Rubber Combination Optimization System and Method Using Artificial Neural Network Model

Abstract

The present invention relates to a method for optimizing rubber formulations using an artificial neural network model, comprising: a second step in which the AI ​​(220) predicts the composition of recommended formulation specifications, the physical properties of the recommended formulation specifications, and whether there are defects, and sends the result to the UI (210) of the AI ​​system (200); a third step in which the UI (210) of the AI ​​system (200) arranges the recommended formulation specifications in order of lowest unit price, queries the SI system (300) for raw material unit prices, receives a response, and outputs the recommended formulation specifications to the user terminal (100); a fourth step in which the UI (210) of the AI ​​system (200) queries the SI system (300) for similar formulation specifications, receives a response, arranges them in order of highest similarity, and outputs them to the user terminal (100); and a fifth step in which the UI (210) of the AI ​​system (200) calculates the processing cost and CT in response to a request for simulation results regarding the similar formulation specifications from the user terminal (100) and outputs the result to the user terminal. The method comprises a sixth step in which, when a user terminal (100) executes a physical property value prediction, the AI ​​(220) of the AI ​​system (200) receives a mixing specification from the UI (210) of the AI ​​system (200), predicts the physical property value of the mixing specification, and sends a response to the UI (210) of the AI ​​system (200); and a seventh step in which, when the user terminal (100) receives the physical property prediction value from the UI (210) of the AI ​​system (200) executes a mixing specification interface on the UI (210) of the AI ​​system (200), the UI (210) of the AI ​​system (200) queries and receives a response to the mixing specification interface from the SI system (300).

Description

Technology Field

[0001] The present invention relates to a rubber compound optimization system and method using an artificial neural network model constructed by utilizing big data accumulated in a solution for production management and product development. More specifically, the invention relates to a rubber compound optimization system and method using an artificial neural network model composed of a layer for predicting the physical properties of rubber products according to the rubber compound and an optimization layer for deriving the optimal rubber compound. Background Technology

[0003] Rubber compound products are produced through a process in which key raw materials such as rubber, carbon, oil, fillers, and chemicals are weighed according to the specified weights in the recipe, fed into a large mixer, and mixed under standard operating conditions such as high temperature and pressure; the mixture is then removed, passed through a large roll to form the shape desired by the customer (sheet or ribbon), cooled, and loaded onto pallets for packaging.

[0004] The domestic rubber industry achieved rapid development in the early 1990s with the introduction of tire technology and the possibility of exports, but it subsequently experienced a period of decline due to rising domestic GNP and competition from late-comer producing nations. However, since the 2000s, as the overall domestic industry has developed, the demand for rubber has begun to increase again; in particular, the synthetic rubber market has started to shift in line with the growth rate of the automotive market, accounting for more than 90% of the total market.

[0005] Although rubber parts used in automobiles account for only 5% of the total vehicle weight, they have begun to establish themselves as important components applied to critical parts that determine the performance of the vehicle. Recently, with the increasing performance and quality of automobiles, high-performance rubber products are receiving a lot of attention and interest.

[0006] The rubber product manufacturing industry as a whole recognizes the need to proactively respond to the crisis in the internal combustion engine parts industry and the emergence of new supply chains in the future automotive sector. Consequently, there is a competitive trend of investing in digital-based manufacturing innovation to address immediate challenges, such as labor shortages, stricter safety and environmental regulations, and rising manufacturing costs.

[0007] With the advancement of artificial intelligence technology, active research and development is being conducted across various industrial environments to derive optimal process conditions by applying AI models to production settings. Such process condition optimization technologies can improve overall productivity by increasing production process efficiency or reducing the probability of defects.

[0008] Furthermore, when optimization algorithms are applied to recipe derivation, recipe optimization technology has the advantage of preventing the waste of costs and resources caused by defects by reducing defects through the prediction of potential defects in advance during the product development and production stages and the derivation of an optimal recipe by modifying the recipe.

