Control system for a mixing production line

Abstract

The invention relates to an adaptive industrial control system (100) comprising at least one manufacturing line (110) which carries out one or more processes for manufacturing rubber products of which a predictive process is part, each manufacturing line (110) incorporating a mixing line (10) which makes it possible to produce rubber mixtures in different quantities and coming from a variety of rubber mixture recipes. The invention also relates to a production site for the rubber products, comprising the disclosed adaptive industrial control system (100). Figure for abstract: Fig. 2

Description

Title of the invention: Control system for a mixing production line Technical field

[0001] The invention relates generally to the production of rubber mixtures and rubber products prepared therefrom. The invention relates more particularly to an adaptive industrial control system which carries out one or more processes for manufacturing rubber products (including mixtures) in which the rubber products are systematically manufactured in accordance with specifications regardless of the variability of the incoming raw materials and the surrounding parameters. Context

[0002] In the field of tire manufacturing, a rubber compound may be produced from a variety of rubber blends, each having different ingredients that are mixed in different amounts and from a variety of recipes. Referring to [Fig.l], a compounding line (or "line") 10 is shown. It is capable of producing rubber compounds having diverse and varied properties as determined by the performance needs of a resulting rubber product(s) (e.g., desired characteristics of a tire(s). The compounding line 10 includes several systems for managing the input, identification, transfer, processing, storage, and output of raw materials. Thus, during a given compounding cycle, products having desired properties during the current manufacturing cycle are readily retrieved from the line 10.

[0003] The mixing line 10 shown comprises a raw material preparation zone (or “preparation zone”) 12 and a mixture production zone (or “production zone”) 13. The preparation zone 12 comprises one or more pieces of equipment for storing the raw materials necessary to produce the chosen mixture recipe and transporting them to the production zone 13.

[0004] The production zone 13 comprises one or more internal mixers MI into which various raw materials for producing the rubber product are introduced. The raw materials introduced to each internal mixer MI include, without limitation, one or more elastomeric materials (e.g., a natural rubber, a synthetic elastomer, and combinations and equivalents thereof) and one or more ingredients, such as one or more processing agents, protective agents, and reinforcing fillers. The raw materials may also include one or more other ingredients such as carbon black, silica, oils and resins. All raw materials are introduced to each MI internal mixer in varying quantities depending on the desired performance of the products obtained from the mixing processes (e.g., tires). An initial blend of elastomeric materials with a carbon black and / or silica filler is often carried out inside an MI internal mixer, where the temperature of the mixture rises (e.g., to values ​​between 130 °C and 180 °C). It is understood that at least one MI internal mixer could carry out the manufacture of a masterbatch (the qualities of a masterbatch are described below).

[0005] The production zone 13 also comprises at least one automated external mixer (also called a "roll mixer" or "feeder homogenizer") HA into which the mixture leaving the internal mixer MI is then transferred. Each HA roll tool circulates this mixture between two rolls so as to transform it into a continuous sheet. Thus, the HA roll mixer cools the produced mixture (for example, to approximately 80°C) to subsequently incorporate the vulcanizing blocks. Vulcanizing products (including, but not limited to, sulfur) may be added to the mixture later in the mixing cycle to obtain the final mixture for commercial use.

[0006] The production zone 13 also comprises at least one finishing homogenizer HF into which the cooled mixture leaving the HA roller mixer is then transferred (four finishing homogenizers HF1, HF2, HF3 and HF4 are shown in [Fig.l], but it is understood that a different number of finishing homogenizers HF could be incorporated in the mixing line 10). The cooled mixture leaving the HA roller mixer arrives at the finishing homogenizer HF, where the vulcanizers of a vulcanization system are added to the cooled mixture. It is known to add to the rubber mixture a product promoting its vulcanization during its subsequent curing. In order not to cause premature partial vulcanization of the rubber, the vulcanizing product is not incorporated at the same time as the other ingredients.Several HF finishing homogenizers could be used in preparation zone 12 in order not to degrade the industrial performance of line 10.

