System and method for generating design space for chemical process

Abstract

Disclosed herein are methods and systems for generating design spaces for determining one or more polymer composition or polymerization operation settings. In an embodiment, a method may include, in response to receiving data, determining input parameters including at least a minimum, a maximum, and an increment for each aggregate operation setting and / or composition. The method may include generating a design space to include one or more of a plurality of polymer formulations, associated performance, and / or aggregation operation settings via iteratively applying experimental data and input parameters to a trained machine learning model based on the input parameters. The method may include determining, in response to a request for polymer formulation and / or polymerization operation settings that meet or exceed a specified performance, a plurality of polymer formulation and / or polymerization operation settings that meet or exceed the specified performance based on the design space.

Description

Technical Field

[0001] This disclosure relates in its entirety to systems and methods for generating design spaces, and more specifically, to systems and methods for generating interactive design spaces associated with chemical processes or operations and compositions and the chemical products (such as one or more polymerization operations, polymer compositions, or polymer products) produced therefrom, and for optimizing the production of chemical processes or operations and compositions. Background Technology

[0002] Experimental data can be generated based on the execution of various chemical process experiments and / or normal chemical processes and / or operations. However, the generation of experimental data does not include measurement data points for many potential variability (such as the amount of variation in different components in the chemical composition), parameters, and / or settings. Therefore, datasets for a specific chemical process or the chemical products produced by a chemical process may lack data points for many scenarios. Furthermore, such datasets will be specific to that particular chemical process and / or the chemical products produced by that chemical process, so any such dataset will simply be applied to a single chemical process and / or chemical composition.

[0003] Furthermore, first-principles-based models, or in other words, physics-based models, are based on first principles (e.g., the laws of thermodynamics). If well-validated, physics-based models can be used to predict quantities not present in the original dataset and allow for some extrapolation. Moreover, physics-based models are highly interpretable because the parameters have physical meaning. However, such physics-based models exhibit various drawbacks, such as requiring prior behavioral assumptions using expertise from the domain of interest and / or taking a long time (e.g., hours or days) to execute such models to achieve meaningful results. Therefore, using first-principles models to generate data points for a dataset would be impractical, as the time required to populate a small number of data points would be impractical, and such output would consider a small number of data points and would not include many different chemical processes and / or chemical compositions. Summary of the Invention

[0004] In view of the foregoing, the applicant has recognized these problems and other problems in the art, and has recognized the need for design spaces or interactive design spaces associated with multiple chemical processes (e.g., polymerization operations), and for enabling the generation of enhanced systems and methods for updated chemical processes and / or chemical formulations for a variety of chemical products.

[0005] This disclosure relates in general to systems and methods for solving the related problems described above, as well as other problems. In particular, such systems and methods enable users to generate multiple alternative chemical processes or operating parameters, chemical composition amounts, and / or chemical component types in less time than in typical experiments, to produce a specific chemical product that meets or exceeds the performance of the selected chemical product. Such a design space can be interactive, further enabling the generation of highly interpretable results, graphs, and figures associated with any of the multiple chemical processes or operations selected by the user and / or computing device, the chemical composition of the produced chemical product, and the performance of those chemical products (such as polymerization processes, polymer formulations, and / or polymer products).

[0006] Such systems and methods can capture data in real time, and / or in another embodiment, can acquire or capture data from previous or historical chemical processes and / or operations (e.g., from a database or other type of storage device), and preprocess the captured data to form a preprocessed dataset. Such preprocessing may include fitting the data to one of a plurality of trained machine learning models (in other words, selecting a model based on several variables, such as data type and / or other factors), and / or defining one or more input parameters that define different quantities and / or types of different variables and / or components of the chemical process or operation. After preprocessing, the preprocessed dataset may be iteratively applied to one or more selected trained machine learning models based on the input parameters, which include at least, for example, minimum, maximum, and increment values ​​of each of a plurality of polymerization operation settings and a plurality of polymer component components. In another embodiment, the preprocessed dataset (or a portion thereof) may first be applied to one or more physics-based models to produce one or more new polymerization operation parameters, new polymer compositions, and / or new polymerization operation results or performance, and then the updated dataset may be applied to one or more trained machine learning models. In another implementation, preprocessing may include determining which portions of the dataset should be applied to which of one or more trained machine learning models. Such determination may be based on one or more types of data (such as temperature, pressure, composition amount and / or composition type, and other factors).

[0007] Iteratively applying data to a machine learning model can produce outputs. In implementations, the outputs may include simulations, graphs, curves, or other values ​​indicating the performance of the resulting chemical product, and in some implementations, predicted variables, products, properties, or chemical process parameters and / or chemical composition and / or type. Such systems and methods may include using and / or based on the output of such trained machine learning models to add to or update a design space and / or update a dataset, and the resulting dataset may be added to or stored in a design space or a design space database. The design space or design space database may include or be connected to a user interface. Users can input requests for updated chemical processes or operations, parameters associated with the chemical processes or operations, the performance of the chemical product, and / or components of the selected chemical composition. The design space database may output one or more graphical representations and / or data related to multiple chemical processes and / or chemical formulations for producing the chemical product based on inputs that include at least the selected chemical product, and in implementations, satisfy or exceed one or more specified parameters. In the implementation scheme, the design space (or a system communicatively connected to the design space) may utilize nearest neighbor search algorithms, linear search algorithms, binary search algorithms, hash table search algorithms, genetic algorithms and / or simulated annealing to generate varying chemical processes and / or chemical compositions in the design space.

[0008] Therefore, users may be able to quickly access multiple different options related to the adjustment of chemical processes and / or chemical composition to obtain chemical products that meet or exceed the selected parameters or performance. Furthermore, the output of the design space can be interactive, allowing users to adjust one or more parameters or compositions, thereby adjusting the graphical representation in real time and allowing users to quickly refine or adjust chemical processes and / or chemical composition.

[0009] Therefore, embodiments of this disclosure relate to a method for generating a design space for determining one or more polymer compositions or polymerization operation settings. The method may include determining input parameters in response to receiving data corresponding to a plurality of experimental data and / or historical polymerization operation data and / or polymerization process data, the input parameters including at least a minimum, maximum, and increment for each of the plurality of polymerization operation settings and each of the plurality of polymer composition components. In embodiments, the plurality of historical polymerization operation data may include data indicating associated performance of a portion of the plurality of polymerization settings, a portion of the plurality of polymer composition components, the output of one or more historical polymerization operations among the plurality of historical polymerization operations, and the output of one or more historical polymerization operations among the plurality of historical polymerization operations. In the implementation scheme, polymerization operation and / or process data may include data from the entire lifecycle of the chemical or polymer, including but not limited to polymerization settings (e.g., including input material flow data and / or process settings), polymerization conditions (e.g., including temperature, pressure, flow rate, and / or other conditions), powder properties (e.g., bulk density and / or flowability, and other powder properties), process parameters, polymer material properties, formulation of the compound, compounding process settings for producing the compound, settings for producing test samples for property measurements, product properties (e.g., density, molecular chain length, and / or stiffness, and other product properties), composite product properties, data on conversion processes for production applications (e.g., such as films, pipes, and / or injection-molded products), extrusion process properties and / or parameters, blending process properties and / or parameters, and / or application performance measurements. The method may include generating a design space by iteratively and incrementally applying data and input parameters corresponding to the current iteration to one or more trained machine learning models. Each application of the trained machine learning model may generate a synthetic output of the polymerization operation and the performance of the synthetic output based on the input parameters of the current iteration. The design space may include indication output or synthesis output and data corresponding to one or more of multiple polymerization operation settings, multiple polymer compositions, or multiple associated properties. The method may include, in response to a request for one or more polymer compositions or polymerization operation settings that meet or exceed specified properties, determining one or more of multiple polymer compositions or multiple polymerization operation settings that meet or exceed specified properties based on the design space.

[0010] In an implementation, the method may include generating one or more visualizations based on one or more of a plurality of polymer compositions or a plurality of polymerization operation settings that satisfy or exceed specified performance based on the design space. The one or more visualizations may include one or more interactive plots or lists of tables.

[0011] In another embodiment, the method may include determining whether a request includes an unknown polymer composition or an unknown polymerization operation setting before determining one or more of a plurality of polymer compositions or a plurality of polymerization operation settings. The method may also include, in response to determining that the request includes an unknown polymer composition or an unknown polymerization operation setting: (a) generating a prompt for additional data, (b) retraining a trained machine learning model in response to receiving the additional data, and (c) updating the design space by applying data corresponding to the unknown polymer composition or unknown polymerization operation setting to the trained machine learning model.

[0012] In another embodiment, the polymerization operation setup may include one or more polymerization reactor setups, process setups, catalyst type, catalyst quantity, temperature, pressure settings, flow rate, residence time, concentration of reactants (e.g., α-olefins, ethylene, propylene, 1-butene, 1-hexene, hydrogen), extrusion setups, blending setups, or solvent type and / or quantity.