[0009] However, conventional rubber compounding methods had the problem of requiring significant cost and time, such as having to find the conditions suitable for the rubber product through repetition and testing of various component combinations. Prior art literature

[0011] Republic of Korea Patent Publication No. 10-2008-0112208. Republic of Korea Patent Registration No. 10-1585448. The problem to be solved

[0012] The present invention, aimed at solving the problems of the conventional technology described above, provides an artificial neural network model-based rubber compound optimization process and system capable of shortening the lead time for rubber compound development by constructing an artificial neural network AI solution using high-quality data accumulated over several years through the operation of production management and product development solutions, and predicting the physical properties of rubber products according to raw material combinations during rubber compound development. This complements the limitations of rubber compound development based on researchers' experience and try-and-error. means of solving the problem

[0014] Specific means for achieving the above objective include: a first step in which a user terminal sets mixing specification conditions in the UI of an AI system and recommends mixing specifications; a second step in which the AI ​​of the AI ​​system, having received the mixing specification conditions from the UI of the AI ​​system, predicts the composition of the recommended mixing specifications, the physical properties of the recommended mixing specifications, and whether defects exist, and sends the result back to the UI of the AI ​​system; a third step in which the UI of the AI ​​system arranges the raw material unit prices in order of lowest unit price, queries and receives the result from a solution for production management and product development (hereinafter referred to as the SI system), and outputs the recommended mixing specifications to the user terminal; a fourth step in which the UI of the AI ​​system queries and receives the result from the SI system regarding similar mixing specifications, arranges them in order of highest similarity, and outputs them to the user terminal; and a fifth step in which, upon a request for simulation results regarding similar mixing specifications from the user terminal, the UI of the AI ​​system calculates the processing cost and CT (cycle time, the sum of the time required for all process steps) and outputs the result to the user terminal. This is achieved by a rubber compound optimization method using an artificial neural network model, comprising: a sixth step in which, when the user terminal executes a physical property value prediction, the AI ​​of the AI ​​system receives a compound specification from the UI of the AI ​​system, predicts the physical property value of the compound specification, and sends a response to the UI of the AI ​​system; and a seventh step in which, when the user terminal that has received the physical property prediction value from the UI of the AI ​​system executes a compound specification interface on the UI of the AI ​​system, the UI of the AI ​​system queries and receives a response from the SI system regarding the compound specification interface.

[0015] According to one aspect of the present invention, the second step is characterized by repeating a sub-step of configuring a blending specification including a combination of raw materials and the input ratio of each raw material, a combination of process steps and process conditions such as RPM, temperature, and time for each process step, in accordance with conditions defined by a user terminal, and a sub-step of predicting physical property values ​​and defects of the blending specification, and recommending the blending specification with the smallest error between the condition value and the predicted value among combinations satisfying the conditions defined by the user terminal as the optimal blending specification.

[0016] According to another aspect of the present invention, the fifth step is characterized by including a step in which a user terminal modifies the mixing ratio and process steps.

[0017] According to another aspect of the present invention, in the seventh step, the predicted physical property values ​​are hardness, specific gravity, tensile strength, elongation, heat aging resistance, change in hardness, change in tensile strength, change in elongation, and compression set.

[0018] According to another aspect of the present invention, the prediction of the formulation specifications and physical property values ​​in the second step is characterized by being performed by an artificial neural network model constructed using big data of physical property values ​​for the overall formulation ratio, rather than the characteristics of each raw material.

