METHOD AND SYSTEM FOR CONTROLLING THE PRODUCTION OF RUBBER PRODUCTS IN RESPONSE TO THE PHYSICAL-CHEMICAL PROPERTIES OF A RUBBER COMPOUND

Description

Technical Field

[0001] The invention relates to the use of essentially "olfactory" measurements in facilities where rubber products (including tires) are manufactured from rubber compounds. More specifically, the invention relates to systems and methods for determining correlations between one or more physical phenomena related to a rubber (for example, stickiness) and the evolution of volatile organic compounds from rubber compounds. Context

[0002] The manufacturing of rubber products (including tires) from rubber compounds involves processes comprising multiple raw material transformation steps, including, but not limited to, grinding and washing, hot convection drying of wet rubber crumbs, packaging of dried rubber, and storage of manufactured rubber. Emissions of volatile organic compounds (VOCs) from different rubbers expose a range of odorants in terms of both concentration and type. During these processes, VOC monitoring presents an important opportunity to assess changes in the composition of VOCs emitted during rubber processing.

[0003] In a study comparing VOCs emitted directly from raw rubber materials or unprocessed rubber, the identified compounds were classified into different chemical groups (e.g., acids, alcohols, aromatic compounds, aldehydes, alkanes, ethers, esters, cyclic hydrocarbons, ketones, sulfur compounds, nitrogen compounds, terpenes, and others) (see "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)) (Kamarulzaman reference). In this study, some of the odorants were subjected to comparative VOC analysis to identify general trends in rubber emission behavior. According to this study, variation in the identified odors can be associated with specific rubber properties, particularly protein levels and moisture content. For example, the Figure 1The table corresponding to Figure 8 in the Kamarulzaman reference represents the set of chemical groups obtained from heated rubber (see also Table 3 in the Kamarulzaman reference). The dominant group of VOCs released by the heated sample were aromatic compounds, followed by ketones, aldehydes, acids, and cyclic compounds. Sulfur and nitrogen groups exhibited a low and similar percentage of odorants; however, they were likely to contribute to the olfactory profile.

[0004] Commercially available tools exist for detecting the presence of a target compound (e.g., a chemical or biological analyte) in a gaseous sample. Among these tools, electronic noses often employ one or more sensors that measure the concentration of a substance by surface plasmon resonance (SPR) (see "Highly-Selective Optoelectronic Nose Based on Surface Plasmon Resonance Imaging for Sensing Volatile Organic Compounds," Sophie Brenet, Aurelian John-Herpin, François-Xavier Gallat, Benjamin Musnier, Arnaud Buhot, Cyril Herrier, Tristan Rousselle, Thierry Livache, and Yanxia Hou, Analytical Chemistry, 90, 9879-9887 (2018). The SPR technique can measure the interaction between molecules using optical principles without a separate labeling substance such as a fluorescent material.Surface plasmons are quantized vibrations of free electrons propagating along the surface of a conductor, such as a metallic surface. These surface plasmons pass through a dielectric medium, such as a prism, and enter a metallic film at an angle greater than the critical angle of the dielectric medium. They are excited by the incident light and induce resonance at a specific angle. The angle of incidence at which this resonance occurs (called the "resonance angle") is sensitive to variations in the refractive index of a material close to the metallic film.

[0005] SPR (Selective Proton Pumping) is a well-established technique for detecting local changes in the optical index (refractive index) that characterize the interaction of the target compound with each sensor on the electronic nose. SPR sensors can quantitatively analyze samples based on the change in the refractive index of the material near the metal film (i.e., a sample exhibiting the properties described above). For VOC analysis, the efficiency of SPR contributes to the high performance of electronic noses, such as their sensitivity, selectivity, repeatability, and stability. SPR imaging solutions for gaseous VOC detection also exist that offer good repeatability and stability (see the solutions disclosed in publications FR3063543, WO2021 / 009440, WO2021 / 053284, and WO2021 / 053285, and commercially available from Aryballe Technologies).

[0006] Patent document FR 3 091 346 A1 describes a method for characterizing target compounds present in a fluid sample by an electronic nose using surface plasmon resonance imaging technology.

[0007] Furthermore, recent improvements in machine learning techniques and data analysis, combined with data computing and storage platforms, have opened avenues for developing new approaches to link olfactory profiles and solutions using the SPR effect. In the field of artificial intelligence (or "AI"), machine learning techniques are well-established, and their foundation lies in being "trained" on a large number of situations. By adjusting the weighting coefficients during a learning phase, machine learning can predict the outcome of a new situation.It is understood that several distinct learning methods are possible, including supervised learning (in which the algorithm trains on a set of labeled data and modifies itself until it is able to achieve the desired result), unsupervised or semi-supervised learning (in which the data is not labeled so that the network can adapt to increase the accuracy of the algorithm), reinforcement learning (in which the algorithm is reinforced for positive results and penalized for negative results) and incremental learning (the algorithm asks for examples and labels as it goes along to refine its prediction) (see https: / / www.lebigdata.fr / reseau-de-neurones-artificiels-definition).

[0008] When attempting to evaluate the parameters of a rubber product, some are inaccessible through measurement. For example, aging is currently managed using empirical expiration dates. Regarding stickiness, workshops primarily use ball tackmeters to estimate a product's stickiness, which requires sampling. In both cases, the measurement is subject to operator error in addition to other measurement errors.

