trained according to historical graph data structures of the production and / or processing facility to generate synthetic graph data structures,• providing the trigger to the generative data-driven model and / or as described herein.In an example embodiment, training the generative data-driven model may include training a discriminator to classify whether the representations generated by the generative data-driven model corresponds to historical representations or synthetic representations.In an example embodiment, training the generative data-driven model may include mapping the historical representations to a numerical representation of the historical representations, in particular by an encoder. The generative data-driven model may be trained to map the digital representations to the synthetic representations.In an example embodiment, the generative data-driven model may be further trained according to instructions for generating synthetic representations. In particular, the synthetic representations may be generated to correspond to the historical representations. The generative data-driven model may be trained according to the historical representations and the instructions for generating synthetic representations, in particular synthetic representations corresponding to the historical representations. The instructions may comprise text and / or string data and / or may be obtained, in particular received, via a user interface. By doing so, the generation of the synthetic data can be tailored via the instructions. Hence, the generated synthetic data is more flexible and directed. Additionally, the human-machine interaction for generating synthetic data is enabled. Ultimately, this contributes to improving operation of production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.In an example embodiment, the historical representations may be obtained, in particular received, via a user interface and / or a database comprising a plurality of representations of one or more production and / or processing facilities. By doing so, the generation of the synthetic data can be tailored via the historical representations provided. Hence, the generated synthetic data is more flexible and directed. Additionally, the human-machine interaction for generating synthetic data is enabled. Ultimately, this contributes to improving operation of2501209production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.In an example embodiment, any one of the methods may further comprise obtaining, in particular receiving, a selection of the historical representations from a plurality of representations comprised by a database via the user interface, and wherein the historical representations are obtained based on the selection. By doing so, the user can define the historical graph data structures to be used for obtaining synthetic graph data structures, i.e. the user is enabled to tailor the generation of synthetic data to the production and / or processing facility at hand. Ultimately, this contributes to improving operation of production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.In an example embodiment, training the generative data-driven model may comprise mapping the obtained instructions to a digital representation of the instructions. The generative data-driven model may be further trained based on the digital representation of the instructions. Preferably, training the generative data-driven model may include training a decoder to generate synthetic instructions, in particular corresponding to the obtained instructions, from a digital representation of the obtained instructions. More preferably, the obtained instructions and the historical representations may be mapped to a combined representation of the obtained instructions and the historical representations. The generative data-driven model may be trained based on the combined representation. The combined representation may be obtained by combining the digital representation of the instructions and the digital representation of the historical representations. Further, training the generative data-driven model may include training an encoder to map the obtained instructions to a digital representation of the instructions.In an example embodiment, the digital representation may comprise a numerical representation such as a tensor.In an example embodiment, the synthetic data comprises data generated by a data-driven model. The historical data comprises data obtained based on real data, i.e. data generated with sensors.In an embodiment, the trigger may comprise a tensor suitable for being received by the generative data-driven model. The trigger may be associated with a random distribution. For example, the trigger may comprise a random tensor.BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS25012010In 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.FIG. 1 illustrates an embodiment of a production and / or processing facility 102.FIG. 2 illustrates an embodiment of obtaining a classification data-driven model.FIG. 3 illustrates an embodiment of obtaining a classification data-driven model.FIG. 4 illustrates an embodiment of obtaining image data from a graph data structure.FIG. 5 illustrates an embodiment of obtaining a generative data-driven model 202 such as a variational autoencoder (VAE).FIG. 6 illustrates an embodiment of obtaining a generative data-driven model 202 such as a variational autoencoder (VAE).FIG. 7 illustrates an embodiment of obtaining a generative data-driven model 202 such as a generative adversarial network (GAN).FIG. 8 illustrates an embodiment of a method for monitoring and / or controlling a production and / or processing facility.DETAILED DESCRIPTIONThe following embodiments are mere examples for implementing the method, the system or application device disclosed herein and shall not be considered limiting.FIG. 1 illustrates an embodiment of a production and / or processing facility 102.The production and / or processing facility 102 may be a chemical plant for producing one or more chemical product(s) and / or may be a processing facility for processing one or more chemical product(s). The production and / or processing facility 102 may be associated with, in particular may comprise, equipment, i.e. one or more equipment item(s). The production and / or processing facility 102 may be monitored and / or controlled to ensure efficient and safe production and / or processing of products. Operation of the production and / or processing facility 102 may depend on the equipment associated with the production and / or processing facility 102. For example, a mixer may require one or more cleaning step(s) between two batches of producing and / or processing products. Further, a reactor may require a predefined type of valve, connection