[0019] Another objective of the present invention is a user terminal that sets mixing specification conditions and executes mixing specification recommendations, requests verification of simulation results, executes physical property value prediction and executes a mixing specification interface; a UI of the AI ​​system that transmits mixing specification conditions set by the user terminal, arranges recommended mixing specifications in order of lowest unit price and outputs them to the user terminal, queries raw material unit prices and similar mixing specifications, arranges them in order of highest similarity and transmits them to the user terminal, and calculates processing costs and CT (cycle time, sum of the time required for all process steps) of similar mixing specifications and outputs them to the user terminal; and an AI of the AI ​​system that receives mixing specification conditions from the UI of the AI ​​system, configures the recommended mixing specifications, predicts the physical property values ​​of the recommended mixing specifications and predicts whether the recommended mixing specifications are defective, and responds to the UI of the AI ​​system, and receives mixing specifications from the UI of the AI ​​system, predicts the physical property values ​​of the mixing specifications, and responds to the UI of the AI ​​system. This is achieved by a rubber compounding optimization system using an artificial neural network model, which is composed of an SI system that responds to raw material unit prices and similar compounding specifications in response to a request from the UI of the AI ​​system and responds to a compounding specification interface in response to the execution of the compounding specification interface of the user terminal to the UI of the AI ​​system.

[0020] According to one aspect of the present invention, the prediction of the mixing specifications and physical property values ​​is characterized by being performed by an artificial neural network model constructed using big data of physical property values ​​for the overall mixing ratio, rather than the characteristics of each raw material. Effects of the invention

[0021] As described above, the rubber compound optimization system and method using an artificial neural network model according to the present invention predicts physical properties according to the rubber compound through an artificial neural network model and provides a rubber compound that satisfies the required physical properties through an optimization model, thereby reducing the number of experiments required for the development of new rubber compounds, which has the effect of shortening the development lead time and reducing research and development costs. Brief explanation of the drawing

[0023] FIG. 1 is a block diagram showing a rubber compound optimization system using an artificial neural network model according to an embodiment of the present invention. FIG. 2 is a flowchart showing the process of optimizing a rubber compound using an artificial neural network model according to an embodiment of the present invention. Specific details for implementing the invention

[0024] Terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings, but should be interpreted in a meaning and concept consistent with the technical spirit of the invention, based on the principle that the inventor can appropriately define the concept of the terms to best describe his invention.

[0025] Therefore, the embodiments described in this specification and the configurations illustrated in the drawings are merely the most preferred embodiments of the present invention and do not represent all of the technical ideas of the present invention; thus, it should be understood that various equivalents and modifications that can replace them may exist at the time of filing this application.

[0027] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings.

[0028] FIG. 1 is a block diagram showing a rubber compound optimization system using an artificial neural network model according to an embodiment of the present invention.

[0029] According to FIG. 1, a rubber compound optimization system using an artificial neural network model according to an embodiment of the present invention comprises: a user terminal (100) that sets compound specification conditions and executes compound specification recommendations, requests verification of simulation results, executes physical property value prediction and executes a compound specification interface; a UI (210) of an AI system (200) that transmits compound specification conditions set by the user terminal (100), arranges recommended compound specifications in order of lowest unit price and outputs them to the user terminal (100), queries raw material unit prices and similar compound specifications and arranges them in order of highest similarity and transmits them to the user terminal (100), and calculates the processing cost and CT (cycle time, sum of the time required for all process steps) of similar compound specifications and outputs them to the user terminal; The AI ​​(220) of the AI ​​system (200) receives mixing specification conditions from the UI (210) of the AI ​​system (200), configures the recommended mixing specification, predicts the physical property value of the recommended mixing specification and predicts whether the recommended mixing specification is defective, and responds to the UI (210) of the system (200), and receives the mixing specification from the UI (210) of the AI ​​system (200), predicts the physical property value of the mixing specification, and responds to the UI (210) of the AI ​​system (200); and the SI system (300) responds to the raw material unit price and similar mixing specification according to the request of the UI (210) of the AI ​​system (200), and responds to the mixing specification interface according to the execution of the mixing specification interface of the user terminal (100) to the UI (210) of the AI ​​system (200).

[0030] FIG. 2 is a flowchart showing the process of optimizing a rubber compound using an artificial neural network model according to an embodiment of the present invention.

[0031] According to FIG. 2, first, a user (user, 100) accesses the UI (210) of the AI ​​system (200) to execute a project (S1) and then sets the mixing specification conditions (S2).