[0007] With machines in current production lines, there is currently no theoretical and applicable physicochemical model explaining all the transformations at work during the production of rubber products (including rubber products intended for incorporation into tires). For example, for the same type and grade of elastomer, the properties of this elastomer can vary from one supplier to another and even from one batch to another for a same supplier. Approaches are therefore largely empirical, and successive experimental designs have made it possible to simplify reality to keep only certain influential parameters of a current mixing process. These influential and controllable parameters (e.g., the chosen recipe of the mixture during manufacture) are then put under control in a statistical approach. Consequently, only deviations outside the tolerance of the standard are analyzed a posteriori.

[0008] This approach, termed static, assumes the reproducibility of the quality of the mixtures if the variability of the influencing factors is under control. However, in order to guarantee the quality of the mixtures produced, it is desired to reduce the variability of the incoming raw materials upstream (for example, the inter-batch variability from the same supplier or the variability between different suppliers). This is due to the fact that it can be difficult and costly to change raw material supplier: if the new raw material does not have the same properties, it may be necessary to replay part of the industrialization process of the mixture to find the new manufacturing recipe. It is also desired to analyze the impacts of hazards on the production lines by isolating the manufactured rubber products and by carrying out additional quality analyses.Following these analyses, the validated rubber products can be reintegrated into the flow (while preventing the integration of non-validated products). In addition, in some cases, it is desirable to adapt the recipe to environmental conditions by proposing different versions of the recipes for manufacturing the mixtures.

[0009] A "data-driven" approach thoroughly questions current classification systems by considering production systems based on real-time prediction of the properties of the mixture being manufactured (i.e., the mixture drop) by models obtained by machine learning.Machine learning actions have been carried out to predict the viscosity in real time (being the Mooney viscosity) with promising results (see, for example, publication CN114290554A which discloses a rubber mixing control system for an internal mixer, characterized in that the system comprises an internal mixer linked to a peripheral computer terminal having a data processing system for optimizing the process parameters of the mixture formula being manufactured) (see also Chinese patent CN108897283B which discloses a data analysis and processing method for a rubber production line, characterized in that it comprises an application server, an MES system, a cloud server, an intermediate database, a data cloud service PC, an auxiliary machine PC, an auxiliary machine on an internal mixer and a temperature collection controller).However, these are applications focused on measuring a material quality criterion and not an approach. across the entire mixing process. Although general algorithms for supervised learning are available, there are some limitations: the algorithms require a lot of training data, or this data comes from laboratory tests by sampling and does not correspond to the expected volume.

[0010] However, in recent years, the volume and nature of the data collected have grown considerably. In parallel, the maturity of data-based learning algorithms now makes it possible to exploit complex correlations in this data in order to build high-performance predictive and prescriptive systems. Systems incorporating artificial intelligence tools are based on these two fundamental elements.

[0011] Thus, the disclosed invention makes it possible to provide an intelligent data-based system for dynamically defining the manufacturing parameters of a mixing line in real time. Such a system makes it possible to best absorb the variabilities of the environment (including raw materials, manufacturing hazards and environmental conditions) by dynamically adjusting the manufacturing line. Indeed, using all of the information related to the manufacturing of the mixtures makes it possible to explain the mechanisms at work during the manufacturing of the mixtures. In particular, the links between the data characterizing the raw materials and the environment, the controllable process parameters and the material model are exploited. Summary of the invention

[0012] The invention relates to an adaptive industrial control system comprising at least one manufacturing line which carries out one or more processes for manufacturing rubber products of which a predictive process is part, each manufacturing line incorporating a mixing line which makes it possible to produce rubber mixtures in different quantities and originating from a variety of rubber mixture recipes, characterized in that the system comprises: - a detection system for collecting information on the physical environment around each manufacturing line, the detection system comprising one or more sensors along the mixing line; - at least one database in which the data obtained from each production line are stored to predict the achievement of the desired properties of a rubber product during production; - a communication network that manages incoming data to the system from at least a portion of at least one manufacturing line, the communication network comprising at least one communication server for executing ins programmed instructions stored in a memory of one or more processors of the system which uses a module for executing an optimization algorithm on data obtained from each manufacturing line to implement a predictive method for generating manufacturing recipes according to a current campaign, the parameters controlled by the optimization algorithm incorporate the signals from one or more sensors arranged along the mixing line, the optimization algorithm comprising: - a mixture quality prediction algorithm for predicting the intrinsic quality of a mixture during production; and - a generative prescription algorithm allowing the output of an identified rubber mixture recipe; such that each manufacturing line is configured on one or more parameters of the identified rubber mix recipe calculated by a data processing module obtained from each manufacturing line; and so that the calculated parameters are controlled by the optimization algorithm incorporating signals from one or more sensors of the mixing line in order to control the mixing line.