and / or tank for supplying one or more input25012011product(s) to the reactor. Hence, the operation of the production and / or processing facility 102 may be monitored and / or adapted based on at least parts of the equipment present in a production and / or processing facility.For monitoring and / or controlling operation, in particular equipment, of the production and / or processing facility 102, classification data-driven models 106 may be utilized. The classification data-driven model 106 may be configured, in particular trained, to provide an indication of a production and / or processing facility, in particular an indication of at least a part of the equipment of the production and / or processing facility, in response to receiving representation associated with the production and / or processing facility 104. The classification data-driven model 106 may be a data-driven model, preferably a classification data-driven model such as a convolutional neural network or a vision transformer. Additionally or alternatively, the classification data-driven model 106 may comprise a clustering algorithm and / or a k-nearest neighbors algorithm. Other algorithms and / or model architectures are known in the art and can be used accordingly. In order to obtain accurate and precise classification data-driven models 106, a high amount of qualitative training data may be required. Typically, availability of training data associated with the production and / or processing facility 104 is limited. Consequently, classification of production and / or processing facilities 102 needs to be improved.This can be improved by utilizing synthetic training data as described herein for obtaining classification data-driven models. The so-obtained synthetic training data is highly qualitive and provides uniform use and representation of the production and / or processing facilities. This allows for comparability of different sets of training data. As a result, the classification data-driven model 106 trained based on this training data is more accurate and / or precise. The reason for this is that the classification data-driven model can be focused on the relevant parts for analyzing the images and is not distracted by changes in representation due to non-uniform training data. Consequently, the features obtained by the classification data-driven model are more relevant for a classification task. Because the output of the classification data-driven models can be used for monitoring and / or controlling production and / or processing facilities, more reliable and meaningful result from the classification data-driven model result in a more efficient, precise and / or accurate monitoring and / or controlling of the production and / or processing facility. This allows for better monitoring and / or controlling when utilized in root cause analysis scenarios for applications like Predictive Maintenance, Failure Detection, etc. For example, finding the possible root causes of a “low alarm” of a reactor can be located by utilizing the P&ID image detection of elements of a tank connected to a preceding pump that failed or a outlet valve being opened. The classification of the P&ID can allow to identify deviations from a target state if visible from the outside. In this case, a deviation and potentially a degree of the deviation, e.g. severe, moderate, slightly, can be displayed to the operator of the chemical plant. The operator will react on the displayed information and trigger certain in-depth analysis of the corresponding equipment, e.g. via sensors. Additionally or alternatively, said in-depth analysis may be triggered automatically. In the case of a severe deviation, the operator may stop at least a part of the production and / or processing associated with the corresponding equipment. Alternatively, the corresponding equipment may be excluded fromthe production and / or processing and / or may be exchanged by another part. Further variants of monitoring and / or controlling the production and / or processing based on said classification are possible and included herein.Ultimately, this disclosure contributes to improving operation of production and / or processing facility due to the improved monitoring and / or controlling of the production and / or processing facility.FIG. 2 illustrates an embodiment of obtaining a classification data-driven model.The classification data-driven model may be trained for monitoring and / or controlling operation of the production and / or processing facility as described in the context of FIG. 1. The classification data-driven model 106 may be trained based on synthetic graph data structures and / or historical graph data structures to provide an indication of a production and / or processing facility upon receiving the synthetic graph data structure and / or the historical graph data structure. The classification data-driven model may be configured to classify the graph data structures, in particular according to at least a part of the production and / or processing facility, preferably according to at least a part of the equipment associated with the production and / or processing facility. The indication of the production and / or processing facility provided by the classification data-driven model 106 may comprise a numerical value associated with a property of the production and / or processing facility, in particular at least a part of the equipment of the production and / or processing facility. For example, the classification data-driven model may be configured to provide a numerical value indicative of a presence of a tank and / or a quantity associated with at least a part of the equipment of the production and / or processing facility. Additionally or alternatively, the classification data-driven model may be configured to provide a tensor such as a vector. The vector may be indicative of a presence of a plurality of equipment items and / or a type of the equipment item.During training, the classification data-driven model 106 may be provided with synthetic graph data structure and / or historical graph data structure and provide a suggested indication of the production and / or processing facility associated with the image data. This suggested indication of the production and / or processing facility may be compared with a ground truth, i.e. the label, associated with the image data provided to the classification data-driven model 106. Depending on a deviation of the suggested indication of the production and / or processing facility from the ground truth, the deviation may be propagated through the classification data-driven model via a backpropagation algorithm. By using a gradient descent or ascent algorithm, one or more model parameter(s) associated with the classification data-driven model 