[0032] That is, the user terminal (100) sets 13 sub-categories of polymer types with a raw material code starting with E, and crosslinking agent types and special function types of raw material groups with a raw material code starting with CV. Here, the polymer types, crosslinking agent types, and special function types may increase or decrease.

[0033] Meanwhile, it is obvious that in the present invention, the above-mentioned raw materials, namely polymer type, crosslinking agent type, special function (special material), etc., are stored and built in the AI ​​system (200) through a database (DB).

[0034] In addition, the user terminal (100) sets the inspection temperature, inspection time, and physical property values, namely hardness, specific gravity, tensile strength, elongation, heat aging resistance, change in hardness, change in tensile strength, change in elongation, and permanent compression set. Here, the types of physical properties may be reduced or increased compared to the above standards.

[0035] And, the user terminal (100) sets the mixing specification type to CMB or FMB and the equipment to the corresponding production equipment.

[0036] The user terminal (100) accesses the UI (210) of the AI ​​system (200) to execute a mixing specification recommendation (S3), wherein the recommended mixing specifications may include the type of mixing specification, details of the mixing ratio, process steps, and predicted values ​​of physical properties by physical property.

[0037] Specifically, CMB or FMB can be recommended as the type of formulation specification.

[0038] In addition, it provides detailed recommendations for mixing ratios, including input classification, input order, raw materials, phr, input quantity, weighing number, and number of divisions; it can recommend information such as the input classification and order, as well as the phr and input quantity for each raw material.

[0039] Input classification is information regarding which of the input raw materials correspond to polymer (P), chemical (CH), carbon black (CB), oil 1 (OA), oil 2 (OB), filler (F), or manual input (M).

[0040] The order of input is the order of input of raw materials, P is alone, CH, OA, and OB can be input together, CB, OA, and OB can also be input together, and P, CH, and CB cannot be input simultaneously.

[0041] phr is the input amount when the polymer weight ratio is set to 100, and the input amount is the actual input amount, with the maximum weight that can be input varying by machine. The weighing number is input information for raw materials corresponding to CH or F and is used in the weighing process. Regarding the number of splits, it is recommended to split once whenever the weight of the input raw material is 15 kg or more.

[0042] The user terminal can recommend detailed information on the equipment capacity and process steps for each process step. Detailed information for each process step, such as process type, temperature-time condition, RPM, time (seconds), and temperature, includes process types such as P, CH, CB, OA, OB, F, M, MIX, SCRAP, and DUMP, temperature-time condition: temperature-time condition indicating whether to satisfy both temperature and time conditions simultaneously (AND) or only one (OR), blade rotation speed (RPM) of each process step, maximum time of each process step, and maximum temperature of each process step.

[0043] As described above, when the user terminal (100) recommends a mixing specification to the UI (210) of the AI ​​system (200), the UI (210) transmits the recommended mixing specification condition to the AI ​​(220) of the AI ​​system (200) (S4).

[0044] Next, the AI ​​(220) of the AI ​​system (200) constructs a list of recommended mixing specification candidates in a mixing specification candidate generation model (S5) and transmits each mixing specification to a physical property prediction model to predict the physical property value of the recommended mixing specification (S6), and at the same time transmits each mixing specification to a defect prediction model to predict whether the recommended mixing specification is defective (S7).

[0045] Specifically, the model for generating candidate mixing specifications uses RULE base, AI models, optimization algorithms, etc., either alone or in combination, to generate 5 sets of input raw material bundles and process step bundles that meet the mixing specification conditions.

[0046] In addition, the physical property prediction model derives predicted physical property values ​​for each recommended mix specification generated by the above mix specification candidate generation model using an AI regression model; at this time, a single prediction model can be configured for each physical property, or a model that predicts multiple physical properties at once can be configured.

[0047] In addition, the defect prediction model derives the defect status of each physical property inspection based on the recommended formulation specifications generated by the above formulation specification candidate generation model using an AI classification model; at this time, a single prediction model can be configured for each physical property, or a model that predicts the results of multiple physical properties at once can be configured.