[0013] In certain embodiments of the system of the invention, the mixing line comprises: - a raw materials preparation area; and - a production area for mixtures including: - one or more internal mixers (IM) into which different raw materials for the production of the rubber product are introduced; - at least one automated external mixer (HA) into which a mixture coming out of the internal mixer (MI) is transferred to circulate it between two cylinders so as to transform it into a continuous sheet; and - at least one finishing homogenizer (HF) into which the cooled mixture leaving the HA cylinder mixer is transferred to add the vulcanizers of a vulcanization system; where the preparation area includes one or more pieces of equipment for storing the raw materials needed to produce the chosen mixing recipe and transporting them to the production area.

[0014] In some embodiments of the system of the invention, during a rubber product manufacturing process performed by the system, the data processing module employs one or more reinforcement learning means that determine the parameters of the identified rubber mix recipe.

[0015] In certain embodiments of the system of the invention, the reinforcement learning means or means are chosen from reinforcement learning means. multi-agent reinforcement.

[0016] In some embodiments of the system of the invention, the data obtained from each manufacturing line by means of the mixing line sensors include: - RAW MATERIAL CHARACTERIZATION (RMP) data representing raw material data, supplier tolerances and blocking rules obtained from the preparation area; - PROCESS data representing the automated data of the production line and its devices (MI, HA, HF), this data including temperatures, waiting times, mass differences and time signals from the signal automatons; - laboratory data including CONTROL data representing characteristics and surrogate quality measurements and QUALITY data representing properties and quality measurement; and - METEO data representing environmental data.

[0017] In some embodiments of the system of the invention, the CMP raw material data is obtained from one or more sensors comprising at least one capture device selected from temperature sensors, infrared sensors, ultrasonic sensors, pressure sensors, sensors employing data obtained by electronic noses and their equivalent devices.

[0018] In some embodiments of the system of the invention, during a rubber product manufacturing process performed by the system, the processor employing the optimization algorithm adopts a two-step predictive strategy comprising the following steps: - a step of building a predictive CONTROL model to predict CONTROL data from PROCESS data, CMP data and METEO data; and - a step of building a predictive QUALITY model to predict the quality of the mixtures during production on the production line; whose QUALITY predictive model uses the prediction coming out of the CONTROL model and the PROCESS data, the CMP data and the METEO data.

[0019] In certain embodiments of the system of the invention, during a process for manufacturing rubber products carried out by the system, the data processing module of the system employs a reinforcement optimization brick which uses the two CONTROL and QUALITY models to train a learning agent making it possible to prescribe the manufacturing recipe of the identified mixture; so that the development of prescriptive algorithms makes it possible to provide controllable parameter values ​​for the production line.

[0020] In some embodiments of the system of the invention, during a rubber product manufacturing process performed by the system, the system updates incoming data from the database to assist in sharing information about physicochemical properties and other physical phenomena of the rubber mixtures at a production site incorporating the system.

[0021] The invention also relates to a production site for rubber products, comprising the disclosed system.

[0022] Other aspects of the invention will become apparent from the following detailed description. Brief description of the drawings

[0023] The nature and various advantages of the invention will become more apparent upon reading the following detailed description, in conjunction with the accompanying drawings, in which like reference numerals designate like parts throughout, and in which: [Fig.l] [Fig.l] represents a schematic view of a known embodiment of a mixing line which is part of a manufacturing line which carries out one or more processes for manufacturing rubber products. [Fig.2] [Fig.2] represents an adaptive industrial control system of the invention which carries out a predictive process making it possible to generate rubber manufacturing recipes according to a current campaign. [Fig.3] [Fig.3] represents data obtained from a manufacturing line during a rubber product manufacturing process carried out by the adaptive industrial control system of [Fig.2]. [Fig.4] [Fig.4] represents an embodiment of a predictive model constructed by the adaptive industrial control system of the invention to predict the Composition Control (CONTROLE) data from the PROCESS data, the Raw Material Characterization (CMP) data and the METEO data. [Fig.5] [Fig.5] represents an embodiment of a predictive model constructed by the adaptive industrial control system of the invention to predict the quality of mixtures during production. [Fig.6] [Fig.6] represents a flow diagram of one embodiment of a reinforcement optimization process using the two models represented respectively in Figures 4 and 5 to prescribe a manufacturing recipe for the identified mixture. [Fig.7] [Fig.7] represents the set of chemical groups obtained from a heated rubber in an example of the data obtained by the adaptive industrial control system of the invention. [Fig.8] [Fig.8] represents a distribution of Shore hardness property measurements from the choice of mixing line settings of [Fig.l] in an example of the data obtained by the adaptive industrial control system of the invention. Detailed description