[0009] Thus, establishing a link between odors and odor-causing substances can enhance the impact of odors on the manufacture of rubber products and facilitate the development of effective production and management practices at a rubber product manufacturing site. The disclosed invention therefore combines SPR data, obtained by electronic noses, with rubber product production facilities to leverage known physicochemical properties in the field of rubber product production (e.g., characteristic odors influencing the stickiness of mixtures in industrial processes) (as used here, the terms "mixture" and "rubber mixture" are interchangeable). Using an artificial intelligence-based system, the data obtained makes it possible to establish a non-obvious link between the SPR physicochemical aspect and the olfactory aspect of the rubber product(s) being produced. Summary of the invention

[0010] The invention relates to a manufacturing control method for rubber products carried out by a rubber product manufacturing system comprising a production installation having at least one mixing means which implements successive mixing steps of a rubber mixture, characterized in that the method comprises the following steps: a step of starting a mixing cycle carried out at the production facility; a step of capturing ambient air around the production facility to obtain at least one gaseous sample during the mixing cycle, this step being carried out by an odor detection device of the system which recognizes the presence of odorous volatile organic compounds present in the ambient air received in the device by measuring its concentration by surface plasmon resonance; a step of detecting odor profiles present in the ambient air and captured by the device; a step of identifying an odor profile to identify a mixture exiting the mixing means and its production state, this step being carried out from at least one sample of the ambient air captured by the device;a step of analyzing the samples captured by the device from one or more SPR fingerprints provided by the device and the extracted data represented therein; a step of constructing at least one olfactory profile model from the physicochemical properties of the identified rubbery mixture, this step including a step of introducing the physicochemical properties of the identified rubbery mixture to a neural network, this step being carried out by the system; and a step of training the olfactory profile model from the olfactory profiles of the identified rubbery mixture; so that the outgoing olfactory profile model will be the identification of the physico-chemical properties of the rubbery mixture being produced at the production facility.

[0011] In embodiments of the method of the invention, the step of constructing the olfactory profile model includes a step of creating a reference of the olfactory profiles sought in the ambient air to be captured by the device.

[0012] In embodiments of the process of the invention, the introduction step of the olfactory profile model construction step includes a step of creating a training base of physico-chemical properties which is introduced into the olfactory profile model.

[0013] In embodiments of the process of the invention, the process further includes a comparison step during which the identification of the physico-chemical properties of the mixture exiting the mixing means, derived from the olfactory profile model, is compared to the olfactory profile of the identified mixture, so that the system can adjust the production installation where the desired properties of the mixture are not achieved.

[0014] In some embodiments of the process of the invention, the process further includes a step for adjusting the production equipment to ensure that the desired properties of the mixture exiting the mixing means are achieved. In such embodiments of the process of the invention, the step for adjusting the production equipment includes a step for determining the stickiness level of the mixture from an SPR imprint provided by the device.

[0015] In embodiments of the process of the invention, during the training step of the olfactory profile model, the system employs a learning method chosen between a machine learning method and a progressive learning method.

[0016] In embodiments of the process of the invention: The odor profile detection step includes a step of removing VOC pollution from the ambient air received by the device, this step being carried out by a filter of the system before the ambient air enters the device; the odor profile identification step includes a step of reducing the humidity and / or temperature in the ambient air received by the device, this step being carried out by a condenser of the system; and the step of analyzing the samples captured by the device further includes a step of analyzing odor profiles in the samples before the analysis carried out by the device, this step being carried out by a multiplexer of the system.

[0017] In embodiments of the process of the invention, the start-up step of a mixing cycle includes a step of introducing, into the mixing means, the raw materials necessary for the production of the mixture.

[0018] In embodiments of the process of the invention, the start-up step of a mixing cycle includes a step of introducing one or more masterbatches into the mixing means.

[0019] In embodiments of the process of the invention, the process further includes a simulation step of a number of mixtures enabling the production of at least one rubbery mixture having predetermined physico-chemical properties.

[0020] The invention also relates to a system for manufacturing rubber products that carries out the disclosed manufacturing control processes for rubber products, characterized in that the system comprises: a production installation which implements successive mixing steps, the production installation comprising at least one mixing means from which one or more rubbery mixtures emerge; an odor detection device which captures ambient air around the production installation to obtain at least one gaseous sample during the mixing cycle; and a control subsystem which employs an odor profile model from the physico-chemical characteristics of the rubbery mixtures recognized by surface plasmon resonance (SPR); so that the device recognizes the presence of odorous volatile organic compounds (VOCs) present in the captured ambient air by measuring its concentration by surface plasmon resonance (SPR);and so that the olfactory profile model learns the physicochemical properties of VOCs associated with the mixtures exiting the mixing unit during the rubber product production cycles carried out by the production facility.

[0021] In embodiments of the system of the invention, the subsystem comprises one or more sensors that are triggered when the output olfactory profile model indicates a discrepancy between the physicochemical properties of the mixture being produced at the production facility and the expected physicochemical properties. In such embodiments of the system of the invention, the subsystem adjusts the operation of the production facility in response to the triggered sensors to obtain a stickiness level of the mixture based on an SPR (Synthetic Product Reduction) profile provided by the device.

[0022] In embodiments of the system of the invention, the mixing means is chosen from one or more extruders and / or one or more internal mixers.

[0023] Other aspects of the invention will become evident from the following detailed description. Brief description of the drawings

[0024] The nature and various advantages of the invention will become more evident upon reading the following detailed description, together with the accompanying drawings, on which the same reference numbers designate identical parts throughout, and in which: [ Fig 1 ] There figure 1 represents a set of chemical groups obtained from heated rubber. [ Fig 2 ] There figure 2 represents an embodiment of a system for manufacturing rubber products of the invention. [ Fig 3 ] There figure 3represents one embodiment of the system of the figure 2 . [ Fig 4 ] There figure 4 represents an embodiment of a manufacturing control process for rubber products carried out by the system of the invention. Detailed description

[0025] Referring now to the figures, in which the same numbers identify identical elements, the figure 2represents a rubber product manufacturing system (or "system") 100 of the invention. System 100 performs a manufacturing control process for rubber products in response to data input obtained by a detection tool that enables physicochemical classification (also called an "electronic nose") of the system. System 100 is usable in facilities where rubber products are manufactured (including tires). It is understood that System 100 can operate in various physical environments without prior knowledge of their parameters (for example, an initial arrangement of a rubber product production line of which System 100 is a part).