[0048] As described above, the AI ​​(220) of the AI ​​system (200) that performed the process from S5 to S7 sends the recommended mixing specifications back to the UI (210) (S8).

[0049] Then, the UI (210) of the AI ​​system (200) calculates the total unit price of five recommended mixing specifications and displays them on the screen in order of lowest total unit price (S9), and queries the SI system (300) for the unit price of each raw material (S10), receives a response (S11), and displays it on the screen of the user terminal (100) (S12).

[0050] Then, the UI (210) of the AI ​​system (200) searches for a mixing ratio that is similar to each recommended mixing specification and mixing specification configuration in the SI system (300) (S13)(S14). At this time, the similarity can be derived as a percentage by combining the types and ratios of raw materials.

[0051] Next, the UI (210) of the AI ​​system (200) arranges the mixing ratios with similar mixing specifications in order of high similarity (15), and outputs the similar mixing specifications on the screen of the user terminal (100) (S16).

[0052] When the user terminal (100) sees this and requests the UI (210) of the AI ​​system (200) to check the simulation results (17), the UI (210) of the AI ​​system (200) calculates the processing cost (CT * standard processing cost per equipment) and CT (cycle time, sum of the time required for all process steps) according to the process steps of the recommended mixing specifications (S18), and outputs the simulation results to the screen of the user terminal (100) again (S19).

[0053] The user terminal (100), having confirmed the simulation results through the screen in this way, accesses the UI (210) of the AI ​​system (200) and modifies the mixing specifications, such as the type of raw material, the amount of raw material input, the process step, the RPM of the process step, time, and temperature.

[0054] Afterwards, when the user terminal (100) accesses the UI (210) of the AI ​​system (200) and executes the prediction of physical properties (S21), the UI (210) of the AI ​​system (200) transmits the mixing specifications to the AI ​​(220) of the AI ​​system (200) (S22), and the AI ​​(220) of the AI ​​system (200) predicts physical properties such as hardness, specific gravity, tensile strength, elongation, heat aging resistance, change in hardness, change in tensile strength, change in elongation, and compression set rate according to the mixing specifications modified in S20, using an AI regression model (S23), and sends the predicted physical properties to the UI (210) of the AI ​​system (200) (S23').

[0055] Then, the UI (210) of the AI ​​system (200) outputs the predicted physical property value to the screen of the user terminal (100) (S24).

[0056] Accordingly, when the user terminal (100) connects to the UI (210) of the AI ​​system (200) and executes the mixing specification interface (S25), the UI (210) of the AI ​​system (200) interfaces the mixing specification selected by the user terminal, such as the mixing specification code, mixing specification details, and process step details, to the SI system (300) (S25').

[0057] Then, finally, the SI system (300) sends the result of the mixing specification interface to the UI (210) of the AI ​​system (200) (S26).

[0058] The present invention is not limited to the above embodiments, and various modifications and variations are possible from this description by those skilled in the art. Accordingly, the concept of the present invention should be understood only by the claims set forth below, and all equivalent or analogous variations thereof shall be considered to fall within the scope of the concept of the present invention. Explanation of the symbols