[0013] Another embodiment of this disclosure relates to a method for generating a design space for determining one or more polymer compositions or polymerization operation settings. The method may include receiving data corresponding to multiple experimental data, real-time polymerization operation data, and historical polymerization operation data. The method may include determining input parameters based on the data, the input parameters including at least a minimum, maximum, and increment value for each of the multiple polymerization operation settings and each of the multiple polymer composition components. The method may include iteratively performing the following steps while the maximum value of the multiple polymerization operation settings has not yet been reached. The steps may include (a) iteratively, while the maximum value of the multiple polymer composition components has not yet been reached, (i) generating a simulation of the polymerization operation based on the input parameters of the current iteration by applying data and input parameters corresponding to the current iteration to one or more trained machine learning models, (ii) determining new entries in the design space based on the simulation, (iii) generating a new selection amount of a selected polymer composition component based on the input parameters, and (iv) repeating steps (i), (ii), and (iii) for the new selection amount of the selected polymer composition component. The method may also include (b) generating a newly selected polymerization operation setting based on the input parameters and (c) repeating steps (a) and (b) for the newly selected polymerization operation setting.

[0014] In another embodiment, the method may include generating a user interface to include user input functionality that allows a user to obtain (a) one or more of a plurality of polymerization operation settings or a plurality of polymer component amounts, and (b) the associated performance of each of the plurality of polymerization operation settings or a plurality of polymer component amounts. The method includes initiating a polymerization operation based on a selection of one of the plurality of polymerization operation settings and one of the plurality of polymer component amounts.

[0015] Another embodiment of this disclosure relates to a system for generating a design space for determining one or more polymer compositions or polymerization operation settings. The system may include communication circuitry configured to receive data corresponding to one or more of continuous polymerization operations, polymerization experiments, or historical polymerization operations. The system may include preprocessing circuitry configured to determine input parameters, which include at least each parameter in the data corresponding to a polymerization operation and a minimum, maximum, and increment of each quantity of a plurality of polymer compositions in the data. The system may include modeling circuitry configured to iteratively generate the design space by incrementally applying the data and input parameters corresponding to the current iteration to one or more trained machine learning models. The design space may include data indicating an output or synthesis output and one or more of a plurality of corresponding polymerization operation settings, a plurality of corresponding polymer compositions, or a plurality of corresponding associated properties. The system may include a chemical operation controller configured to, in response to receiving a request for one or more of the polymer compositions or polymerization operation settings that meet or exceed specified properties based on the design space, determine one or more of a plurality of variable polymer compositions or a plurality of polymerization operation settings that meet or exceed specified properties based on the design space.

[0016] In another embodiment, the system may include visualization circuitry. The visualization circuitry may be configured to generate one or more visualizations based on one or more of a plurality of variable polymer compositions or a plurality of polymerization operation settings that satisfy or exceed specified performance based on the design space. The visualization circuitry may be configured to display options to a user interface, thereby allowing selection of one or more visualizations based on the type of each of the one or more visualizations. The visualization circuitry may be configured to display the selected visualization on the user interface in response to the selection of one or more visualizations.

[0017] In another embodiment, the system may include design space circuitry configured to process multiple outputs of one or more trained machine learning models to produce data formatted for the design space.

[0018] In another implementation, the modeling circuitry may be configured to retrain one or more trained machine learning models in response to the selection of one of a plurality of variable polymer compositions and one of a plurality of polymerization operation settings.

[0019] In another embodiment, one of the trained machine learning models may include an image-based machine learning model trained on a series of images of the polymer product, corresponding data, and acceptance or rejection of the polymer product. The request may include a visual aspect of the polymer product based on one or more of the polymer composition or polymerization operation settings.

[0020] In another implementation, the modeling circuitry may be configured to generate the missing data via a physics-based model and based on the data and design space in response to missing data in one or more of a plurality of polymer compositions, a plurality of associated properties, or a plurality of polymerization operation settings.

[0021] Another embodiment of this disclosure relates to a controller for generating a design space for determining one or more polymer compositions or polymerization operation settings. The controller may include inputs / outputs that communicate with polymerization equipment signals. The controller may be configured to determine input parameters in response to receiving experimental data, the input parameters including at least a minimum, maximum, and increment of process parameters in the experimental data. The controller may be further configured to generate a design space to include one or more of a plurality of polymer compositions, associated properties, or polymerization operation settings by iteratively and incrementally applying experimental data and input parameters corresponding to the current iteration based on the input parameters to one or more trained machine learning models. The controller may be configured to determine one or more of a plurality of variable polymer formulations or a plurality of polymerization operation settings that satisfy or exceed the specified properties based on the design space in response to receiving a request for one or more of a polymer formulation or a plurality of polymerization operation settings that satisfy or exceed the specified properties based on the design space. The controller may be configured to display one or more visualizations generated based on satisfying or exceeding one or more of the specified properties based on the design space.

[0022] In another embodiment, one or more visualizations include one or more graphical or tabular visualizations. Furthermore, the request includes applying Quality Key (CTQ) inputs, and wherein the controller is configured to convert the applied CTQ inputs into material for processing the CTQs.

[0023] Additional and / or alternative objects, features, and advantages of this disclosure will become apparent to those skilled in the art from the accompanying drawings, detailed description, and examples herein. However, the applicant hereby states that while the drawings, detailed description, and examples illustrate certain embodiments of this disclosure, they are provided for illustrative purposes only and are not intended to limit or imply any particular limitations. Furthermore, certain changes and modifications within the spirit and scope of the disclosed technology will become apparent to those skilled in the art based on this detailed description. Attached Figure Description

[0024] The disclosed aspects, features, and advantages of this disclosure will be better understood with reference to the following description, examples, claims, and drawings.

[0025] However, the applicant hereby states that the accompanying drawings illustrate certain embodiments of this disclosure and should not be considered as a limitation on the breadth and scope of this disclosure: Figure 1 This is a schematic diagram of a system for generating design space according to certain embodiments of this disclosure; Figure 2 This is another schematic diagram of an apparatus for generating design space according to certain embodiments of the present disclosure; Figure 3 This is a schematic diagram of a controller for generating a design space according to certain embodiments of this disclosure; Figure 4A and Figure 4B This is an example of a user interface for displaying results generated via a design space, according to certain embodiments of this disclosure; Figure 5A An exemplary information flow for use in a virtual experiment with an extruder is illustrated; Figure 5B This is a flowchart for generating a design space according to certain embodiments of this disclosure; and Figure 6A and Figure 6B These are other flowcharts for generating design spaces according to certain embodiments of this disclosure. Detailed Implementation

[0026] To gain a more detailed understanding of the features and advantages of embodiments of the systems and methods disclosed herein, as well as other features and advantages that will become apparent, a more specific description of embodiments of the systems and methods briefly outlined above can be obtained by referring to the following detailed description of their embodiments, one or more of which are further illustrated in the accompanying drawings, which form part of this specification. However, it should be noted that the drawings only illustrate various embodiments of the systems and methods disclosed herein and should not be considered as limiting the scope of the systems and methods disclosed herein, as it may also include other effective embodiments.

[0027] This disclosure relates in general to systems and methods for solving design space generation and other related problems as described herein. In particular, such systems and methods can enable faster analysis than typical analyses, providing multiple options to meet or exceed specified performance levels of chemical products. Furthermore, these multiple options can be presented interactively, allowing users to adjust or optimize the chemical composition and / or chemical processes of a specific or selected chemical product. Therefore, users can obtain more efficient and optimized chemical processes and compositions in a relatively short time (compared to the time spent on experiments), rather than conducting lengthy experiments.

[0028] Furthermore, the design space can be used to adjust chemical operating settings and / or chemical composition, such as, in a non-limiting example, adjustments to the temperature and / or pressure within the polymerization reactor, adjustments to the type and / or amount of chemicals or polymers included in or in the manufacture of the polymer composition, and / or catalysts used to produce the chemical composition, as well as other adjustments to the chemical operating settings.

[0029] This disclosure relates in general to systems and methods for solving the related problems described above, as well as other problems. In particular, such systems and methods enable users to generate multiple alternative chemical processes or operating parameters, chemical composition amounts, and / or chemical component types to produce a specific chemical product in less time than in a typical experiment. Such a design space can be interactive, further enabling the generation of highly interpretable results, graphs, and figures associated with any of the multiple chemical processes or operations selected by the user and / or computing device, and the chemical composition of the chemical product (such as polymerization processes and / or polymer formulations).

[0030] Such systems and methods can utilize data (whether captured in real time or from existing data sources) to construct a design space. This design space can include data indicating multiple outputs and / or synthetic outputs (e.g., outputs of polymerization operations such as chemical conversion, extrusion, or blending operations) and data corresponding to one or more of multiple polymerization operation settings, multiple polymer composition components, or multiple associated properties. The system and methods can utilize such data to determine or define input parameters. Input parameters can include (for each of one or more different types of input parameters) minimum or starting values, maximum or ending values, and / or incremental values ​​(in other words, incrementing the variable until a maximum value is reached). The system or method can define input parameters with multiple values ​​based on various factors. For example, variables selected for input parameters can be based on the properties of the desired final chemical product (e.g., tensile strength, modulus of elasticity, toughness, pattern or visible pattern, and / or temperature resistance or threshold, and other performance variables). In other words, variables that can be applied to one or more trained machine learning models can include variables that have or are determined to provide or produce the desired performance.