[0024] Referring now to the Figures, in which like numbers identify like elements, [Fig. 2] shows an adaptive industrial control system (or "adaptive system" or "system") 100 that performs a predictive process for generating rubber manufacturing recipes (or "manufacturing recipes" or "recipes") according to a current campaign (or a potential campaign) in order to guarantee the realization of future planned campaigns. The word "campaign" refers to the duration of operation of a line for manufacturing a mixture according to a chosen recipe. By using the disclosed invention, a known product will be delivered with homogeneous and stable characteristics.

[0025] Future manufacturing recipes are generated in anticipation of planned future campaigns in which predetermined products are required, being in specified quantities during a specified period. The specified period depends on the fulfillment of the current manufacturing recipe and future manufacturing recipes. To satisfy a current campaign, the system of the invention generates one or more manufacturing recipes. In some embodiments, potential recipes may be identified, for example, by satisfying the current requirements of the current manufacturing recipe, and then generating various manufacturing recipes that will meet the needs of one or more subsequent campaigns based on the specified time frames for the execution of those campaigns.

[0026] The system 100 of the invention determines the feasibility of a mixture associated with the use of a manufacturing recipe to carry out a current campaign. "Mixture feasibility" refers to the ability to comply with the technical specifications of the raw materials used in the mixture. "Mixture feasibility" also refers to the ability to achieve a final technical result for the mixture (i.e., the products used retain their own technical characteristics while contributing to the desired characteristics of the mixture in which they are integrated). In some embodiments, the manufacturing recipes may be ranked using self-learning approaches in which some production projects are more desirable than others. The system may then rank the different manufacturing recipes based on the feasibility of the associated mixture.It is contemplated that the system 100 of the invention is capable of producing rubber blends having diverse and varied properties as determined by the performance needs of a resulting rubber product(s) (e.g., characteristics. desired characteristics of one or more tires).

[0027] The system 100 comprises at least one rubber product manufacturing line (or "line") 110 that performs one or more processes for manufacturing rubber products. The line 110 is usable in facilities where rubber products are manufactured (including tires). The line 110 may comprise a mixing line of the type represented in [Fig.l] by the mixing line 10. It is understood that the line 110 could operate in several physical environments without knowledge of their parameters in advance (for example, a facility in which the line 110 is part).

[0028] Referring again to Figures 1 and 2, and further to [Fig. 3], the system 100 comprises at least one communication network (or "network") that manages incoming data to the system from various sources (e.g., from at least a portion of the manufacturing line 110 and its associated detection system). The communication network incorporates one or more communication servers (or "servers") each comprising one or more processors operatively connected to a memory. The memory is configured to store a module for executing an algorithm for optimizing the data obtained from the line 110 (or "optimization algorithm") (such data being shown in and described with respect to [Fig. 3]).The optimization algorithm enables continuous improvement across the entire line 110, ensuring that the system 100 improves with the experience it gains, particularly in following instructions to achieve the desired properties of the rubber product being produced.

[0029] The term "processor" (or, alternatively, the term "programmable logic circuit") means one or more devices capable of processing and analyzing data and including one or more software programs for their processing (e.g., one or more integrated circuits known to those skilled in the art as being included in a computer, one or more controllers, one or more microcontrollers, one or more microcomputers, one or more programmable logic controllers (or "PLCs"), one or more application-specific integrated circuits, one or more neural networks, and / or one or more other known equivalent programmable circuits).The processor includes one or more software programs for processing the data captured by the manufacturing line detection system 110 (and the corresponding data obtained) as well as one or more software programs for identifying and locating variances and identifying their sources to correct them.