106 may be adapted to decrease the deviation of the suggested indication from the ground truth.The training of the classification data-driven model 106 may be terminated if the deviation between the suggested indications and the ground truth may be within a predefined deployment range. After training, the classification data-driven model 106 may be deployed for providing and / or generating indications of the production and / or processing facilities. For this purpose, the classification data-driven model 106 may be provided by13representations of the production and / or processing facility, for example data generated via a sensor, or synthetic data, for example obtained as described in the context of FIG. 5 - FIG. 8.Examples of the classification data-driven model include but are not limited to a convolutional neural network, a graph neural network or a vision transformer. Additionally or alternatively, the classification data-driven model 106 may comprise a clustering algorithm and / or a k-nearest neighbors algorithm. Other algorithms and / or model architectures are known in the art and can be used accordingly.FIG. 3 illustrates an embodiment of obtaining a classification data-driven model.The classification data-driven model may be trained to be applied for monitoring and / or controlling operation of the production and / or processing facility as described in the context of FIG. 1. The classification data-driven model 106 may be trained based on synthetic image data and / or historical image data to provide an indication of a production and / or processing facility upon receiving the synthetic image data and / or the historical image data. The image data may be obtained based on graph data structures associated with the production and / or processing facility as described in further detail in the context of FIG. 4.The classification data-driven model may be configured to classify the synthetic image data and / or the historical image data, in particular according to at least a part of the production and / or processing facility, preferably according to at least a part of the equipment associated with the production and / or processing facility. The indication of the production and / or processing facility provided by the classification data-driven model 106 may comprise a numerical value associated with a property of the production and / or processing facility, in particular at least a part of the equipment of the production and / or processing facility. For example, the classification data-driven model may be configured to provide a numerical value indicative of a presence of a tank and / or a quantity associated with at least a part of the equipment of the production and / or processing facility. Additionally or alternatively, the classification data-driven model may be configured to provide a tensor such as a vector. The vector may be indicative of a presence of a plurality of equipment items and / or a type of the equipment item.During training, the classification data-driven model 106 may be provided with synthetic image data and / or historical image data and provide a suggested indication of the production and / or processing facility associated with the image data. This suggested indication of the production and / or processing facility may be compared with a ground truth, i.e. the label, associated with the image data provided to the classification data-driven model 106. Depending on a deviation of the suggested indication of the production and / or processing facility from the ground truth, the deviation may be propagated through the classification data-driven model via a backpropagation algorithm. By using a gradient descent or ascent algorithm, one or more model parameter(s) associated with the classification data-driven model 106 may be adapted to decrease the deviation of the suggested indication from the ground truth.14The training of the classification data-driven model 106 may be terminated if the deviation between the suggested indications and the ground truth may be within a predefined deployment range. After training, the classification data-driven model 106 may be deployed for providing and / or generating indications of the production and / or processing facilities. For this purpose, the classification data-driven model 106 may be provided by image data generated via a sensor or synthetic data, for example obtained as described in the context of FIG. 10.FIG. 4 illustrates an embodiment of obtaining image data from a graph data structure.The image data and the graph data structure may represent the production and / or processing facility. A node of the graph data structure may represent an entity, i.e. an equipment item, of the production and / or processing facility. An edge of the graph data structure may represent a connection between two or more entities, i.e. two or more equipment items. Rules may be predefined for converting the graph data structure to the image data and vice versa. The predefined rules may specify a relation between the one or more node(s) and / or the one or more edge(s) and a corresponding visual representation via the image data, i.e. a corresponding part of the image data associated with the production and / or processing facility.FIG. 5 illustrates an embodiment of obtaining a generative data-driven model 202 such as a variational autoencoder (VAE).Training the generative data-driven model 202 may result in the generative data-driven model 202 configured to provide synthetic graph data structures, in particular in response to receiving a vector, in particular a latent space vector, associated with a random distribution of features associated with the production and / or processing facility. The vector may be suitable for being received by the generative data-driven model 202. The features associated with the production and / or processing facility may be obtained by providing the graph data structure associated with the production and / or processing facility to an encoder 514. T raining the VAE may result in obtaining an encoder configured to map graph data structures provided to a numerical representation of the graph data structure such as a latent space vector. The latent space may be a multi-dimensional space. The vector may indicate, in particular comprise one or more features associated with the graph data structure.Training the VAE may include mapping the historical graph data structures to vectors, in particular latent space vectors associated with the historical graph data structure via the encoder and mapping the vectors to the synthetic graph data structures. Further, the synthetic graph data structures may be compared to the historical graph data structures to obtain a deviation between the synthetic graph data structures generated by the generative