[0060] 100 : User Terminal 200 : AI System 210 : UI 220 : AI 300 : SI System

Claims

Claim 1 A first step in which the UI (210) of the AI ​​system (200) is connected to a user terminal (100) to execute the setting of mixing specification conditions and the recommendation of mixing specifications; a second step in which the AI ​​(220) of the AI ​​system (200), having received the mixing specification conditions from the UI (210) of the AI ​​system (200), predicts the composition of the recommended mixing specifications, the physical properties of the recommended mixing specifications, and whether there are defects, and sends a response to the UI (210) of the AI ​​system (200); a third step in which the UI (210) of the AI ​​system (200) arranges the raw material unit prices in order of lowest unit price, queries the SI system (300) for the unit prices of raw materials, receives a response, and transmits the recommended mixing specifications to the user terminal (100); and a third step in which the UI (210) of the AI ​​system (200) queries the SI system (300) for similar mixing specifications, receives a response, arranges the similar mixing specifications in order of highest similarity, and transmits the similar mixing specifications to the user terminal (100). A fourth step of transmitting to a terminal (100); a fifth step in which the UI (210) of the AI ​​system (200), having received a request from the user terminal (100) for a simulation result regarding a similar mixing specification, calculates the processing cost and CT and transmits the simulation result to the user terminal (100); a sixth step in which, when a physical property value prediction is executed by the user terminal (100), the UI (210) of the AI ​​system (200) transmits the mixing specification to the AI ​​(220) of the AI ​​system (200) and receives a response for the predicted physical property value of the mixing specification predicted by the AI ​​(220) of the AI ​​system (200). A rubber compound optimization method using an artificial neural network model, comprising a seventh step in which, when a compound specification interface is executed by a user terminal (100) that receives a physical property prediction value from the UI (210) of the AI ​​system (200), the UI (210) of the AI ​​system (200) queries the SI system (300) for the compound specification interface and receives a response. Claim 2 A method for optimizing rubber formulation using an artificial neural network model according to claim 1, wherein the second step is characterized by repeating a sub-step for configuring a formulation specification including a combination of raw materials and the input ratio of each raw material, a combination of process steps and process conditions such as RPM, temperature, and time of each process step, in accordance with conditions defined by a user terminal, and a sub-step for predicting physical property values ​​and defects of the formulation specification, and recommending the formulation specification with the smallest error between the condition value and the predicted value among combinations satisfying the conditions defined by the user terminal as the optimal formulation specification. Claim 3 A method for optimizing rubber formulation using an artificial neural network model, wherein, in claim 1, the fifth step includes a step in which a user terminal (100) modifies the formulation ratio and process steps. Claim 4 A method for optimizing rubber formulations using an artificial neural network model, characterized in that, in the 7th step of claim 1, the predicted physical property values ​​are hardness, specific gravity, tensile strength, elongation, heat aging resistance, change in hardness, change in tensile strength, change in elongation, and permanent compression set. Claim 5 A method for optimizing rubber formulation using an artificial neural network model, wherein, in claim 1, the prediction of the formulation specifications and physical property values ​​in the second step is performed by an artificial neural network model constructed using big data of physical property values ​​for the total formulation ratio, rather than the characteristics of each raw material. Claim 6 A user terminal (100) that sets mixing specification conditions and executes mixing specification recommendations, requests verification of simulation results, executes physical property value prediction and executes a mixing specification interface; a UI (210) of an AI system (200) that transmits mixing specification conditions set by the user terminal (100), arranges recommended mixing specifications in order of lowest unit price and outputs them to the user terminal (100), searches for raw material unit prices and similar mixing specifications, arranges them in order of highest similarity and transmits them to the user terminal (100), and calculates processing costs and CTs of similar mixing specifications and outputs them to the user; a UI (210) of an AI system (200) that receives mixing specification conditions from the UI (210) of the AI ​​system (200), configures recommended mixing specifications, predicts physical property values ​​of recommended mixing specifications and predicts whether there are defects in recommended mixing specifications, and responds to the UI (210) of the AI ​​system (200), and receives mixing specifications from the UI (210) of the AI ​​system (200). A rubber compounding optimization system using an artificial neural network model, comprising: an AI (220) of an AI system (200) that predicts the physical property values ​​of the compounding specifications and replies to the UI (210) of the AI ​​system (200); and an SI system (300) that replies to the raw material unit price and similar compounding specifications in response to a request from the UI (210) of the AI ​​system (200), and replies to the compounding specification interface in response to the execution of the compounding specification interface of the user terminal (100) to the UI (210) of the AI ​​system (200). Claim 7 A rubber compounding optimization system using an artificial neural network model, characterized in that, in claim 6, the prediction of the compounding specifications and physical property values ​​is performed by an artificial neural network model constructed using big data of physical properties for the total compounding ratio, rather than the characteristics of each raw material.

Images