[0031] As a non-limiting example, the production of the selected chemical product may include up to 28 components or ingredients (in other examples, more or fewer components or ingredients may be used). The amount of each of the components or ingredients may vary or remain static before being applied to the machine learning model. Furthermore, for such a chemical product, the selected corresponding outputs (in other words, material properties) may be expected, for example, up to 7 corresponding outputs (and in other examples, additional corresponding outputs). The incremental values ​​of the amounts of the components or ingredients and each of the corresponding outputs can be as small as a tenth of a digit, a hundredth of a digit, or even smaller. Therefore, thousands, and in some cases even more, of variations of the components or ingredients and their corresponding outputs can be applied to one or more machine learning models. In another example, the incremental values ​​of any component or ingredient and its corresponding output can be small (such as a tenth or a hundredth of a digit, or even smaller) and / or random.

[0032] Different variables and / or combinations of variables can be applied to different trained machine learning models. Systems and methods can determine which model applies different variables and / or combinations of variables based on multiple factors, including but not limited to the type of variables and / or previous models used for similar variables, as well as other variables.

[0033] Depending on the input parameters, applying such data to one or more machine learning models can be iterative and incremental. For example, the amount of a specific component in a chemical composition can be varied or adjusted based on incremental values, and then applied to one or more trained machine learning models until a maximum value is reached. After reaching the maximum value, another variable in the chemical process or operation can be incremented, and the amount of the specific component in the chemical composition can then be changed based on the incremental values ​​until the maximum value is reached again, and so on. Such examples are not limiting, and it will be understood that any variable can be incremented while others remain static, and then applied to one or more machine learning models and incremented again until the incremented variable reaches its maximum value. Furthermore, after reaching that maximum value, one of the other static variables can be incremented and the process repeated.

[0034] This iterative application of data to a machine learning model or a trained machine learning model can produce outputs. In embodiments, the outputs may include simulations, graphs, curves, or other numerical indicators of predictor variables and / or chemical properties, and in some embodiments, predictions of chemical process parameters and / or chemical composition and / or type. Such systems and methods may include using the outputs of such trained machine learning models to add to or update a design space and / or update a dataset, and the resulting dataset may be added to or stored in the design space or a design space database. The design space or design space database may include or may be connected to a user interface. Users can input requests for updated chemical processes or operations, parameters associated with the chemical processes or operations, properties of chemical products, and / or components of selected chemical compositions. The design space database may output one or more graphical representations and / or data related to the production of chemical products that meet or exceed one or more specified parameters, based on inputs that include at least selected chemical products and the properties of the chemical products. In embodiments, the design space may utilize nearest neighbor search algorithms or other search algorithms to generate varying chemical processes and / or chemical compositions.

[0035] Therefore, users may be able to quickly access multiple different options related to the adjustment of chemical processes and / or chemical composition to obtain chemical products that meet or exceed the selected parameters. Furthermore, the output of the design space can be interactive, allowing users to adjust one or more parameters or compositions, thereby adjusting the graphical representation in real time and allowing users to quickly refine or adjust chemical processes and / or chemical compositions.

[0036] The following definitions are provided to clarify certain terms and phrases in this disclosure and are in no way intended to unnecessarily or inappropriately limit any implementation or aspect thereof.

[0037] When used in conjunction with the terms “comprising,” “including,” “containing,” or “having” in the claims or description, the word “a” or “an” may mean “a”, but it is also consistent with the meanings of “one or more,” “at least one,” and “one or more.”

[0038] The words “comprising” (and any form of inclusion, such as “comprise” and “comprises”), “having” (and any form of having, such as “have” and “has”), “including” (and any form of inclusion, such as “includes” and “include”), or “containing” (and any form of containing, such as “contains” and “contain”) are inclusive or open-ended and do not exclude additional, unlisted elements or steps of procedure.

[0039] Figure 1This is a schematic diagram of a system 100 for generating a design space according to certain embodiments of this disclosure. Such a design space can be used to generate variable chemical process or operating parameters and / or chemical composition components that meet or exceed selected performance characteristics of a chemical product based on inputs from computing devices such as via user interfaces (UI) 122A, 122B, or up to 122N, or from computing devices 124A, 124B, or up to 124N. The design space can be a database including data indicating multiple outputs (e.g., outputs of physical generation and outputs of synthetic generation, in other words, outputs determined by applying data to a model) and data indicating one or more of multiple polymerization operation settings, multiple polymer composition components, or multiple associated performance characteristics. Such a system 100 may include a design space system 102. Design space system 102 may include a processor 104 or multiple processors and a memory 106. The memory 106 may store or include instructions. Instructions may include preprocessing instructions 108, design space instructions 110, machine learning model 112 (or trained machine learning model), physics-based model 114, and / or visualization instructions 120. Design space system 102 may be communicatively connected to controller 126 and / or database 132. Controller 126 may be connected to various equipment 130 and / or polymerization reactor 128 (or other types of reactors) located in the plant. Controller 126 may be further connected to various sensors, analyzers, and / or instruments associated with equipment 130 and / or polymerization reactor 128 (and other components and / or equipment located in the plant or other locations), thus enabling controller 126 and design space system 102 to capture data during polymerization operations and allowing the design space to be continuously updated and / or refined over time. In another embodiment, design space system 102 may be directly connected to equipment 130 and polymerization reactor 128 (and / or associated sensors, analyzers, and / or instruments and / or other components or equipment). In such examples, the design space system 102 may be or may include the functions of a controller (such as, for example, controlling the operating parameters and / or settings of equipment 130 and / or polymerization reactor 128). In embodiments, sensors, analyzers, and / or instruments may include temperature sensors, pressure sensors, flow meters, densitometers, spectrometers, gas chromatographs, other chemical analyzers, and / or other sensors or instruments to determine or measure the properties of equipment 130 and / or polymerization reactor 128 or the fluids therein.In another embodiment, equipment 130 may include pumps, compressors, control valves, extruders, other types of reactors, furnaces, heaters, coolers, distillation columns, separation columns, heat exchangers, flash vessels, purge tanks, storage containers, decanters, cyclone separators, pressure sensors, membrane separators, flow sensors, temperature sensors, composition measurement devices (such as gas chromatographs, Raman spectroscopy, NMR, or other measurement devices, as those skilled in the art will understand), and / or other equipment utilized at a plant configured to perform polymerization operations. Furthermore, as described in further detail below, design space system 102 may be connected to computing devices 124A, 124B, and up to 124N (e.g., via UI 122A, 122B, and up to 122N). Design space system 102 may generate UI 122A, 122B, and up to 122N to enable computing devices 124A, 124B, and up to 124N to interact with design space system 102. For example, a user can request multiple different compositional and / or process or operational parameters for a specific chemical product from computing devices 124A, 124B, or up to 124N via UIs 122A, 122B, or up to 122N. Design space system 102 can utilize instructions stored in memory 106 (such as visualization instructions 120) to generate such UIs 122A, 122B, and up to 122N and / or other visualization and interactive functions, as described further in detail below.

[0040] As noted, memory 106 may include preprocessing instructions 108. In embodiments, preprocessing instructions 108 may be configured to receive data, when executed by processor 104, from controller 126 (and / or, in another embodiment, directly from equipment 118, polymerization reactor 116, and associated sensors, instruments, and / or analyzers), from computing devices 124A, 124B, and up to 142N, and / or from database 132 and other data sources. Before generating the design space, design space system 102 may obtain, request, and / or receive a large amount of data corresponding to various different chemical products, the chemical processes or operations that manufacture these chemical products, the different compositions or formulations that manufacture these chemical products, and the performance exhibited by the chemical products after production according to the chemical processes or operations and / or chemical compositions. To populate, generate, or determine the initial design space, preprocessing instructions 108 may first determine the minimum, maximum, and increment values ​​of each or a portion of the chemical operation parameters or settings and / or the amount of each component in the chemical composition. Such values ​​can be used to initialize the design space. As noted, the design space can be updated or refined over time. New data associated with each situation can also be preprocessed when new chemical operations are performed and / or when chemical compositions or formulations are adjusted.

[0041] In one implementation, the design space system 102 can receive data in real time and / or continuously while a chemical operation (such as polymerization) is in progress, in operation, or being performed. In another implementation, the data can be transferred to or stored in a database 132. In such an implementation, the design space system 102 can periodically obtain or receive data from the database 132. The preprocessing instruction 108 can also determine, upon execution, whether the received data includes known chemical operations and / or chemical compositions. Upon execution, the preprocessing instruction 108 can compare the results corresponding to the received data with the current design space. In another implementation, each data point can include a label. The label can indicate a specific chemical process or operation. Based on the chemical operation to be analyzed, data points with selected or associated labels (indicating a specific polymerization operation) can be separated and stored as subsets of data. In such examples, the preprocessing instruction 108 can determine one or more types of machine learning models 112 for different portions of the data (e.g., in addition to physics-based models in some implementations, utilizing specific models to populate temperature data for selected chemical operations and / or chemical products). In another example, multiple chemical operations can be performed within a selected timeframe. During such a timeframe, data received at selected times can be relevant to polymerization experiments, for example, to determine (i) how a particular component behaves, (ii) how the component is prepared, (iii) the properties exhibited by the component, and / or (iv) how various factors affect the final product (such as temperature, pressure, flow rate, and other factors). As noted, data indicative of such experiments can be separated and stored as a subset of data. The remainder of the data can be stored (e.g., in memory 106 or in database 132) for later use, or, in another embodiment, deleted or removed from the design space system 102.