data-driven model 202 and the historical graph data structures provided to the encoder 514. This deviation may be propagated through the VAE including the generative data-driven model 202 via a backpropagation algorithm. By using a gradient descent or ascent algorithm, one or more model parameter(s) associated with the VAE including the generative data-driven model 202 may be adapted to decrease the25012015deviation of the suggested indication from the ground truth. Similarly, the encoder can be adapted together with the generative data-driven model 202. In the context of VAE, the generative data-driven model 202 may also be referred to as decoder. This generative data-driven model 202 may be deployed for generating synthetic graph data structures for obtaining the classification data-driven model as described in further detail in the context of FIG. 1 to FIG. 3 and FIG. 8.FIG. 6 illustrates an embodiment of obtaining a generative data-driven model 202 such as a variational autoencoder (VAE).The VAE in FIG. 6 may comprise the VAE in FIG. 5 and a further encoder, i.e. a text encoder and a further decoder, i.e. a text decoder. During training, the text encoder may be configured to map historical instructions for generating the synthetic graph data structure 622 to a vector, i.e. a latent space vector associated with the instructions 624. This vector associated with the historical instructions may be combined with the vector associated with the historical graph data structures into a combined representation associated with the historical graph data structures and the historical instructions. The combined representation may be a multi-modal representation of the production and / or processing facility. Combining the two vectors may include concatenating the vectors, element-wise multiplication applied to the two vectors, cross-modal attention applied to the two vectors, or multimodal hashing applied to the two vectors. The combined representation may be a representation of the graph data structures and the instructions in a joint latent space.A Variational Autoencoder (VAE) may be a generative model that may combine principles from probabilistic graphical models and neural networks. It may be designed to learn a latent representation of input data in a way that allows it to generate new data points that are similar to the training data. The training objective for the VAE may be to maximize the Evidence Lower Bound (ELBO), which can be expressed in terms of the data likelihood and a regularization term. By maximizing the ELBO, the VAE balances the fidelity of reconstruction with the need for a well-structured latent space. To optimize the ELBO with respect to the parameters of the encoder and decoder, we need to compute gradients. This can be done by using reparameterization allowing gradients to flow through the encoder during backpropagation.During training the generative data-driven model 202, i.e. the decoder associated with the graph data structures, may map the combined representation to the synthetic graph data structures. The decoder associated with the instructions may map the combined representation to the synthetic instructions. The so-obtained synthetic instructions and graph data structures may be compared to the historical instruction and graph data structures for adapting one or more model parameter(s) of the VAE including the generative data-driven model 202. Adapting the one or more model parameter(s) may include using the backpropagation algorithm and the gradient descent or ascent algorithm as described in the context of FIG. 5.25012016FIG. 7 illustrates an embodiment of obtaining a generative data-driven model 202 such as a generative adversarial network (GAN).In an example embodiment, the generative data-driven model 202 may be at least a part of a GAN. During training, the generative data-driven model 202 may be provided with a random vector, i.e. a latent space vector associated with a random distribution. The random vector may be suitable to be received by the generative data-driven model 202. The generative data-driven model 202 may generate synthetic graph data structures 712 based on the provided vectors. A discriminative data-driven model 510 may be provided with the generated synthetic graph data structures 712 and historical graph data structures, i.e. graph data structures of real production and / or processing facilities. The discriminative data-driven model 510 may classify which provided data set may be real data and which may be synthetic data. Hence, during training, the generative data-driven model 202 may be adapted to generate realistic synthetic data while the discriminative data-driven model 510 may be adapted to distinguish between real and synthetic data. Adapting the discriminative data-driven model 510 and the generative data-driven model 202 may include determining a discriminator loss and a generator loss. Said losses can be backpropagated through the corresponding networks to adapt the model parameters of the networks.The training process of GANs can be formalized using game theory, where we may define a minimax game between the Generator and the Discriminator. The goal may be to find a Nash Equilibrium where neither player can improve their strategy given the other player's strategy. Let ( z ) be a random noise vector sampled from a latent space ( Z ). The generator ( G ) may take this noise vector as input to produce fake data ( G(z) ). The discriminator ( D ) may take either real data ( x ) from the training set or fake data ( G(z) ) and may output a probability ( D(x) ) that may indicate whether the input may be real or fake. The objective functions for the Discriminator and Generator can be defined as a combination of the discriminator objective defined via the discriminator loss and the generator objective defined via the generator loss. The Discriminator may aim to maximize its ability to differentiate between real and fake data. The Generator may aim to minimize the probability that the Discriminator correctly identifies the generated data as fake. This may mean the Generator may be trying to maximize ( D(G(z)) ), which may equivalent to minimizing the negative log probability.The training process of the GAN may include the following steps:1. Initialization: Start with random weights for both the Generator and Discriminator.2. Training Loop:o Sample a batch of real data ( x ) from the dataset.o Sample a batch of noise vectors ( z ) from the latent space.o Compute the Discriminator loss and update the Discriminator parameters using backpropagation. o Compute the Generator loss and update the Generator parameters using backpropagation.3. Repeat until convergence or a predefined number of epochs.Optionally, the generative data-driven model 202 may be instructed how to