[0030] In the system 100 of the invention, the memory may include both volatile and non-volatile memory devices. The non-volatile memory may include solid-state memories, such as NAND flash memory, keep-alive memory (KAM) for saving data. various operating variables while the processor is powered off, magnetic and optical storage media, or any other suitable data storage device that retains data (e.g., when the mixing line detection system is disabled or loses power). Volatile memory may include static and dynamic RAM that stores program instructions and data, including a learning application.

[0031] In embodiments of an optimization method performed by the system 100, the processor may configure the system (and in particular line 110) on one or more parameters of a target rubber mixture recipe calculated by a data processing module. In these embodiments, it is understood that one or more means of reinforcement learning (or "reinforcement learning") could be employed (including deep reinforcement learning). For example, one or more means of multi-agent reinforcement learning (or "MARL") could be employed which is based on a Markov decision process (or "MDP") framework. Those skilled in the art in this field will recognize that numerous data processing techniques can be used to choose and determine the parameters of the rubber mixtures.Several commercially available data processing systems can be used.

[0032] The processor may also refer to a reference (e.g., a table of rubber compound recipes and their various properties) to make a final determination of a parameter or parameters of an identified recipe. The reference may include known recipe parameters corresponding to a plurality of known commercially available rubber products (e.g., tires). For example, after the data processing module calculates one or more identified recipe parameters, the processor may compare the calculated parameters with the known parameters recorded in the reference.

[0033] The system 100 implements an optimization algorithm that employs two learning algorithms: - A compound quality prediction algorithm that is capable of predicting the intrinsic quality of a compound being produced based on all available upstream data (including, without limitation, the quality of incoming raw materials, sensor values ​​on the manufacturing line, environmental conditions at the time of manufacturing, and end-of-line rheological characterization). The prediction algorithm is usable as a simulator that allows understanding the desired compound properties (e.g., descriptors expected by a tire designer); and - A generative prescription algorithm that returns an optimal rubber compound manufacturing recipe to be applied in a given context. This manufacturing recipe can be completely new (i.e., never proposed in the past) to adapt to a unique variability combination.

[0034] The processor may retrieve the corresponding known recipe parameters that most closely match the parameters of the mixture being manufactured to configure the manufacturing line 110. Referring again to [Fig. 3], the processor relies on the use of all data collected from the production line 110 as well as its production environment, including, without limitation, the following data: - Raw Material Characterization (CMP) data representing raw material data, supplier tolerances and blocking rules obtained from preparation area 12; - The PROCEDE data representing the automated data of the production line 110 and its MI, HA and HF devices, this data including, without limitation, the temperatures, the waiting times, the mass deviations and the time signals of the automated devices (for example, the time signals such as the power curves, the rotation speed, etc.); - Laboratory data including Control Composition CONTROL data (representing characteristic curves and points and surrogate quality measurements) and QUALITY data (representing low volume cooked properties and quality measurement); and - METEO data representing environmental data (including, without limitation, humidity level, ambient air pressure, ambient air temperature, etc.).

[0035] The PROCESS data is obtained from one or more sensors known to generate or capture data corresponding to the operation of the production line 110, including at least one capturing device selected from temperature sensors (including sensors of the temperature in an internal mixer MI), infrared sensors, ultrasonic sensors, pressure sensors, sensors employing data obtained by electronic noses, and / or other equivalent devices. The sensors could further comprise a combination of sensors for collecting data concerning one or more aspects of the dynamic situation of a portion of the line 110 (for example, one or more acceleration and / or speed sensors of one or more belts of the mixing line) (see [Fig. 2]).The sensors could include one or more virtual sensors (or "soft sensors") that could replace the hardware sensors (or "hardware sensors") or work in parallel with them.

[0036] The sensors may store this data in at least one database 112 of the system 100 to perform the provisioning of one or more software programs incorporating the optimization algorithm (see [Fig.2]). The system 100 updates the incoming data to assist in sharing information with respect to the mixing line and the environment in which it is installed. The system 100 implements a predictive method by using "stored" measurement data and the "updated" data to prepare customized reminders to be sent to the mixing line (or to a facility incorporating the mixing line) and to predict the achievement of desired properties of a rubber product being produced. The data is retrieved and processed in order to extract only characteristic points from the time series from the sensors of the mixing line.Thus, the system 100 may provide adjustments to production cycle parameters (including adjustments to the mixing cycles performed by the mixing line) to improve the rubber product exiting the mixing line.