[0042] As noted, physics-based models or first-principles models can be used to predict or determine the value of a specific input that has not yet been measured. Such input can be used as input to machine learning model 112. For example, a physics-based model can use measured temperature and volume to determine pressure, which can be used as input to machine learning model 112.

[0043] Once a portion or subset of the data has been selected, preprocessing instruction 108 can smooth the data subset. In other words, when executing preprocessing instruction 108, any peripheral data points or errors in the data subset can be removed to form a smoothed data subset. As those skilled in the art will understand, such a smoothing process can include various smoothing algorithms, such as moving average smoothing, exponential smoothing, double exponential smoothing, triple exponential smoothing, and other techniques. In an implementation, the smoothing algorithm can be selected by the user.

[0044] Once the dataset is available, design space instructions 110 can be executed. Design space instructions 110 can determine which part of the dataset is applied to machine learning model 112 (and in some implementations, to a physics-based model 114). Design space instructions 110 can apply the dataset to machine learning model 112 when executed. Design space instructions 110 can subsequently increment one or more specific variables applied to the next application of machine learning model 112 when executed.

[0045] In implementations, one or more machine learning models 112 may be trained based on historical data. Historical data may include previous polymerization operations, the resulting output or polymer composition, any adjustments made based on data collected from such operations, and / or the performance of the resulting output chemical products. In implementations, one or more different machine learning models may be trained based on the performance of such models for such sets or types of data, and each such machine learning model may be of a different type. For example, a chosen model may be used when considering temperature over time, while another type of model may be considered for pressure. In such examples, relevant portions of a dataset of a particular chemical operation or chemical composition may be applied to the corresponding one or more machine learning models 112. Based on applying data (e.g., chemical process or operation parameters and / or chemical composition) to one or more machine learning models 112, one or more machine learning models 112 may generate one or more graphs, performance predictions, output predictions (e.g., “synthetic output” or predicted output rather than physically tested output), and / or other operational or process data. Such outputs (in some implementations, also including measured data and / or synthetic data) may be added to a design space. After multiple iterative applications, the design space may include multiple different chemical processes or operations, chemical compositions, performance, and / or other relevant data. Computing devices, controllers, and / or users can leverage this design space to quickly and without experimentation optimize chemical processes and / or adjust chemical compositions to produce chemical products that meet or exceed specified performance.

[0046] In an implementation, the design space system 102 may include visualization instructions 120. When executed, visualization instructions 120 may generate UIs 122A, 122B, and up to 122N, or a portion of UIs 122A, 122B, and up to 122N, for one or more computing devices UIs 124A, 124B, and up to 124N. UIs 122A, 122B, and up to 122N may include a search bar or other search functionality (such as file or data upload functionality) to allow or enable a user to request or search for variable chemical compositions, chemical operations, and / or specific chemical product properties. Upon such a search or request, visualization instructions 120 may generate results including a selected number of chemical operation parameters and / or chemical compositions that meet or exceed specified characteristics. Visualization instructions 120 may utilize various search algorithms to generate such results, including but not limited to nearest neighbor search algorithms or another algorithm configured to select a specified number of results that match or approximate the requested chemical composition, chemical operation, and / or chemical product properties.

[0047] In implementations, the design space system 102 may be connected to multiple controllers, plants (e.g., connected to computing devices and / or other devices located at the plant) and / or other locations. In such implementations, the design space system 102 may provide the optimization capabilities described above and herein to each of these locations. Furthermore, the design space system 102 may utilize data collected at each location to further refine machine learning models and / or add data related to new chemical processes and / or chemical compositions. In implementations, polymerization operations or settings may include converting monomers / oligomers or other chemicals into polymers, blending chemicals, or polymers, and / or extruding chemicals or polymers. In implementations, polymerization operation settings may include one or more polymerization reactor settings, process settings, catalyst type, catalyst quantity, co-catalyst type and quantity, temperature, pressure settings, flow rate, residence time, blending time, concentration of reactants (e.g., α-olefins, ethylene, propylene, 1-butene, 1-hexene, hydrogen), or solvent type and quantity. In another embodiment, polymerization operation and / or process data may include data from the entire lifecycle of the chemical or polymer, including but not limited to polymerization setup, process parameters, polymer material properties, formulation of compounds, compounding process setup for producing compounds, setup for producing test samples for property measurements, composite product properties, data on conversion processes for producing applications (e.g., membranes, pipes, and / or injection molded products), blending data, extrusion data, and / or application performance measurements.

[0048] In some examples, the design space system 102 may be a computing device. The term "computing device" is used herein to mean any or all of a programmable logic controller (PLC), controller, programmable automation controller (PAC), industrial computer, server, virtual computing device or environment, desktop computer, personal data assistant (PDA), laptop computer, tablet computer, smartbook, handheld computer, personal computer, smartphone, virtual computing device, cloud-based computing device, and any other similar electronic device equipped with at least one processor and any other physical components necessary to perform the various operations described herein.

[0049] The term "server" or "server device" is used to refer to any computing device capable of being used as a server, such as a primary switching server, web server, mail server, document server, or any other type of server. A server can be a dedicated computing device or a server module (e.g., an application) hosted by the computing device, enabling the computing device to operate as a server. A server module (e.g., a server application) can be a full-featured server module or a lightweight or secondary server module (e.g., a lightweight or secondary server application) configured to provide synchronization services between dynamic databases on the computing device. A lightweight or secondary server can be a simplified version of server-type functionality that can be implemented on a computing device such as a smartphone, enabling it to function as an internet server (e.g., an enterprise email server) only to the extent necessary to provide the functionality described herein.

[0050] As used herein, "non-transitory machine-readable storage medium" or "memory" can be any electronic, magnetic, optical, or other physical storage device that contains or stores information such as executable instructions, data, etc. For example, any machine-readable storage medium described herein can be any one or a combination of random access memory (RAM), volatile memory, non-volatile memory, flash memory, storage drives (e.g., hard disk drives), solid-state drives, any type of storage disk, etc. Memory may store or include instructions executable by a processor.

[0051] As used herein, "processor" or "processing circuitry" may include, for example, one or more processors included in a single device or distributed across multiple computing devices. Processors (such as...) Figure 1 The processor 104 shown may be at least one or a combination of a central processing unit (CPU), a semiconductor-based microprocessor, a graphics processing unit (GPU), a field-programmable gate array (FPGA) for retrieving and executing instructions, a real-time processor (RTP), or other electronic circuits suitable for retrieving and executing instructions stored on a machine-readable storage medium.

[0052] In the implementation scheme, one or more machine learning models 112 may be supervised or unsupervised learning models. In the implementation scheme, one or more machine learning models 112 may be based on decision trees, random forest models, random forests utilizing bagging or boosting (such as gradient boosting), K-nearest neighbors, neural network methods, support vector machines (SVMs), lasso-based models, other supervised learning models, other semi-supervised learning models, other unsupervised learning models, or some combination thereof, as will be readily understood by those skilled in the art.

[0053] In another embodiment, one of the one or more machine learning models may be an image and / or recognition machine learning model. Such a model can be trained using multiple images of chemical products exhibiting a specific pattern. In this embodiment, for a particular chemical product, the design space may include image or pattern data, as well as labels or other indicators for identifying such images or patterns. In such embodiments, a user can search for specific chemical products exhibiting such patterns. Such searches may be text- and / or image-based.

[0054] Figure 2 This is another schematic diagram of a system for generating a dataset associated with an aggregation operation, according to certain embodiments of this disclosure. Such a system may consist of processing circuitry 202, memory 204, communication circuitry 206, preprocessing circuitry 208, design space circuitry 210, modeling circuitry 212, visualization circuitry 214, and chemical operation controller circuitry 216, each of which will be described in more detail below. Although the various components are only... Figure 2 The device is shown as being connected to the processing circuit 202, but it should be understood that the device 200 may also include a bus for transmitting information between any combination of the various components of the device 200. Figure 2 (Not explicitly shown in the text). The device 200 can be configured to perform various operations described herein, such as those described above. Figure 1 And the following text combined Figures 3 to 6B The operations described.

[0055] Processing circuitry 202 (and / or a coprocessor or auxiliary processor or any other processor otherwise associated with the processor) may communicate with memory 204 via a bus for transferring information between components of the device. Processing circuitry 202 may be embodied in a variety of different ways and may, for example, include one or more processing devices configured to execute independently. Furthermore, the processor may include one or more processors configured in series via a bus to enable independent execution of software instructions, pipelines, and / or multithreading.