[0026] System 100 includes a production plant (or "plant") 110 which implements successive mixing and end-of-line stages. As shown in the figure 2The production installation 110 includes at least one mixing means 112 that carries out successive mixing stages of a rubber compound. By way of example, the mixing means 112 may include at least one extruder 112a with a frame having assembled common parts, which may include, without limitation, a screw-barrel assembly (with or without its optional heating and cooling accessories), a drive unit (reducer and coupling), a main motor, devices for feeding material (for example, metering units or hoppers 112b), a control cabinet that houses the motor drives, starting and safety devices, and regulating, controlling, displaying, and measuring devices. The extruder 112a may be a single-screw extruder, a twin-screw extruder, or a multi-screw extruder.

[0027] It is understood that the mixing means 112 may incorporate, instead of the extruder 112a, one or more known internal mixers (not shown) that produce an initial mixture of elastomeric materials with a carbon black and / or silica filler. By "internal mixer" is meant a machine consisting of a rammer and two half-tanks (or "tanks"), each containing a rotor with one or more blades (for example, a Banbury or Intermix type machine for polymers). In another example, the mixing means 112 may incorporate at least one automated external mixer (also called a "roller mixer" or "roller tool") into which this working mixture is then transferred by circulating it again between two rollers so as to transform it into a continuous sheet.Vulcanizing products (including, but not limited to, sulfur) may be added to the mixture later in a mixing cycle to obtain the final mixture for commercial use.

[0028] For example, the production installation 110 further includes one or more roller tools 114 that shape the mixture M exiting the extruder 112a into a strip 115. The strip 115 passes to one or more roller tools 114 for continuous shaping to a predetermined width and for cooling of the manufactured mixture. It is understood that each roller tool may include internal cooling means as known to those skilled in the art.

[0029] The production installation 110 may include a device for cutting or shaping the extruded material. For example, the continuous strip 115 formed may be wound into rolls (represented, for example, by roll 117 of the figure 2) to facilitate the storage of the produced mixture.

[0030] It is understood that the configuration of production installation 110 is given as an example and that other configurations may be incorporated into system 100.

[0031] Referring again to the figure 2 The system 100 also includes an odor detection device (or "device") 120 for recognizing the presence of a target compound by measuring its concentration using surface plasmon resonance ("SPR"). The device 120, which receives ambient air from around the production facility 110, may include a fan 121 (or equivalent suction device) that draws in ambient air and feeds it into the device. The ambient airflow may be regulated by a venting means that can selectively retain and exhaust ambient air (for example, a valve that can be selectively opened and closed).

[0032] The device 120 also includes one or more sensors 122 that detect the presence of odorous volatile organic compounds (or "VOCs") in the ambient air around the production facility 110. The sensors 122 incorporate a sensitive part that interacts with the VOCs (for example, VOCs emitted by the mixture M being produced at the production facility 110 and represented by molecules A and B in the figure 2 Each 122 sensor can detect compounds from a predetermined family of compounds. As understood by those skilled in the art, this sensitive part can consist of non-biological materials (such as semiconductor metal oxides (or "MOS") and semiconductor polymers) or organic molecules (e.g., peptides).

[0033] The device 120 further comprises a thin film, and in particular a metallic layer 124, having a surface that remains in contact with the ambient air entering the device. The metallic layer 124 comprises a layer of a metal known to produce the SPR effect (for example, gold or silver). The sensors 122 are arranged on this metallic layer 124 in a predetermined position.

[0034] The device 120 also includes a prism 126 having an entrance face for light entry, an exit face for light exit, and a support face on which the metallic layer is disposed. In a process achieving the SPR effect, a lighting device 127 of the device 120 emits collimated light through the entrance face of the prism 126 onto a surface of the metallic layer 124, this surface having a known reflectivity. The lighting device 127 can be selected from commercially available lighting devices, including, but not limited to, LED (light-emitting diode) lights.

[0035] The lighting device 127, being sensitive to the refractive index of the ambient air present in the device 120, produces a plasmon resonance on the surface of the metallic layer 124. This resonance, being sensitive to the refractive index of the ambient air present in the device 120, decreases the reflectivity of the metallic layer 124 so that it varies in the vicinity of each sensor 122. The device 120 can be chosen from commercially available odor detection devices (or "electronic noses") (for example, of the "NeOse" type offered by the company Aryballe, but it is understood that other equivalent devices can be used).

[0036] The device 120 incorporates at least one processor 128 which is configured to detect VOCs present in the ambient air and captured by the sensors 122. The term "processor" (or, alternatively, the term "programmable logic circuit") means one or more devices capable of processing and analyzing data and comprising one or more software programs for their processing (for example, 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 128 includes one or more software programs for processing volatile compounds captured by the device 120 (and the corresponding data obtained), as well as one or more software programs for identifying and classifying odor profiles to enable the identification of VOCs during the production of rubber compounds. The processor 128 also includes one or more software programs for processing subsystems associated with system 100 (and the corresponding data obtained), as well as one or more software programs for identifying variances and their sources in order to correct them.