[0037] Referring again to Figures 1 to 3, and further to Figures 4 and 5, the processor of the system 100 employing the optimization algorithm adopts a two-step predictive strategy to prescribe the manufacturing recipes. First, a predictive model (a "CONTROL model") is constructed to predict the CONTROL data from the PROCESS data, the CMP data and the WEATHER data (see [Fig. 4]). Next, a predictive model (a "QUALITY model") is constructed to predict the quality of the mixtures being produced (see [Fig. 5]). The QUALITY model uses the prediction output from the CONTROL model and the PROCESS data, the CMP data and the WEATHER data. During the training of both the CONTROL and QUALITY models, one or more known machine learning algorithms could be employed as understood by those skilled in the art.In embodiments of the system 100, one or more neural networks or decision trees for predicting QUALITY data. These algorithms may be based on multi-label, multi-mixing line and / or multi-production site data. Other known solutions may be used to compensate for a lack of QUALITY label such as transfer learning or the use of intermediate models based on CONTROL data.

[0038] It is understood that the order of construction of the two CONTROL and QUALITY models is not limited to the order shown. Thus, the QUALITY model could be constructed upstream of the construction of the CONTROL model, or the two models could be constructed substantially simultaneously. The two CONTROL and QUALITY models could be constructed alongside the construction of one or more other models to properly achieve the performance parameters of the rubber product being produced.

[0039] Referring now to [Fig.6], a reinforcement optimization brick uses the two COMPOSITION CONTROL and QUALITY models to train a learning agent for prescribing the recipe for manufacturing the mixture. Once collected and aggregated, this data of different forms (for example, time signals and tabular data) is assembled into a data set allowing the development of prescriptive algorithms for providing the controllable parameter values ​​of the production line 110. In one embodiment, the prescriptive algorithms for providing the controllable parameter values ​​of the production line typically comprise optimization algorithms of the reinforcement learning and / or evolutionary genetic algorithm type.These algorithms use previous predictive models and their training is on adapted hardware architectures.

[0040] The field of manufacturing rubber products (including tires) from rubber mixtures involves processes comprising multiple steps of processing raw materials, including, without limitation, grinding and washing steps, hot convection steps for drying wet rubber crumb, packaging of dried rubber, and storage of manufactured rubber. Thus, the system 100 is capable of preventing fallout of mixtures that are initially out of tolerance. The system 100 can make necessary adjustments in the mixing line that ensure compliance with tolerances during subsequent mixing cycles.

[0041] For example, the system 100 may adjust the operation of the mixing line 110 in response to the emissions of volatile organic compounds (or "VOCs") from different rubbers that exhibit a number of odorants in terms of concentrations as well as odorant type. Referring to [Fig. 7], this figure represents the set of chemical groups obtained from a heated rubber. The variation in the identified odors may be associated with specific properties of the rubber, in particular protein levels and moisture content ([Fig. 7] corresponds to [Fig. 8] of the reference "Quantification of VOCs and the Development of Odor Wheels for Rubber Processing", Nor H. Kamarulzaman, Nhat Le-Minh, Ruth M. Fisher, Richard M. Stuetz, Science of the Total Environment 657 (2019))("the Kamarulzaman reference") (see also Table 3 of the Kamarulzaman reference)).The dominant group of VOCs released from the heated sample were aromatic compounds, followed by ketones, aldehydes, acids, and cyclic-type compounds. Sulfur and nitrogen groups had a low and similar percentage of odorants, however, they were likely to contribute to the odor profile. Thus, the system 100 can employ the obtained prediction models to establish a . link between odors and odorous substances. The system 100 can therefore take advantage of the impact of odors in the manufacture of rubber products and develop efficient production and management practices at a rubber product production site (for example, by using data obtained by electronic noses to take advantage of known physicochemical properties in the field of rubber product production (see application FR2112099 and application FR2302251 of the Applicant).