[0056] Processing circuitry 202 may be configured to execute software instructions stored in memory 204 or otherwise accessible to processing circuitry 202 (e.g., software instructions stored on a separate storage device). In some cases, processing circuitry 202 may be configured to execute hard-coded functions. Thus, whether configured by hardware or software methods, or by a combination of hardware and software, processing circuitry 202 represents an entity (e.g., physically embodied in circuitry) capable of performing operations according to various embodiments of this disclosure when configured accordingly. Alternatively, as another example, when processing circuitry 202 is embodied as an executor of software instructions, the software instructions may specifically configure processing circuitry 202 to perform the algorithms and / or operations described herein when executing the software instructions.

[0057] Memory 204 is non-transitory and may include, for example, one or more volatile and / or non-volatile memories. In other words, memory 204 may be, for example, an electronic storage device (e.g., a computer-readable storage medium). Memory 204 may be configured to store information, data, content, applications, software instructions, etc., to enable device 200 to perform various functions according to the example embodiments contemplated herein.

[0058] Communication circuit 206 can be any means configured to receive and / or transmit data from / to a network and / or any other device, circuit, or module communicating with device 200, such as devices or circuits embodied in hardware or a combination of hardware and software. In this regard, communication circuit 206 may include, for example, a network interface for enabling communication with a wired or wireless communication network. For example, communication circuit 206 may include one or more network interface cards, antennas, buses, switches, routers, modems, and supporting hardware and / or software, or any other device suitable for enabling communication via a network. Furthermore, communication circuit 206 may include processing circuitry 202 for transmitting such signals to the network or for processing signals received from the network. In embodiments, communication circuit 206 may enable the reception of aggregation operation data (in this example, including polymer composition or other data related to aggregation operation) and the transmission of aggregation operation settings to associated equipment and / or devices.

[0059] The apparatus 200 may include preprocessing circuitry 208 configured to preprocess received data. Preprocessing the received data may include determining input parameters for the received data and / or fitting the data to one or more machine learning models. Preprocessing circuitry 208 may perform these operations using processing circuitry 202, memory 204, or any other hardware components included in the apparatus 200, as described below in conjunction with Figures 5 to 6. Figure 6BThe pretreatment circuit 208 may further utilize the communication circuit 206 as described above to collect data (such as real-time chemical operation data and / or historical chemical operation data) from various sources (e.g., one or more different components or devices in the plant, such as polymerization reactors and associated sensors or analyzers; databases; and / or other data sources). The output of the pretreatment circuit 208 may be transmitted to other circuits of the apparatus 200 (such as design space circuit 210).

[0060] Additionally, the device 200 includes a design space circuit 210 that can apply or transmit data to a modeling circuit 212 based on input parameters. For example, the design space circuit 210 can generate sets or subsets of data based on input parameters and then send each set to the modeling circuit 212 for application to the model. The design space circuit 210 can also receive output from the modeling circuit 212 and, upon receiving output, apply or add it to a design space database. Other information, such as labels and / or original corresponding data, can be added to the output added to the design space database. In another embodiment, the design space circuit 210 can iteratively apply data to the modeling circuit 212. For example, the design space circuit 210 can apply a portion of data corresponding to chemical operations and chemical compositions to the modeling circuit 212. The design space circuit 210 can then increment one or more variables in the data corresponding to chemical operations and chemical compositions according to the input parameters and apply the adjusted and / or incremented data to the modeling circuit 212. Depending on the input parameters, the design space circuit 210 can continue to increment variables and apply the adjusted and / or incremented data to the modeling circuit 212 until one of the variables reaches its maximum value. The design space circuit 210 can then increment other variables and continue applying such data to the modeling circuit 212. Therefore, the design space circuit 210 can form, generate, or produce a design space including multiple chemical operations, chemical compositions, and corresponding chemical product properties. The design space circuit 210 can perform these operations using the processing circuit 202, the memory 204, or any other hardware components included in the device 200, as described below in conjunction with Figures 5 to 6. Figure 6B The design space circuit 210 can further utilize the communication circuit 206 to collect data (e.g., preprocessed data and / or input parameters) from various sources such as the preprocessing circuit 208, the controller 126, or the polymerization reactor 116 and / or the equipment 118 at the plant; receive outputs from the modeling circuit 212; and / or add, post-process, and / or combine the outputs from the modeling circuit 212 into the design space. The outputs of the design space circuit 210 can be transmitted to other circuits of the device 200 (such as the modeling circuit 212).

[0061] The device 200 also includes modeling circuitry 212, which can apply received data to one or more trained machine learning models based on the type of received data, indicators transmitted with the data, and / or determination of which of the one or more trained machine learning models is best suited for the data. Applying the data to one or more trained machine learning models in this way can produce outputs including one or more predictions, probabilities, simulations (such as simulated performance of potential chemical products), and / or graphs or plots associated with the performance of the resulting chemical product (e.g., a chemical product is the result of chemical operations and chemical composition). Modeling circuitry 212 can perform these operations using processing circuitry 202, memory 204, or any other hardware components included in the device 200, as described below in conjunction with Figures 5 to 6. Figure 6B The machine learning model circuit 212 can also receive data from the design space circuit 210 using the communication circuit 206. The output of the modeling circuit 212 can be transmitted to other circuits of the device 200, such as the visualization circuit 214 and / or the chemical operation controller circuit 214.

[0062] The device 200 also includes visualization circuitry 214, which can generate a user interface allowing a user to search for or determine multiple variable chemical operations and / or chemical compositions that meet or exceed specified performance based on data in the design space, and / or enabling the user to initiate chemical operations with chemical compositions to produce chemical products. In another embodiment, visualization circuitry 214 can be configured to automatically initiate polymerization operations or other chemical operations based on a search for specified performance of a particular chemical operation and / or chemical composition. In such embodiments, a user can initiate such a search, and visualization circuitry 214 can then initiate a chemical operation based on performance matching from the design space, available chemical compositions, and / or current equipment availability and / or downtime. Visualization circuitry 214 can generate a user interface including interactive fields to allow the user to further adjust the chemical operation and / or chemical composition and view the corresponding chemical performance. Furthermore, visualization circuitry 214 enables the user to initiate or start chemical operations to produce chemical products. Visualization circuitry 214 can perform these operations using processing circuitry 202, memory 204, or any other hardware components included in device 200, as described below in conjunction with Figures 5 to 6. Figure 6B The visualization circuit 214 can also receive data from the design space circuit 210 using the communication circuit 206. In an embodiment, in response to a received signal initiating a chemical process, the visualization circuit 214 can transmit such an initiation signal to the chemical operation controller circuit 216.

[0063] The apparatus 200 also includes a chemical operation controller circuit 216, which can control or initiate polymerization operations based on signals from the visualization circuit 214. The chemical operation controller circuit 214 can perform these operations using the processing circuit 202, the memory 204, or any other hardware components included in the apparatus 200, as described below in conjunction with Figures 5 to 6. Figure 6B The chemical operation controller circuit 216 may further utilize communication circuit 206 to receive initiation signals and / or communicate with the selected plant or facility controller, such that the chemical operation controller circuit 216 may initiate a chemical operation at the selected plant or facility.

[0064] Although components 202-216 are described in part using functional language, it should be understood that a particular implementation necessarily includes the use of particular hardware. It should also be understood that some of these components 202-216 may include similar or common hardware. For example, in some embodiments, preprocessing circuitry 208, design space circuitry 210, modeling circuitry 212, visualization circuitry 214, and chemical manipulation controller circuitry 216 may each from time to time utilize the use of processing circuitry 202, memory 204, or communication circuitry 206, so that duplicate hardware is not required to facilitate the operation of these physical elements of device 200 (although in some embodiments, dedicated hardware elements may be used for any of these components, such as those where enhanced parallelism may be desired). Therefore, the use of the term "circuit" with respect to elements of the device should be interpreted as necessarily including particular hardware configured to perform the functions associated with the particular element described herein. Of course, while the term "circuit" should be understood broadly to include hardware, in some embodiments, the term "circuit" may additionally refer to software instructions that configure the hardware components of device 200 to perform the various functions described herein.

[0065] Although the preprocessing circuit 208, design space circuit 210, modeling circuit 212, visualization circuit 214, and chemical operation controller circuit 216 described above may utilize processing circuit 202, memory 204, or communication circuit 206, it should be understood that any of these elements of device 200 may include one or more dedicated processors, specially configured field-programmable gate arrays (FPGAs), or dedicated interface circuits (ASICs) to perform their respective functions, and may accordingly utilize processing circuit 202 which executes software stored in memory or memory 204, and communication circuit 206 which enables any functions not performed by dedicated hardware elements. However, in all embodiments, it should be understood that preprocessing circuit 208, design space circuit 210, modeling circuit 212, visualization circuit 214, and chemical operation controller circuit 216 are implemented via specific machines designed to perform the functions described herein in conjunction with such elements of device 200.