[0037] To effectively manage the recognition of odor entering device 120, it is necessary to identify, within the VOCs present in the ambient air around the production facility 110, the odor characterizing the rubber product(s) being produced. In particular, identifying olfactory profiles corresponding to several volatile compounds is relevant for determining correlations between a physicochemical property related to rubber (e.g., stickiness) and changes in VOCs from rubber compounds. The identification of olfactory profiles can be carried out by constructing one or more models associated with the chemical SPR aspects and the olfactory aspects of one or more target compounds (also called the "olfactory profile model").An olfactory profile model can be used by learning the physicochemical properties associated with rubbery mixtures during rubbery product production cycles.

[0038] The term "target compound" (in the singular or plural) is used here to refer to a compound associated with a rubbery product (including mixture M) being produced in the physical environment of system 100 and identified based on SPR data obtained by device 120. To create a "black box" of rubbery mixtures and their odor profiles, the parameters of different rubbery mixtures can be used to form one or more odor profile models. This data accumulated in the black box can be used to make decisions regarding the control of parameters in the production facility 110 of system 100 by examining the odor profiles of rubbery mixtures exiting the facility, current available odor profiles, odor profiles of similar mixtures, and / or the time spent detecting particular VOCs in a mixing cycle.

[0039] Referring again to the figure 2An artificial intelligence embodiment can be used by the system 100 to construct at least one olfactory profile model. In some embodiments, the processor 128 (alone or in combination with one or more other processors) can configure the system 100 (and in particular the device 120) on one or more physicochemical properties recorded in the olfactory profile model. The processor 128 can also refer to a reference to perform a final determination of the expected physicochemical property or properties. The reference can include at least one reference base incorporating, for example, a table of VOCs of various rubbery mixtures (including, without limitation, physicochemical properties corresponding to the VOCs at a specific time during a mixing cycle of a rubbery mixture).The processor 128 can compare the physicochemical properties of the mixture M exiting the mixing means 112 with the odor profile of the identified mixture, including the VOCs revealed by the SPR effect, so that the system 100 can adjust the production plant 110 where the desired properties of the mixture M are not being achieved (for example, by sending an adjustment command to the production plant 110 to predict when the desired properties of the mixture will be achieved). The processor 128 can retrieve the physicochemical properties of the ambient air captured in the device 120 that most closely match the known properties for configuring the device. An ambient air reference can also include VOC measurements corresponding to a plurality of known odor profiles (see, for example, the . figure 1 ).

[0040] The ambient air data captured by device 120 is transferred to and stored in the memory of processor 128. Processor 128, which executes instructions from a data processing module, analyzes the data to determine one or more odor profiles corresponding to the rubber compound being produced at production facility 110. These odor profiles are generally indicators of the physicochemical characteristics of the compound being produced. The detection of different chemical groups, such as VOCs, allows for the differentiation of emissions from compound M. Processor 128 can detect changes in the properties of compound M to identify at least one associated volatile compound. Other characteristics of compound M can also be determined.

[0041] The captured data can be applied to a determinant of physicochemical properties, including VOCs, which can leverage one or more machine learning models to generate the target compounds. While the examples described here focus on the use of neural networks (specifically convolutional neural networks, or CNNs) as the machine learning model, other types of machine learning models can be employed.These include, but are not limited to, models using linear regression, logistic regression, decision trees, support vector machines, naive Bayes, nearest neighbor (knn), random forest, dimensionality reduction algorithms, gradient descent algorithms, neural networks (e.g., autoencoders, CNNs, RNNs, perceptrons, log short-term memory (LSTM), Hopfield, Boltzmann, deep belief, deconvolution, generative confrontation (GAN), etc.) and their complements and equivalents. The CNN(s) can be trained with ground truth data represented in a VOC reference created during a rubber product manufacturing control process performed by System 100.

[0042] In one embodiment, System 100 could be optimized using the principle of progressive learning (or "shaping"). Progressive learning allows the learning of new tasks by reusing features learned in previous tasks and by utilizing features currently being learned. At some point, the learned features can be used to perform offline tasks, which may be unsupervised, supervised, or semi-supervised. The model, having prior experience with the policy to adopt in a similar problem (for example, assigning VOCs to a sticky, rubbery mixture), progresses faster than the performance obtained through learning from scratch.For the disclosed invention, training a neural network to identify the primary factors of SPR, thereby enabling the estimation of a physicochemical property (e.g., stickiness, aging, pollution, etc.), can require an enormous amount of data to obtain the desired result. Therefore, incremental learning can be chosen to optimize training time.

[0043] System 100 of the invention therefore uses the SPR effect to provide one or more digital fingerprints (or "SPR fingerprints" or "fingerprints") of the captured VOCs. For example, the EA and EB fingerprints of the figure 2 represent the VOCs emitted by mixture M.

[0044] Referring again to the figure 2The system 100 further includes a control subsystem (or "subsystem") 130. Subsystem 130 uses the output olfactory profile model from the neural network to control the operation of the production facility 110 in response to the presence of target compounds in the ambient air of the facility. The sensors of subsystem 130 can trigger when the output olfactory profile model indicates a discrepancy between the physicochemical properties of the mixture M being produced at the production facility 110 (represented by the SPR fingerprint of the mixture supplied by device 120) and the expected physicochemical properties. The sensor(s) 122 of device 120 can detect the presence of target compounds in the ambient air around the production facility 110, which triggers subsystem 130 to change the rotational speed of the screws in the extruder 112a.As an example, the sensors 122 of the device 120 can detect the presence of the target compounds at a level indicating a mixture that is too sticky, allowing a modification of the raw materials entering the extruder 112a. In another example, the detection of ambient air by the device 120 can signal a flow requiring a change of screw (for example, the screws can be chosen from interpenetrating co-rotating profiles with conjugate profiles, conjugate or non-conjugate profiles, parallel or conical screws, single-thread or multi-thread profiles, and modular or non-modular screws).