[0042] In another example, the system 100 may adjust the mixing line based on other physical phenomena being studied. Referring to [Fig. 8], this figure represents a distribution of Shore hardness property measurements. Bars 8A, 8B define the lower and upper tolerances, respectively. The dotted bar 8C indicates the values ​​to be targeted. It is known that dam quality data is directly impacted by the choice of mixing line settings. In order to select the most relevant algorithm, performance tests were performed using various implementations (e.g., performance tests performed in Python).In one embodiment, it has been found that a "Covariance Matrix Adaptation for Evolution Strategy" (or "covariance matrix adaptation for evolution strategy" or "CMA-ES") algorithm has advantageous performance in terms of the quality of the solutions found. Indeed, all of the results are found at the end of the optimization within the tolerances. Furthermore, certain adjustment proposals make it possible to get significantly closer to the target values.

[0043] The system 100 updates incoming data (see, for example, database 112) to assist in sharing information about the physicochemical properties and other physical phenomena of rubber mixtures at a production site incorporating the system 100. The system 100 implements the rubber product manufacturing processes using "stored" data and the "updated" data to prepare customized reminders to be sent to the mixing line and to predict achievement of desired properties of the mixture being produced. Thus, the system 100 can propose adjustments to parameters of the mixing cycle to improve the rubber mixture exiting the mixing line.

[0044] The parameters controlled by the optimization algorithm incorporate the signals from the sensors of the mixing line. The actual degrees of freedom of the mixing line include the instructions of the automatons which induce these signals. Therefore, a signal to instruction inversion system is necessary in order to control the mixing line 110 by the prescriptions.

[0045] The system 100 of the invention can easily repeat one or more steps of a process of manufacturing rubber products in an order to properly achieve the performance parameters of the rubber product being produced.

[0046] In advance of launching the predictive method which is part of a rubber product manufacturing process, orders for the rubber products could be received by known means (for example, by one or more communication networks integrated into a rubber product production site incorporating the system 100). The site allows the orders to be received up to a specific time to ensure the achievement of the desired properties of the ordered rubber products in a "just in time" manner (where the predictive method of a rubber product manufacturing process starts).

[0047] The rubber product manufacturing processes performed by the system 100 may be done by the control of the PLC and may include preprogramming of the production information of a rubber product. For example, a setting of the manufacturing process may be associated with the parameters of an identified rubber mixture recipe so that each manufacturing line 110 is configured on one or more parameters calculated by a data processing module obtained by the system 100 from each manufacturing line. The system 100 (and / or a rubber product production site incorporating the system 100) may easily repeat one or more steps of a rubber product manufacturing process in a determined order to properly provide ordered rubber products.

[0048] A setting of a rubber product manufacturing process performed by the system 100 may be associated with the parameters of the typical physical environments in which the system 100 operates. In embodiments of the invention, the system 100 (and / or a rubber product production site incorporating the system 100) may receive voice commands or other audio data representing, for example, a step or a stop of the mixing line. The request may include a request for the current status of a rubber product production cycle. A generated response may be represented audibly, visually, tactilely (for example, using a haptic interface), and / or virtually and / or augmentedly. This response, associated with the corresponding data, may be recorded in one or more neural networks.

[0049] For all embodiments of the system 100, a monitoring system could be implemented. At least a portion of the monitoring system may be provided in a portable device such as a mobile network device (e.g., a mobile phone, a laptop, one or more network-connected wearable devices (including “augmented reality” and / or “virtual reality” devices), network-connected wearables, and / or any combinations and / or equivalents). It is conceivable that detection and comparison steps can be carried out iteratively.

[0050] The terms "at least one" and "one or more" are used interchangeably. Ranges that are presented as being "between a and b" encompass the values ​​"a" and "b".

[0051] Although particular embodiments of the disclosed system have been illustrated and described, it will be understood that various changes, additions, and modifications may be practiced without departing from the spirit and scope of the present disclosure. Accordingly, no limitations should be imposed on the scope of the disclosed invention except those set forth in the appended claims.