generate the synthetic graph data structures via instructions for generating the synthetic graph data structure. For this purpose, the generative data-driven model 202 receives a digital, i.e. numerical representation, associated with the instructions. This digital representation may be obtained by providing the instructions to an embedding model 714. The embedding model 714 may be configured and / or trained to map data to a digital representation of the data, i.e. map the instructions to a digital representation of the instructions. In particular, the embedding model 714 may be configured to adapt the data format associated with the instructions and / or extract one or more latent space features from the instructions.In an example, the generative data-driven model 202 may be a conditional GAN.FIG. 8 illustrates an embodiment of a method for monitoring and / or controlling a production and / or processing facility.Historical graph data structures associated with a production and / or processing facility may be obtained, in particular received 802. The historical graph data structures may be obtained, in particular received, via a user interface and / or a database comprising a plurality of graph data structures. In an example embodiment, a selection of the historical graph data structures from the plurality of graph data structures comprised by the database may be obtained, in particular received via the user interface. Based on the selection, the historical graph data structures may be obtained, in particular received from the database. By doing so, the user can define the historical graph data structures to be used for obtaining synthetic graph data structures, i.e. the user is enabled to tailor the generation of synthetic data to the production and / or processing facility at hand.A plurality of instructions, in particular historical instructions, for generating synthetic graph data structures associated with the production and / or processing facility may be obtained, in particular received 804. Preferably the instructions may be suitable for generating the synthetic graph data structure corresponding to the historical graph data structures. Hence, the instructions may be historical instructions for triggering generating of synthetic graph data structures corresponding to the historical graph data structures. The instructions may be received via a user interface. The instructions may comprise text and / or string data. The instructions may describe the synthetic graph data structure to be generated.18A generative model may be trained according to the historical graph data structure and the instructions for generating synthetic graph data structures 806. In particular, the synthetic graph data structures may be generated to correspond to the historical graph data structures. The generative data-driven model 202 may be trained according to the historical graph data structure and the instructions for generating synthetic graph data structures, in particular synthetic graph data structures corresponding to the historical graph data structures. In an example embodiment, training the generative model may be as described in the context of FIG. 5 to FIG. 7 and / or the generative data-driven model 202 may be associated with an architecture as described in the context of FIG. 5 to FIG. 7.After training the generative data-driven model 202, the generative data-driven model 202 can be deployed for generating synthetic graph data structures 808. Said generated synthetic graph data structures may follow a distribution of data points associated with the historical graph data structures. Hence, training the generative data-driven model 202 may enable to the generative data-driven model 202 to generate realistic synthetic graph data structures. Generating the synthetic graph data structures may include providing a random vector to the generative data-driven model 202 for triggering the generative data-driven model 202 to generate the synthetic graph data structure and optionally providing instructions for generating the synthetic graph data structure to the generative data-driven model 202. The instructions may guide the generative data-driven model 202 to generate the synthetic graph data structure according to the instructions. Providing the instructions to the generative data-driven model 202 may include providing a digital representation of the instructions to the generative data-driven model 202. The digital representation of the instructions may be obtained by providing the instructions to an embedding model. The embedding model may be configured to map data to a digital representation of the data. The digital representation of the data may be associated with a different format and / or size than the data.Indications of the production and / or processing facility associated with the synthetic graph data structures may be obtained, in particular received, e.g. via a user interface and / or a database comprising a plurality of indications of a plurality of production and / or processing facilities 810. The indications may be labels associated with the synthetic graph data structures. The indications may characterize one or more equipment item(s) of the production and / or processing facility and / or a connection between two or more equipment items of the production and / or processing facility. An example of said indications can be seen in FIG. 1 and FIG. 2.In an example embodiment, a classification data-driven model may be trained based on the synthetic graph data structures and the associated indications, i.e. labels for classifying graph data structures according to the production and / or processing facility 812. The training may be further described in the context of FIG. 2.After training, the classification data-driven model may be deployed for classifying graph data structures associated with the production and / or processing facility 814. See FIG. 1 for an example.19In an example embodiment, synthetic image data may be obtained from the synthetic graph data structures 816. The synthetic image data may be associated with the indications. In particular, the synthetic image data may be associated with the indications associated with the corresponding synthetic graph data structures. Obtaining the synthetic image data may be described in further detail in the context of FIG. 4.The classification data-driven model may be trained based on the obtained synthetic image data and the corresponding indications 818. This may be described in further detail in the context of FIG. 3.After training, the classification data-driven model may be deployed for classifying image data associated with the production and / or processing facility 814. See FIG. 1 for an example.The publication Prior Art Disclosure; Issue 684; paragraphs