[0045] In some embodiments, subsystem 130 includes one or more sensors (not shown) that trigger when an image of the production facility 110, together with the outgoing odor profile model, indicates a discrepancy between the physicochemical properties of the rubber compound being produced at the production facility 110 and the expected physicochemical properties. In this sense, a sensor of subsystem 130 may include a camera, a still camera, an optical sensor, and / or other equivalent detection methods.

[0046] Referring again to the figure 2 and furthermore to the figure 3 Another embodiment of system 100 is schematically represented. In this embodiment of system 100, system 100 comprises the production installation 110 and the device 120 as described above, in relation to the embodiment shown in the figure 2System 100 also includes at least one filter 140 to remove VOC pollution from the environment surrounding the production facility 110. Filter 140 reduces the VOC pollution (in sulfur, for example), thus providing a stable reference for building the odor profile model. In one embodiment of filter 140, the filter includes an activated carbon filter, which is well known as an air purifier. To capture VOCs from the mixture(s) being produced, system 100 further includes a gas concentration dome 142 that communicates with a condenser 144 of system 100 (the dome 142 is chosen from commercially available domes, for example, of the type offered by SoluProTech). The condenser 144, which is chosen from among the known and commercially available condensers, makes it possible to reduce the humidity and temperature of the ambient air.System 100 also includes a multiplexer 146 allowing the analysis of multiple target compounds before the SPR analysis performed by device 120.

[0047] Referring again to figures 2 And 3 , and furthermore to the figure 4 , a detailed description is given by way of example of a manufacturing control method for rubber products in response to the input of data obtained by the device 120 (or "manufacturing control method" or "method") 200 of the invention carried out by the system 100.

[0048] By initiating a manufacturing control process for rubber products 200 of the invention, the process includes a start-up step 202 of a mixing cycle in the production plant 110. This step includes a step of introducing, into the mixing means 112, the various raw materials necessary for the production of rubber products having desired properties (see arrow A of the figure 2These raw materials include, but are not limited to, an elastomeric material (e.g., natural rubber, 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, resins, and crosslinking or vulcanizing agents (e.g., sulfur). All ingredients are introduced in varying quantities depending on the desired performance of the products obtained from the mixing processes (e.g., tires).

[0049] The mixing cycle can also be initiated by starting the cycle with a pre-mixed product that does not contain all the ingredients of the chosen mixing recipe (called a "masterbatch"). For example, resins and vulcanizing agents are not present in the masterbatch. In this case, the masterbatch is either retrieved hot from a mixer upstream of the production unit 110 (such as an internal or external mixer), or it is cold because it has been manufactured and packaged several hours or even days beforehand.

[0050] During this step, the mixing means 112 is implemented (for example, in cases where the mixing means includes the extruder 112a, the screw(s) of the extruder are rotated to cause movement of the rubbery mixture downstream of the extruder as soon as the raw materials are introduced) (see arrow B of the figure 2It is understood that the mixing cycle can be easily adapted for all embodiments of the production plant 110 (for example, embodiments incorporating internal mixers instead of the extruder 112a). During this step, the roller tools 114 shape the strip 115 of the mixture M exiting the mixing means 112, and the continuous strip 115 formed is wound into rolls 117 to facilitate storage of the mixture.

[0051] The manufacturing control method of the invention also includes a step 204 for capturing ambient air around the production unit 110 to obtain one or more gaseous samples during the production of the strip 115. This step, performed by the device 120 of the system 100, can be carried out iteratively depending on the number of samples intended to feed the neural network. The capture step implements an SPR effect using a local change in the refractive index of the rubbery mixture produced by the production unit 110.

[0052] The manufacturing control process 200 of the invention further includes a step of detecting odor profiles, including VOCs, present in the ambient air and captured by the sensors 122 of the device 120. During this step, these detected odor profiles are recorded in one or more databases 205. Different parameters, such as the characteristics of the gases (including the nature and concentration of the chemicals, the physico-chemical properties of the chemicals, for example, solubility, saturation vapor pressure and biodegradability), the operating conditions and variations in the humidity level of the environment can influence the duration of this step.

[0053] In embodiments of system 100 incorporating a filter 140, this step includes a step of removing VOC pollution from the ambient air before it enters the device 120.

[0054] The manufacturing control process 200 of the invention further includes an odor profile identification step 208 to identify the rubbery mixture being produced and its production state. This step is performed using at least one ambient air sample captured by the device 120. Identifications can be made depending on the number of samples to be captured. During this step, the extracted data is processed and normalized by the device 120 so that VOCs can be measured by the SPR effect. The device 120 provides a fingerprint of the VOCs present in each measured sample (see, for example, the EA and EB fingerprints shown in the figure 4 ). This "olfactory" imprint, being an image of the olfactory profile at the moment of its measurement, varies according to the treatments carried out on the mixture or its expected evolution during the execution of a mixing cycle.

[0055] In embodiments of system 100 incorporating condenser 144, this step includes a step of reducing the humidity and / or temperature in the ambient air captured by device 120.

[0056] In one embodiment, during the olfactory profile identification step 208, the device 120 can automatically produce several fingerprints (for example, during a chemical calibration as described in publication FR3063543). The signal representing the local optical index of a gaseous medium is intentionally variable to increase the robustness of the neural network and to ensure its acuity.