Claims

1. Claims An adaptive industrial control system (100) comprising at least one manufacturing line (110) which carries out one or more processes for manufacturing rubber products of which a predictive process is part, each manufacturing line (110) incorporating a mixing line (10) which makes it possible to produce rubber mixtures in different quantities and coming from a variety of rubber mixture recipes, characterized in that the system (100) comprises: - a detection system for collecting information on the physical environment around each manufacturing line (110), the detection system comprising one or more sensors along the mixing line (10); - at least one database (112) in which the data obtained from each manufacturing line (110) are stored to predict the achievement of the desired properties of a rubber product during production; - a communication network that manages the data entering the system (100) from at least a part of at least one manufacturing line (110), the communication network comprising at least one communication server for executing programmed instructions stored in a memory of one or more processors of the system (100) which employs a module for executing an optimization algorithm on data obtained from each manufacturing line (110) to implement a predictive method for generating manufacturing recipes according to a current campaign, the parameters controlled by the optimization algorithm incorporate the signals from one or more sensors arranged along the mixing line (10), the optimization algorithm comprising: - a mixture quality prediction algorithm for predicting the intrinsic quality of a mixture during production; and - a generative prescription algorithm for outputting an identified rubber mixture recipe; such that each manufacturing line (110) is configured on one or more parameters of the identified rubber mixture recipe calculated by a data processing module obtained from each manufacturing line (110); and so that the calculated parameters are controlled by the algorithm optimization incorporating signals from one or more sensors of the mixing line in order to control the mixing line (10).

2. The system (100) of claim 1, in the mixing line comprising: - a preparation zone (12) for the raw materials; and - a production zone (13) for the mixtures comprising: - one or more internal mixers (MI) into which different raw materials for producing the rubber product are introduced; - at least one automated external mixer (HA) into which a mixture leaving the internal mixer (MI) is transferred to circulate it between two cylinders so as to transform it into a continuous sheet; and - at least one finishing homogenizer (HF) into which the cooled mixture leaving the cylinder mixer HA is transferred to add the vulcanizers of a vulcanization system; where the preparation zone (12) comprises one or more pieces of equipment for storing the raw materials necessary to produce the chosen mixture recipe and conveying them to the production zone (13).

3. The system (100) of claim 1 or claim 2, wherein, during a rubber product manufacturing process performed by the system (100), the data processing module employs one or more reinforcement learning means that determine parameters of the identified rubber mix recipe.

4. The system (100) of claim 3, wherein the one or more reinforcement learning means are selected from multi-agent reinforcement learning means.

5. The system (100) of any one of claims 1 to 4, wherein the data obtained from each manufacturing line (110) by means of the sensors of the mixing line comprise: - RAW MATERIAL CHARACTERIZATION (RMP) data representing the raw material data, supplier tolerances and blocking rules obtained from the preparation zone (12); - PROCESS data representing the automated data of the production line (110) and its devices (MI, HA, HF), these data comprising the temperatures, waiting times, mass deviations and time signals of the signal automatons; - laboratory data including CONTROL data representing characteristics and surrogate quality measurements and QUALITY data representing properties and quality measurement; and - METEO data representing environmental data.

6. The system (100) of claim 5, wherein the CMP raw material data is obtained from one or more sensors comprising at least one capturing device selected from temperature sensors, infrared sensors, ultrasonic sensors, pressure sensors, sensors employing data obtained by electronic noses and their equivalent devices.

7. The system (100) of claim 5 or claim 6, wherein, during a process for manufacturing rubber products performed by the system (100), the processor employing the optimization algorithm adopts a two-step predictive strategy comprising the following steps: - a step of building a CONTROL predictive model for predicting CONTROL data from PROCESS data, CMP data and WEATHER data; and - a step of building a QUALITY predictive model for predicting the quality of the mixtures being produced on the manufacturing line (110); of which the QUALITY predictive model uses the prediction coming out of the CONTROL model and the PROCESS data, CMP data and WEATHER data.

8. The system (100) of claim 7, wherein, during a rubber product manufacturing process performed by the system (100), the data processing module of the system employs a reinforcement optimization brick that uses the two CONTROL and QUALITY models to train a learning agent for prescribing the identified recipe for manufacturing the mixture; such that the development of prescriptive algorithms makes it possible to provide the controllable parameter values of the production line (110).

9. The system (100) of any one of claims 1 to 8, wherein, during a rubber product manufacturing process performed by the system (100), the system (100) updates incoming data from the database (112) to assist in sharing the information on physicochemical properties and other physical phenomena of rubber mixtures at a production site incorporating the system (100).

10. A rubber product production site, comprising the system (100) of any one of claims 1 to 9.

Images