[1000] to

[8005] ; ISSN: 2198-4786; published: February 12, 2024 will be regarded as Reference RF1, which is incorporated herein by reference in its entirety. Preferably, the product is a product as described in Reference RF1; paragraphs

[1000] to

[8005] , Preferably, the method / process described herein is further a method / process for the production of a product.The converting step to obtain the product preferably comprises one or more step(s) as described below and can be performed by conventional methods well known to a person skilled in the art. The converting step preferably comprises one or more step(s) selected from:• recycling, preferably depolymerizing, gasifying, pyrolyzing, and / or steam cracking; and / or• purifying, preferably crystallizing, (solvent) extracting, distilling, evaporating, hydrotreating, absorbing, adsorbing and / or subjecting to ion exchanger; and / or• assembling, preferably foaming, synthesizing, chemical conversion, chemically transforming, polymerizing and / or compounding; and / or• forming, preferably foaming, extruding and / or molding; and / or• finishing, preferably coating and / or smoothing.In addition, the one or more step(s) are described in detail in Reference RF1; paragraphs

[1000] to

[8005] ,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 one20node 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.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.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.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,21etc., finite state machines, memory such as static random-access memory or embedded dynamic random-access memory, custom designed circuitry, programmable logic arrays, etc.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.All terms and definitions used herein are understood broadly and have their general meaning.The publication Prior Art Disclosure; Issue 684; paragraphs

[1000] to

[8005] ; ISSN: 2198-4786; published: February 12, 2024 will be regarded as Reference RF1, which is incorporated herein by reference in its entirety. Preferably, the product is a product as described in Reference RF1; paragraphs

[1000] to

[8005] , Preferably, the method / process described herein is further a method / process for the production of a product.The converting step to obtain the product preferably comprises one or more step(s) as described below and can be performed by conventional methods well known to a person skilled in the art. The converting step preferably comprises one or more step(s) selected from:• recycling, preferably depolymerizing, gasifying, pyrolyzing, and / or steam cracking; and / or• purifying, preferably crystallizing, (solvent) extracting, distilling, evaporating, hydrotreating, absorbing, adsorbing and / or subjecting to ion exchanger; and / or• assembling, preferably foaming, synthesizing, chemical conversion, chemically transforming, polymerizing and / or compounding; and / or• forming, preferably foaming, extruding and / or molding; and / or• finishing, preferably coating and / or smoothing.In addition, the one or more step(s) are described in detail in Reference RF1; paragraphs

[1000] to

[8005] ,

Claims

25012022CLAIMSWhat is claimed is:

1. A method for monitoring and / or controlling a production and / or processing facility, the method comprising:obtaining, in particular receiving, a representation of the production and / or processing facility, providing the representation to a classification data-driven model for obtaining an indication of the production and / or processing facility, wherein the classification data-driven model is obtained according to any one of claims 2 to 5,providing the indication of the production and / or processing facility, in particular for monitoring and / or controlling a production and / or processing facility.