[0057] The manufacturing control process 200 of the invention further includes an analysis step 210 of the samples captured by the device 120. From the fingerprint provided by the device 120 and the extracted data represented therein, characteristics are extracted that feed into a VOC classification algorithm based on physicochemical properties. The characteristics mentioned herein may include, without limitation, the characteristics of sticky mixtures, de-cohesive mixtures, the addition of water to the mixtures, and the aging of the mixtures.

[0058] In embodiments of system 100 incorporating the multiplexer 146, this step includes a step of analyzing multiple target compounds in each sample before the analysis performed by the device 120.

[0059] The manufacturing control process 200 of the invention further includes a step of constructing the olfactory profile model based on the physicochemical properties of the identified rubbery mixture. To construct the olfactory profile model, this step includes a step 214 of introducing the physicochemical properties of the identified rubbery mixture into the neural network, this step being carried out by the system 100. During this step, physicochemical properties of the identified rubbery mixture are obtained by the device 120. The physicochemical properties obtained include data corresponding to the fingerprint provided by the device 120. This data is recorded (for example, in one or more databases 150 of the system 100) (see the figure 2 ), and they are updated during the manufacturing control process on a continuous or intermittent basis.

[0060] The introductory step 214 includes a step to create a training set (or "base") 215 of the physicochemical properties that are fed into the olfactory profile model. The physicochemical properties correspond to the olfactory profiles of the samples captured by the sensors 122 of the device 120. This step includes a step to create a reference of the desired olfactory profiles in the ambient air to be captured by the device 120. The reference of olfactory profiles created during this step includes expected olfactory profiles corresponding to the known olfactory profiles of the rubber compounds being produced at the production facility 110. This step can be performed in advance of other process steps to feed the neural network the true expected olfactory profiles by analyzing the ambient air captured during several mixing cycles.

[0061] The training set created may include at least one pre-existing reference set 160 (for example, a VOC table of various rubbery mixtures discussed above). The set may include parameters corresponding to a plurality of known rubbery mixture recipes (including parameters that are part of the background information). In embodiments of process 200, at least part of the odor profile reference is created by one or more persons skilled in the art (for example, during a recording of data captured from the ambient air of other installations producing a mixture similar to mixture M exiting production installation 110).

[0062] The manufacturing control process 200 of the invention further includes a training step 216 for the olfactory profile model. During this step, a machine learning method takes as input the olfactory profiles obtained from the rubbery mixture as well as the data from the created training set. After the system 100 has obtained the olfactory profiles of the identified rubbery mixture, the processor 128 can retrieve its corresponding known physicochemical properties to construct the olfactory profile model.

[0063] In one embodiment, the machine learning method used during training step 216 includes a supervised learning method. The supervised learning method may include training one or more neural networks as discussed above.

[0064] Thus, the output olfactory profile model will identify the physicochemical properties of the rubber compound being produced at production facility 110 (characterized, for example, by the achievement of the desired properties of the compound at a predetermined time). This allows, for example, preventive actions to be taken to ensure that the rubber compound achieves the desired properties. As an example, in certain use cases, once the learning process is complete, system 100 can estimate the stickiness produced from an SPR fingerprint and other basic data (e.g., temperature and humidity).

[0065] In embodiments of the manufacturing control method 200 of the invention, the method further includes a comparison step 218. During this step, the identified physicochemical properties of the mixture M exiting the mixing means 112, derived from the odor profile model, are compared to the odor profile of the identified mixture, so that the system 100 can adjust the production installation 110 where the desired properties of the mixture M are not achieved. In these cases, the method 200 further includes a control step 220 for adjusting at least one element of the production installation 110. During this step, the adjustment command is sent to the subsystem 130 to manage the production installation 110 (for example, to trigger the sensors of the subsystem 130 so that the speed of the extruder screws is slowed down).

[0066] During this step, System 100 updates the incoming data (see database 221) to facilitate the sharing of information on the physicochemical properties of rubber mixtures with a production site incorporating Production Facility 110 (including operators at the site incorporating Production Facility 110). System 100 implements Process 200 of the invention by using "stored" VOC measurement data and the "updated" data to prepare customized reminders to be sent to Production Facility 110 (or to a site incorporating Production Facility 110 or to one or more operators at the site incorporating Production Facility 110) and to predict the achievement of the desired properties of the mixture. In this embodiment, System 100 can suggest adjustments to mixing cycle parameters to improve the rubber mixture exiting Production Facility 110.

[0067] As used here, "operator" or "user" refers to a single operator or a group of operators. An operator includes, but is not limited to, an individual participant in a task of a mixing cycle performed by Production Facility 110, an individual member of a team or group participating in a mixing cycle, one or more machines associated with an individual or team participating in at least one task of a mixing cycle, a digital community associated with a mixing cycle, and combinations and equivalents thereof. The operator may also be a spectator who observes a mixing cycle, in whole or in part, physically or virtually (for example, by remotely managing a predetermined live operation in order to view Production Facility 110 in real time).As used here, "operator" can also refer to any electronic system or device configured to receive command input and configured to automatically send data to at least one other operator.

[0068] In embodiments of process 200 of the invention, process 200 further comprises a simulation step of several mixtures for producing rubbery products with predetermined physicochemical properties. Each simulation is subsequently used to predict the odor profile of a rubbery mixture with expected physicochemical properties via an odor profile model as described above, ultimately yielding a predicted distribution of odor profiles for the identified physicochemical properties.