[0066] In some implementations, various components of device 200 can be remotely hosted (e.g., hosted by one or more cloud servers) and therefore do not need to physically reside on the corresponding device 200. Consequently, some or all of the functions described herein can be provided by third-party circuitry. For example, a given device 200 can access third-party circuitry via any kind of network connection facilitating the transmission of data and electronic information between device 200 and one or more third-party circuitries. The device 200 can then communicate remotely with one or more of the other components described above that include device 200.

[0067] As will be understood based on this disclosure, the exemplary embodiments contemplated herein may be implemented by device 200 (or by controller 302). Furthermore, some exemplary embodiments (such as those for...) Figure 1 and Figure 3 The described implementation may take the form of a computer program product comprising software instructions stored on at least one non-transitory computer-readable storage medium (such as memory 204). Any suitable non-transitory computer-readable storage medium may be utilized in such implementations; some examples are non-transitory hard disks, CD-ROMs, flash memory, optical storage devices, and magnetic storage devices. It should be understood that for implementations using... Figure 2 The apparatus 200 described herein embodies certain devices that load software instructions onto a computing device or apparatus to produce a dedicated machine including means for implementing the various functions described herein.

[0068] Figure 3 This is a schematic diagram of a controller, according to certain embodiments of the present disclosure, for generating a complete or substantially complete dataset associated with an polymerization operation for adjusting polymerization operation settings. As described herein, the control system may be controller 302, one or more controllers, a PLC, a SCADA system, a computing device, and / or other components to generate a design space for adjusting and / or initiating chemical operations (e.g., Figure 1 and Figure 2 (The components, devices, or apparatus described herein). Controller 302 may include one or more processors (e.g., processor 304) to execute instructions stored in memory 306. In this example, memory 306 may be a machine-readable storage medium.

[0069] As used herein, “signal communication” refers to electrical or wireless communication, such as hardwiring two components together, as understood by those skilled in the art. For example, wireless communication can be or includes Wi-Fi. ® ,Bluetooth ®ZigBee, near-field communication, or other wireless communication methods as will be understood by those skilled in the art. Additionally, signal communication may include one or more intermediate controllers, relays, or switches positioned between elements that communicate with each other.

[0070] As noted, memory 306 may store instructions, such as preprocessing instructions 308, that can be executed by processor 304 to preprocess data. Controller 302 may be connected to and / or receive data from one or more of the following: polymerization reactor 322, sensors or other devices that determine some characteristics of polymerization reactor 322, equipment 324 at one or more locations (such as a plant, facility, or other location), sensors or other devices that determine some characteristics of equipment 324, database 326, and / or receive data from a user interface 328. Controller 302 may obtain data from each of one or more data sources. Preprocessing instructions 308 may define input parameters and / or preprocess data. In response to receiving such data, preprocessing instructions 308 may be executed to preprocess such data. Preprocessing instructions 308 may determine input parameters of the data at execution. The determination of input parameters may be based on the type of data received, the type of machine learning model available, and / or provide an indication of which variables in the data can utilize the input parameters. Input parameters may include the minimum value of one or more corresponding variables, the maximum value of one or more corresponding variables, and / or an increment value that indicates the next value to which the variable should be adjusted when the variable and other corresponding data are applied to one or more machine learning models.

[0071] The controller 302 may include a design space generation instruction 312. The design space generation instruction 312 can, upon execution, apply data to a trained machine learning model 314 based on input parameters and then increment the data. The design space generation instruction 312 can, upon execution, increment variables and then reapply the data to the machine learning model. Furthermore, the design space generation instruction 312, upon execution, can determine which of the trained machine learning models 314 to utilize based on factors such as the component and / or data type, the available trained machine learning model 314, whether the chemical operation and / or chemical composition is unknown, and / or input received via the user interface 328. The design space generation instruction 312 can also receive output from the trained machine learning model 314 and, upon receiving the output, update the design space using the output and the corresponding data used to generate the output. The output may include the simulation performance of the resulting chemical composition.

[0072] The controller 302 may include data visualization instructions 316. Data visualization instructions 316, when executed, can generate a graphical user interface (GUI) or a portion of a GUI to be displayed via a user interface. Data visualization instructions 316 can enable the user interface to search for various chemical operations and / or chemical compositions that meet specified performance requirements. Data visualization instructions 316 can further generate and display one or more interactive charts, graphs, and / or tables, which enable further refinement of the chemical composition and / or chemical operations.

[0073] Controller 302 may include process and / or composition adjustment and / or initialization instructions 318. When executed, process and / or composition adjustment and / or initialization instructions 318 may result in adjustments or updates to chemical operating parameters and / or chemical composition. In one embodiment, adjustments or updates to chemical operating parameters and / or chemical composition may be transmitted to polymerization reactor 322 and / or other equipment 324, thus causing polymerization reactor 322 and / or other equipment 324 to operate under those settings for subsequent operations. In another embodiment, a user may select chemical operating parameters and / or chemical composition via a GUI generated by data visualization instructions 318, and based on such selection, process and / or composition adjustment and / or initialization instructions 318 may cause polymerization reactor 322 and / or other equipment 324 to operate under those selected settings.

[0074] Go to Figure 4A and Figure 4B A graphical user interface (GUI) 402 is provided, which exemplifies a user interface corresponding to a design space or a search within a design space. As noted, a user can search for specific chemical processes and / or chemical product properties. Such a GUI 402 may include a search box 404 for user input of a search. The GUI 402 may also include buttons 406 for selecting the view type of the data (e.g., as a graph, chart, interactive plot, and / or list of tables or table visualization) and whether to download the data, a portion of the data, and / or the currently displayed chart, graph, and / or table. Such a search can generate, for example, one or more interactive graphical displays. For example, a spider diagram 410 with multiple formulation or composition suggestions 412 can be generated. A user can move any point on the spider diagram to adjust the desired chemical process and / or chemical composition properties. Based on this adjustment, adjusted chemical formulations and / or chemical compositions can be generated. As noted, the GUI 402 can generate additional views; for example, the GUI 402 can generate a table including the properties of one or more chemical compositions.

[0075] In the example, to utilize design space, as described above, system 100, device 200, or controller 302 may receive datasets, as shown in Table 1 below. In other embodiments, substantially more data may be received. Such data may include various polymers in a formulation or composition, modifiers added to the compound, formulation, or composition, and / or measured outputs such as ash content (measured as a weight percentage), melt flow rate (measured as decigrams per minute), tensile modulus (measured as megapascals), Charpy impact at various temperatures (measured as kilojoules per square meter), fracture type, and / or visual aesthetics, as well as other measurable outputs. For example, measured data or sample data may include the weight percentage of polymers (e.g., polymer 1, polymer 2, and / or polymer 3) in the compound, the weight percentage of external modifiers (e.g., external modifier 1 and / or external modifier 2) in the compound, the weight percentage of talc in the compound, the weight percentage of short glass in the compound, and / or the weight percentage of additives in the compound.

[0076] Table 1

[0077] Once system 100, device 200, or controller 302 receives data, system 100, device 200, or controller 302 can preprocess the data. For example, as shown in Table 2, system 100, device 200, or controller 302 can, for instance, filter the data. Other preprocessing steps can be performed as described above.

[0078] Table 2

[0079] Based on the filtered data, system 100, device 200, or controller 302 can apply such data to one or more machine learning models to generate synthesis data and / or prediction data and / or performance (or in other words, unmeasured data points) for various other formulations or compositions, as shown in Table 3 below.

[0080] Table 3

[0081] Based on synthetic or predicted data and / or performance, and other data in the design space, system 100, device 200, or controller 302 can provide suggested formulations or compositions and / or operating or process settings and / or parameters, such as Figure 4A The charts and / or examples shown in the text Figure 4B The table illustrated in .

[0082] In another example, design space data can be generated for extrusion operations. In such implementations, the aggregated operation data can include various extrusion process parameters and / or components, including but not limited to feed composition, recycled components, final output or product, filler, temperature in one or more zones of the extruder, pressure in one or more zones of the extruder, barrel torque, feed rate, ash content, product modulus or stiffness, and / or product impact. All these data points can be obtained for many historical extrusion operations. As noted, the data may be incomplete. In other words, many variables may not be represented by existing actual data, including the resulting outputs of those different variables. Therefore, the systems and methods described herein can be used to generate a design space for extrusion operations. For example, such systems and methods can utilize historical data to generate multiple inputs. These inputs can include variations in extrusion process parameters and / or components. The inputs can then be applied to a trained machine learning model to produce a synthetic output with predicted parameters or features. Each output can be added to and / or stored in the design space.

[0083] In the implementation, a portion of the data can be collected in real time. The systems and methods described herein can determine alternative process data based on real-time data and apply this alternative process data to a trained machine learning model to continue filling the design space. Furthermore, the systems and methods can proactively adjust the extrusion process using new entries entering the design space, for example, by adjusting fillers, additives, temperature, pressure, and / or other parameters.