2. A method for obtaining a classification data-driven model for monitoring and / or controlling a production and / or processing facility, the method comprising:obtaining, in particular receiving, synthetic representations of the production and / or processing facility as generated according to any one of claims 6 to 13,obtaining, in particular receiving, indications of the production and / or processing facility associated with the synthetic representations,training the classification data-driven model based on the obtained synthetic representations and the obtained indications to classify representations according to the associated production and / or processing facility,optionally providing the classification data-driven model, in particular for monitoring and / or controlling a production and / or processing facility.

3. The method of claim 1 or 2, wherein the classification data-driven model is trained to classify the synthetic representations according to at least a part of the equipment and / or a connection between two or more equipment items associated with the production and / or processing facility.

4. The method of any one of claims 1 to 3, wherein the indication of the production and / or processing facility comprises an indication of at least a part of the equipment and / or a connection between two or more equipment items associated with the production and / or processing facility.

5. The method of any one of claims 2 to 4, wherein the obtained synthetic representation of the production and / or processing facility comprises synthetic image data associated with the production and / or processing facility, and wherein the synthetic image data is obtained from a synthetic graph data structure associated with the production and / or processing facility, and wherein the synthetic graph data structure is generated by:25012023obtaining, in particular receiving, a trigger for triggering a generative data-driven model to generate the synthetic graph data structure of a production and / or processing facility, wherein the generative data-driven model is trained according to historical graph data structures of the production and / or processing facility to generate synthetic graph data structures,providing the trigger to the generative data-driven model.

6. A method for obtaining synthetic representations of a production and / or processing facility, in particular for monitoring and / or controlling the production and / or processing facility, the method comprising:obtaining, in particular receiving, a trigger for triggering a generative data-driven model to generate the synthetic representation of the production and / or processing facility, wherein the generative data- driven model is trained according to any one of claims 7 to 12 and / or according to historical representations of the production and / or processing facility to generate synthetic representations of the production and / or processing facility,providing the trigger to the generative data-driven model for generating the synthetic representation, providing the generated synthetic representation.

7. A method for training a generative data-driven model to provide synthetic representations of a production and / or processing facility, in particular for monitoring and / or controlling the production and / or processing facility, the method comprising:obtaining, in particular receiving, historical representations of the production and / or processing facility,training a generative model according to the historical representations for generating synthetic representations, in particular synthetic representations corresponding to the historical representations,optionally, providing the trained generative data-driven model, in particular for monitoring and / or controlling the production and / or processing facility.

8. The method of any one of claims 1 to 7, wherein the representation of the production and / or processing facility comprises a graph data structure associated with the production and / or processing facility.

9. The method of any one of claims 6 to 8, wherein the generative data-driven model is further trained according to instructions for generating synthetic representations.

10. The method of claim 9, wherein the instructions comprise text and / or string data and / or are obtained, in particular received, via a user interface.2501202411. The method of any one of claims 6 to 10, wherein the historical representations may be obtained, in particular received, via a user interface and / or a database comprising a plurality of representations of one or more production and / or processing facilities.

12. The method of any one of claims 6 to 11 , further comprising obtaining, in particular receiving, a selection of the historical representations from a plurality of representations comprised by a database via the user interface, and wherein the historical representations are obtained based on the selection.

13. An apparatus and / or system comprising at least one processor and at least one memory storing instructions that, when executed by the at least one processor, cause the apparatus at least to carry out the steps of the method according to any one of claims 1 to 12.

14. Use of a synthetic representation of a production and / or processing facility as obtained according to any one of claims 6 to 12 for monitoring and / or controlling an operation of a production and / or processing facility.

15. Use of synthetic image data as obtained according to any one of claims 6 to 12 for training a classification data-driven model to provide an indication of a production and / or processing facility upon receiving representations of the production and / or processing facility for monitoring and / or controlling an operation of the production and / or processing facility.