[0069] System 100 of the invention is therefore based on a neural network whose fundamental principle is to be trained on a large number of situations (for example, ambient air samples) so that it can then describe a new rubbery mixture presented to it. On the one hand, the neural network must be told what it should recognize, and then taught this information ("annotation"). On the other hand, the performance of the neural network must be assessed, and relevant samples must be provided to avoid certain biases, such as overtraining, which would reduce the network's performance. Thus, the principle of System 100 is simply to perform what could be considered a difference between the marker elements of ambient air and the measured elements. The effects of environmental pollution are mathematically eliminated.

[0070] It is conceivable that one or more steps of the process could be carried out iteratively. EXAMPLE

[0071] Using reference data (here, the stickiness and associated SPR modeling), a minimally trained neural network can determine the stickiness level based on the SPR data. Considering the stickiness level of mixtures in manufacturing workshops, the properties of the mixture composition influence this level, notably: The ratio of reinforcing filler to plasticizer, with this ratio indicating a stickier mixture as the ratio decreases. The resin ratio in the mixture, with this ratio indicating a stickier mixture as the ratio increases. The elastomer volume fraction, which indicates a stickier mixture as the volume fraction decreases.

[0072] The characteristic of the elastomer that can contribute to the stickiness of the mixtures is also considered. In this example, a Mooney ML1+4 100°C of dry elastomer <50 points, or the presence of isoprene, is considered. The Mooney, also known as viscosity or plasticity, characterizes solid substances in a well-established manner. An oscillating consistometer, as described in the ASTM D1646 standard (1999), is used. This plasticity measurement is performed according to the following principle: the sample being analyzed in its raw state (i.e., before baking) is molded (formed) in a cylindrical chamber heated to a given temperature (e.g., 35°C or 100°C). After one minute of preheating, the rotor rotates inside the specimen at 2 revolutions per minute, and the torque required to maintain this rotation is measured during 4 minutes of rotation.The Mooney viscosity (ML 1 + 4) is expressed in "Mooney units" (with 1 MU = 0.83 Nm) and corresponds to the value obtained at the end of 4 minutes. The term "stickiness" defines a property of a rubber compound that varies depending on the formulations of the compounds. A very sticky compound is not necessarily an "extreme" compound in one of the criteria, but it is the combination of all the criteria that gives it its stickiness and therefore its manufacturing difficulty. By multiplying these criteria, a "stickiness index" is thus found and represented by the table below: [Table 1] 1 - Black / Plasticizer Ratio Black / Plasticizer index 150.38 140.30 130.23 120.15 110.08 100.00 89.92 79.85 69.77 59.70 49.62 Severity value 1 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00 2 - Presence of Resin Tackifying resin index 0.00 20.00 40.00 60.00 80.00 100.00 120.00 140.00 160.00 180.00 200.00 Severity value 1 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00 3 - Cohesion Mixture Volume fraction of index elastomer 143 134 126 117 109 100 91 83 74 66 57 Severity value 1 1.10 1.20 1.30 1.40 1.50 1.60 1.70 1.80 1.90 2.00

[0073] Empirically, and according to various industrial feasibility criteria, an assessment is carried out on the industrial feasibility of several mixtures. The feasibility limit is established for stickiness indices above 3.5 and 4.5 points. Examples of mixtures :

[0074]

[0075] Measuring the stickiness of a rubber compound is essential in the manufacture of certain rubber products. For example, in semi-finished products manufactured in the tire industry, stickiness determines the strength of the casing assembly and the cohesion of the products during shaping.

[0076] Thanks to system 100 and its use of the SPR effect by device 120, the disclosed invention is not only highly efficient but also flexible and adaptable on a case-by-case basis if needed or if operating conditions change. System 100 is therefore suitable for rubber products composed of a variety of rubber compounds. Consequently, the invention takes into account the quality of the measured and analyzed parameters to ensure the quality of the rubber products obtained.

[0077] It is planned to incorporate other electronic nose-type devices into the manufacturing processes of rubber products. For example, an electronic nose used may employ gas chromatography-based technology, electrochemical cell-based technology (e.g., for the analysis of target gases such as NH3, H2S, and CO2), photoionization detection (PID)-based technology, and interferometry-based technology (e.g., Mach-Zehnder interferometry used in cases where reduced sensitization to sulfur and other aggressive compounds is desired).

[0078] A cycle of the manufacturing control process of the invention can be carried out by PLC control and can include pre-programmed management information. For example, a process setting can be associated with the desired properties of a mixture produced by the production installation 110, including the properties of the mixing means 112, the properties of the raw materials entering the mixing means, and the properties of the mixture exiting the mixing means.

[0079] In embodiments of the invention, the system 100 (and / or a site incorporating the system 100) can receive voice commands or other audio data representing, for example, a start or stop of capturing samples of ambient air entering the device 120. The request may include a request for the current status of a mixing cycle carried out by the production facility 110. A generated response may be represented audibly, visually, tactilely (for example, using a haptic interface), virtually and / or augmented.

[0080] For all implementations of system 100, a monitoring system could be put in place. At least part of the monitoring system can be provided in a portable device such as a mobile network device (e.g., a mobile phone, a laptop, one or more portable network-connected devices (including "augmented reality" and / or "virtual reality" devices, network-connected wearable clothing and / or any combinations and / or any equivalents).

[0081] System 100 enables continuous measurement of rubber compounds resulting from mixing cycles and non-destructive testing that eliminates operator effect. System 100 of the invention makes the measurement of physicochemical characteristics more objective by eliminating the need for rubber sampling and ensuring repeatable measurements.

[0082] The terms "at least one" and "one or more" are used interchangeably. Ranges presented as being "between a and b" encompass the values ​​"a" and "b".