[0084] Predictive Example: Virtual Experiment - Extruder Figure 5A An exemplary information flow 501 for a virtual experiment is illustrated. The exemplary system may include one or more extruders 503. "Feed 1" 505 may enter one or more extruders 503 at a first inlet and may be, for example, a polypropylene ("PP") recycled feed stream. "Feed 1" 505 may be provided by one or more suppliers 507 and may be associated with various feed data 509, including but not limited to quality data for post-consumer recycled ("PCR") materials. PCR quality data may include, but is not limited to, analytical certificates containing information such as (but not limited to) melt flow index, ash content, modulus (stiffness), and / or variability properties. In some embodiments, the PCR material may have high variability.

[0085] "Feed 2" component 511 can be fed online into extruder 503. "Feed 2" 511 may include, for example, virgin polypropylene, additives, fillers, and / or elastomers. Online process 513 can provide various online PCR quality data 515, including but not limited to viscosity (e.g., for manufacturer measurement purposes), ethylene content, and / or ash content. Other online processes 517 can provide various processing data 519, including but not limited to barrel torque, melt pressure, temperature (in some embodiments based on zones within extruder 503), "Feed 1" quantity, and / or "Feed 2" quantity. In various embodiments, additional feed can be added to extruder 503 as needed.

[0086] The “final product” 521 can leave the extruder 503. The offline process 523 can provide offline final product quality data 525, including but not limited to ash content, modulus (stiffness), impact strength, melt flow rate and / or quality / assurance tests.

[0087] A pre-trained machine learning model 527 can be used. Model 527 can be used to determine what to add to extruder 503 as "feed 2" composition 511 based on the unknown and / or variable mass of "feed 1" composition 505. In some embodiments, model 527 can provide input about the amount and / or type of the component added as "feed 2" composition 511 to produce a "final product" 521 with desired and / or known specifications from extruder 503. Known materials informatics tools can be used to obtain the model. However, in order to anticipate batch-to-batch variability in PCR materials, online measurement of PCR quality can be applied. This allows for real-time adjustment of the "feed 2" composition to ensure the desired performance of the "final product" is met.

[0088] Model training To train model 527, training data / input 529 can be generated based on the key identifier of PCR in "feed 1" 505.

[0089] The training data can be divided into sets of training data and validation data. The ratio of training data can vary from approximately 99:1 to approximately 1:99, but is preferably approximately 75:25. The training data can be used to train a machine learning model using a folding scheme (k-fold cross-validation) or a similar process. Validation data (or unseen data) may not be used for training and will be used to validate the model.

[0090] To train model 527, performance data 525 with offline measurements from available "final products" may be required (e.g., via experimental design). Utilizing analytical credentials from a PCR vendor can improve model performance. The exact features (i.e., the quantities measured) that can be used to obtain optimal model performance can be obtained by investigating feature importance and may depend on the measurement capabilities of the exact equipment used.

[0091] Model application Once the model is trained, the composition of "feed 2" 511 can be tuned using online measured characteristics (including but not limited to online PCR quality 515 and online data processing 519) and the desired final product performance.

[0092] Figure 5B This is a flowchart illustrating, according to certain embodiments of this disclosure, a design space for generating a design space comprising multiple chemical operations (such as polymerization operations), chemical compositions (such as polymer compositions), and corresponding properties of the resulting chemical products. Unless otherwise stated, the actions of method 500 may be performed within system 100, apparatus 200, and / or controller 302. Specifically, method 500 may be included in one or more programs, protocols, or instructions loaded into memory 106 of design space system 102 and executed on processor 104 or one or more processors. The order of the described operations is not intended to be construed as limiting, and any number of the described blocks may be combined in any order and / or in parallel to implement these methods.

[0093] At box 502, the design space system 102 can receive a dataset. At box 506, the design space system 102 can define input parameters for each or a portion of each variable in each chemical operation and / or chemical composition. In embodiments, the design space system 102 can determine the variables defining the input parameters based on multiple factors, such as the type of data received, the number of data points for a particular variable, and / or indicators or markers included in the data. Input parameters may include, for example, a minimum value, a maximum value, and / or an increment value for a variable. At box 508, the design space system 102 can iteratively apply the data to one or more machine learning models based on the input parameters. When applying the data to one or more machine learning models, the machine learning models can produce outputs, such as predictions, probabilities, and / or simulations, which exemplify or represent the performance or characteristics of the chemical products produced by the chemical operations and / or chemical compositions. This iterative application of data to machine learning models and the resulting outputs, along with the addition of other data to the design space, can form the design space (in some embodiments, in addition to measurement data and / or other synthetic data generated by one or more physics-based models). Boxes 506 and 508, encompassed by box 504, may include processes or subprocesses for generating the design space. Box 504 may include additional steps or boxes, and in some embodiments, for example, the design space system 102 may further preprocess the dataset before iteratively applying it to one or more machine learning models and / or the design space system 102 may perform post-processing on the output generated each time the data is applied to the machine learning model before adding the output to the design space.

[0094] Once the design space is generated, the design space system 102 can receive inputs at boxes 516 and / or 510. Specifically, at box 516, the design space system 102 can receive user constraint inputs. Such inputs may include the type of chemical product sought, the properties of such product, and / or the chemical operations and / or chemical composition used to produce such chemical product. In addition to or instead of user constraint inputs, at box 510, the design space system 102 can receive applied quality criticalities (CTQs) (in other words, the desired performance or other aspects of a chemical product and / or chemical process). At box 512, the design space system 102 and / or another user can convert the applied CTQs into material CTQs and / or process CTQs. If performed by the design space system 102, such conversions may include applying the applied CTQs to a model or other algorithm configured to generate one or more of material CTQs (in other words, materials that can meet the desired performance) and / or process CTQs (in other words, processes that can meet the desired performance) based on the desired performance specified in the input applied CTQs. Based on the receipt of applied CTQs and / or user input constraints, at box 518, the design space system 102 can generate multiple chemical compositions or formulations that satisfy or exceed user inputs (such as applied CTQs and / or user constraint inputs). Furthermore, the design space system 102 can generate one or more visualizations, such as, for example, the properties of the resulting chemical product and / or the chemical composition and / or chemical operations used to produce the chemical product.

[0095] Figure 6A and Figure 6B This is an additional flowchart, according to certain embodiments of the present disclosure, for generating a design space comprising multiple chemical operations (such as polymerization operations), chemical compositions (such as polymer compositions), and corresponding properties of the resulting chemical products. Unless otherwise stated, the actions of methods 600 and 601 can be performed within system 100, apparatus 200, and / or controller 302. Specifically, methods 600 and 601 can be included in one or more programs, protocols, or instructions loaded into memory 106 of design space system 102 and executed on processor 104 or one or more processors. The order of the described operations is not intended to be construed as limiting, and any number of the described blocks can be combined in any order and / or in parallel to implement these methods.

[0096] First, turning to method 600, at block 602, the design space system 102 can receive data. The data may include real-time chemical operation data and / or historical chemical operation data. In another embodiment, the data may include experimental data. At block 604, the design space system 102 may define input parameters for the received data, such as minimum, maximum, and / or incremental values ​​of one or more variables included in the received dataset. In embodiments, design space creation may be a continuous or sequential process. For example, the design space system 102 may receive data at different times and update the design space based on additional data received. Such received data may include chemical operations, chemical composition, and / or the resulting chemical products, and the corresponding properties of the chemical products.

[0097] At box 606, the design space system 102 can iteratively apply data to one or more machine learning models based on input parameters. As noted, in addition to the performance of the resulting chemical product, such application to one or more machine learning models can also produce chemical operations and / or chemical compositions. The output of such machine learning models can be a collection of values, probabilities, predictions, and / or simulation results. In addition to the received data, the design space system 102 can also utilize the output from the machine learning models to generate a design space.

[0098] At box 608, the design space system can determine whether input constraints and / or CTQs have been received. If no input constraints and / or CTQs have been received, the design space system 102 can wait until such inputs are received and / or until new data is received to populate the design space.

[0099] If input is received, at box 610, the design space system 102 can determine whether the input is unknown or whether the relevant results are included in the design space. If the input is unknown, the design space system 102 can determine whether the input includes sufficient additional data to be added to the design space (via application to one or more machine learning models). For example, if input including chemical composition and / or chemical parameters and resulting or potential performance data is received, the design space system 102 can determine that such data is sufficient (further depending on whether any data points are missing).

[0100] If the data is insufficient to add to the design space, at box 614, the design space system 102 may prompt another computing device and / or user to submit additional data. Once the data is received and / or deemed sufficient, the design space system 102 may define additional input parameters and iteratively apply the data to one or more machine learning models within the machine learning model.

[0101] If the input is known or includes relevant entries in the design space, the design space system 102 can determine one or more formulations or compositions and / or operating settings that meet or exceed the specified performance of the resulting chemical product. At box 618, the design space system 102 can generate one or more visualizations of each of the one or more formulations or compositions and / or operating settings. In another embodiment, at box 620, if a user selects a specific chemical composition and / or operating setting, and then, after implementing such a setting, transmits the results of those operations with the selected chemical composition (e.g., the performance of the resulting chemical compound), the design space system 102 can use such data to further refine one or more machine learning models among one or more machine learning models.