[0083] Although specific embodiments of the disclosed apparatus have been illustrated and described, it will be understood that various changes, additions, and modifications can be made without departing from the spirit or scope of this disclosure. Therefore, no limitations should be imposed on the scope of the invention described except those set forth in the appended claims.

Claims

1. A method (200) for controlling the manufacture of rubber products produced by a rubber-product manufacturing system (100) comprising a production facility (110) having at least one mixing means (112) that performs successive steps of mixing a rubber mixture, characterized in that the method comprises the following steps: - a step (202) of initiating a mixing cycle that is performed at the production facility (110); - a step (204) of capturing ambient air of the production facility (110) in order to obtain at least one gas sample during the mixing cycle, this step performed by an odor detection device (120) of the system (100) that recognizes the presence of the odorous volatile organic compounds (VOCs) present in the ambient air received in the device (120) by measuring its concentration using surface plasmon resonance (SPR); - a step of detecting olfactory profiles present in the ambient air and captured by the device (120); - a step (208) of identifying an olfactory profile in order to identify a mixture (M) exiting the mixing means (112) and the production status thereof, this step performed on the basis of at least one sample of the ambient air captured by the device (120); - a step (210) of analyzing the samples captured by the device (120) against one or more SPR imprints (EA, EB) supplied by the device (120) and the extracted data represented therein; - a step of constructing at least one olfactory-profiles model from the physicochemical properties of the identified rubber mixture, this step comprising a step (214) of introducing the physicochemical properties of the identified rubber mixture into a neural network, this step being performed by the system (100); and - a step (216) of training the olfactory-profiles model using the olfactory profiles of the identified rubber mixture; such that the output olfactory-profiles model will be the identification of the physicochemical properties of the rubber mixture in the process of being produced in the production facility (110).

2. The method (200) of Claim 1, wherein the step of constructing the olfactory-profiles model comprises a step of creating a reference of the olfactory profiles sought to be captured in the ambient air by the device (120).

3. The method (200) of Claim 2, wherein the introduction step (214) of the step of constructing the olfactory-profiles model comprises a step of creating a learning database (215) of the physicochemical properties that is introduced into the olfactory-profiles model.

4. The method (200) of any one of Claims 1 to 3, further comprising a comparison step (218) during which the identification of the physicochemical properties of the mixture (M) exiting the mixing means (112), output from the olfactory-profiles model, is compared against the olfactory profile of the identified mixture so that the system (100) can adjust the production facility (110) when the desired properties for the mixture (M) are not attained.

5. The method (200) of Claim 4, further comprising a step (220) of controlling the adjustment of the production facility (110) such that the mixture (M) exiting the mixing means (112) attains the desired properties.

6. The method (200) of Claim 5, wherein the step (220) of controlling the adjustment of the production facility (110) comprises a step of determining a level of tack of the mixture (M) from an SPR imprint supplied by the device (120).

7. The method (200) of any one of Claims 1 to 6, wherein, during the step (216) of training the olfactory-profiles model, the system (100) employs a learning method selected from a machine learning method and a progressive-learning method.

8. The method (200) of any one of Claims 1 to 7, wherein: - the step of detecting olfactory profiles comprises a step of eliminating VOC pollution from the ambient air received by the device (120), this step performed by a filter (140) of the system (100) before the ambient air enters the device (120); - the step (208) of identifying an olfactory profile comprises a step of reducing the moisture content and / or the temperature of the ambient air received by the device (120), this step performed by a condenser (144) of the system (100); and - the step (210) of analyzing the samples captured by the device (120) further comprises a step of analyzing olfactory profiles in the samples prior to the analysis performed by the device (120), this step performed by a multiplexer (146) of the system (100).

9. The method (200) of any one of Claims 1 to 8, wherein the step (202) of initiating a mixing cycle comprises a step of introducing, into the mixing means (112), the raw materials needed for performing the production of the mixture (M).

10. The method (200) of any one of Claims 1 to 8, wherein the step (202) of initiating a mixing cycle comprises a step of introducing, into the mixing means (112), one or more masterbatches.

11. The method (200) of any one of Claims 1 to 10, further comprising a step of simulating a number of mixing operations yielding at least one rubber mixture (M) having predetermined physicochemical properties.

12. A system (100) for manufacturing rubber products that performs a method (200) for controlling the manufacture of rubber products, characterized in that the system (100) comprises: - a production facility (110) that performs successive mixing steps, the production facility comprising at least one mixing means (112) from which one or more rubber mixture(s) (M) exit; - an odor detection device (120) that captures ambient air of the production facility (110) in order to obtain at least one gas sample during the mixing cycle; and - a control subsystem (130) that employs an olfactory-profiles model based on physicochemical characteristics of the rubber mixtures recognized by surface plasmon resonance (SPR); such that the device (120) recognizes the presence of odorous volatile organic compounds (VOCs) present in the captured ambient air by measuring the concentration thereof using surface plasmon resonance (SPR); and such that the olfactory-profiles model learns the physicochemical properties of the VOCs associated with the mixtures (M) exiting the mixing means (112) during the rubber-product production cycles performed by the production facility (110).

13. The system (100) of Claim 12, wherein the subsystem (130) comprises one or more sensors triggered when the output olfactory-profiles model indicates an offset between the physicochemical properties of the mixture (M) that is in the process of being produced in the production facility (110) and its expected physicochemical properties.

14. The system (100) of Claim 13, wherein the subsystem (130) adjusts the operation of the production facility (110) in response to the triggered sensors so as to obtain a level of tack of the mixture (M) from an SPR imprint supplied by the device (120).

15. The system (100) of any one of Claims 12 to 14, wherein the mixing means (112) is selected from one or more extruder(s) and / or one or more internal mixing mill(s).