[0102] Go to Figure 6B Method 601 illustrates another process for filling a design space. In such an implementation, the design space system may receive data at block 602 and define the input parameters for the data at block 604, as described above and herein.

[0103] At block 622, the design space system 102 can then determine whether the maximum value of the variables in the operation settings has been reached. In one embodiment, this determination can be made for a combination of multiple variables and variations in the operation settings. If the maximum value has been reached, at block 632, the design space system 102 can determine whether additional or new data has been received, and if so, the process described for method 601 can be repeated; otherwise, the design space system 102 can wait for such data to further populate the design space.

[0104] If the maximum value of the variables set in the operation has not yet been reached, the design space system 102 can determine whether the maximum value of the amount of components in the chemical composition (such as polymer composition) has been reached. If the maximum value has been reached, at box 630, the design space system can increment one or more variables set in the operation and return to box 622. If the maximum value has not yet been reached, at box 626, the design space system 102 can generate a simulation of the chemical operation using the chemical composition and use the resulting performance data to populate the design space. Then, at box 628, the design space system 102 can generate a new chemical composition based on the input parameters (e.g., incrementing the amount of components in the previous chemical composition).

[0105] While specific terms and concepts are incorporated herein, the applicant emphasizes that the disclosed terms and concepts are for descriptive purposes only and therefore should not be construed or interpreted as limiting in any way. Certain embodiments and aspects of the disclosed systems, processes, and methods have been described in detail with particular reference to the illustrated embodiments. However, it will be apparent that many and various modifications and changes can be made to the spirit and scope of the embodiments of the systems, processes, and methods described herein, and such modifications and changes are considered equivalents and within the breadth and scope of this disclosure.

Claims

1. A method for generating a design space for determining one or more polymer compositions or polymerization operation settings, the method comprising: In response to receiving multiple historical polymerization operation data, input parameters are determined, the input parameters including at least the minimum, maximum, and increment of each of the multiple polymerization operation settings and each of the multiple polymer composition components, the multiple historical polymerization operation data including data indicating the associated performance of a portion of the multiple polymerization settings, a portion of the multiple polymer composition components, the output of one or more of the multiple historical polymerization operations, and the output of one or more of the multiple historical polymerization operations; The design space is generated by iteratively and incrementally applying the data and input parameters corresponding to the current iteration to one or more trained machine learning models. Each application generates a synthetic output of the polymerization operation and the performance of the synthetic output based on the input parameters of the current iteration. The design space includes data indicating the output or synthetic output and corresponding to one or more of multiple polymerization operation settings, multiple polymer component sets, or multiple associated performance characteristics. In response to a request for one or more of the polymer compositions or polymerization operation settings that meet or exceed specified performance, one or more of the plurality of polymer compositions or polymerization operation settings that meet or exceed the specified performance based on the design space are determined.

2. The method according to claim 1, further comprising: One or more visualizations are generated based on one or more of the plurality of polymer compositions or the plurality of polymerization operation settings that satisfy or exceed the specified performance based on the design space, and wherein the one or more visualizations include one or more interactive plots or tabular lists.

3. The method according to claim 1, further comprising: Before determining one or more of the plurality of polymer compositions or plurality of polymerization operation settings, it is determined whether the request includes an unknown polymer composition or an unknown polymerization operation setting; and In response to determining that the request includes an unknown polymer composition or an unknown polymerization operation setting: Generate prompts for additional data. In response to receiving the additional data, the trained machine learning model is retrained, and The design space is updated by applying data corresponding to the unknown polymer composition or unknown polymerization operation settings to the trained machine learning model.

4. The method according to claim 1, wherein the polymerization operation settings include one or more polymerization reactor settings, process settings, catalyst type, catalyst amount, temperature, pressure settings, flow rate, residence time, reactant concentration, solvent type, extrusion equipment settings, blending equipment settings, or solvent amount.

5. A method for generating a design space for determining one or more polymer compositions or polymerization operation settings, the method comprising: Receive data corresponding to multiple experimental data, real-time aggregated operation data, and historical aggregated operation data; The input parameters are determined based on the data, and the input parameters include at least the minimum, maximum and incremental values ​​of each of the multiple polymerization operation settings and each of the multiple polymer component settings. as well as While the maximum value set by the plurality of aggregation operations has not yet been reached, iteratively: (a) When the maximum value of the plurality of polymer components has not yet been reached, iteratively: (i) A simulation of the aggregation operation is generated based on the input parameters of the current iteration by applying the data and input parameters corresponding to the current iteration to one or more trained machine learning models. (ii) Determine new entries in the design space based on the simulation. (iii) Based on the input parameters, generate a new selection amount of the component of the selected polymer composition, and (iv) Repeat steps (i), (ii) and (iii) for the new selection amount of the components of the selected polymer composition. (b) Generate newly selected aggregation operation settings based on the input parameters, and (c) Repeat steps (a) and (b) for the newly selected aggregation operation.

6. The method according to claim 5, further comprising: The user interface is generated to include user input functionality, which allows the user to obtain (a) one or more of a plurality of polymerization operation settings or a plurality of polymer composition component amounts, and (b) the associated performance of each of the plurality of polymerization operation settings or the plurality of polymer composition component amounts. In response to the selection of one of the plurality of polymerization operation settings and one of the plurality of polymer component amounts, a polymerization operation is initiated based on the selection.

7. A system for generating a design space for determining one or more polymer compositions or polymerization operation settings, the system comprising: Communication circuit, the communication circuit being configured as follows: Receive data corresponding to one or more of an aggregation operation, an aggregation experiment, or a historical aggregation operation; a preprocessing circuit configured to: Determine the input parameters, which include at least each parameter in the data corresponding to the polymerization operation and the minimum, maximum, and increment of each quantity of multiple polymer components in the data; Modeling circuit, the modeling circuit being configured as follows: The design space is generated iteratively by incrementally applying the data and input parameters corresponding to the current iteration to one or more trained machine learning models. The design space includes data indicating the output or synthesis output, as well as data on one or more of a plurality of corresponding polymerization operation settings, a plurality of corresponding polymer compositions, or a plurality of corresponding associated properties. as well as A chemical operation controller, wherein the chemical operation controller is configured to: In response to receiving a request for one or more of the polymer composition or polymerization operation settings that meet or exceed the specified performance, determine one or more of the multiple variable polymer compositions or multiple polymerization operation settings that meet or exceed the specified performance based on the design space.

8. The system according to claim 7, wherein the system comprises: A visualization circuit, wherein the visualization circuit is configured to: One or more visualizations are generated based on one or more of the plurality of variable polymer compositions or the plurality of polymerization operation settings that satisfy or exceed the specified performance based on the design space; Display options to the user interface, thereby allowing selection of the one or more visualizations based on the type of each visualization in the one or more visualizations; as well as In response to the selection of one or more visualizations, the selected visualization is displayed on the user interface.

9. The system according to claim 7, wherein the system comprises: Design a spatial circuit, wherein the design spatial circuit is configured as follows: The outputs of the one or more trained machine learning models are processed to produce data formatted for the design space.

10. The system of claim 7, wherein the modeling circuitry is configured to retrain the one or more trained machine learning models in response to selection of one of the one or more polymerization operation settings of one or more variable polymer compositions.

11. The system of claim 7, wherein one of the one or more trained machine learning models includes an image-based machine learning model trained on a series of images of the polymer product, corresponding data, and acceptance or rejection of the polymer product, and wherein the request includes the visual appearance of the polymer product produced based on one or more of the polymer composition or the polymerization operation settings.

12. The system of claim 7, wherein the modeling circuit is configured to: In response to missing data in one or more of the plurality of polymer compositions, the plurality of associated properties, or the plurality of polymerization operation settings, the missing data is generated via a physics-based model and based on the data and the design space.

13. A controller for generating a design space for determining one or more polymer compositions or polymerization operation settings, the controller comprising: The controller is configured to provide inputs / outputs for signal communication with the aggregation equipment, and is configured to: Input parameters are determined in response to received experimental data, the input parameters including at least the minimum, maximum and increment of the process parameters in the experimental data; A design space is generated by iteratively and incrementally applying experimental data and input parameters corresponding to the current iteration to one or more trained machine learning models. The design space includes one or more of a plurality of polymer compositions, associated properties, or polymerization operation settings. In response to receiving a request for one or more polymer formulations or polymerization operation settings that meet or exceed specified performance, determine one or more of a plurality of variable polymer formulations or a plurality of polymerization operation settings that meet or exceed the specified performance based on the design space; and Displays one or more visualizations generated based on one or more of the plurality of variable polymer formulations or the plurality of polymerization operation settings that meet or exceed the specified performance based on the design space.

14. The controller of claim 13, wherein the one or more visualizations include one or more of graphical visualizations or tabular visualizations.

15. The controller of claim 13, wherein the request includes an application quality critical (CTQ) input, and wherein the controller is configured to convert the application CTQ input into material for processing the CTQ.

Images