Tire performance prediction apparatus, performance prediction method, and method for generating learned model

Abstract

To provide a tire performance prediction apparatus capable of estimating distribution data of physical quantities acting on an outer surface of a tread portion or inside the tread of a target tire by using a learned model.SOLUTION: A tire performance prediction apparatus includes a derivation unit which inputs data 21 including at least a pattern image 21A of a target tire to a learned model 24B, which has been machine-learned so as to output distribution data 22 of physical quantities acting on a tread portion when the tread portion is in contact with a road surface in a case where the data 21 including at least the pattern image 21A of the tread portion of a tire that is not in contact with the road surface is input, and outputs distribution data 22 of the physical quantities of the target tire.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present invention relates to a tire performance prediction device, a performance prediction method, and a method for generating a trained model. [Background technology]

[0002] Patent Document 1 listed below describes a tire image recognition method. This method involves inputting an image of the contact patch of a target tire to be recognized into a machine learning model that has learned the relationship between an image representing the tire contact patch and the contour of the tire contact patch, deriving an output, and estimating the contour of the contact patch of the target tire based on the derived output. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2024-048936 Summary of the Invention [Problem to be solved by the invention]

[0004] Because the tread portion of a tire is the part that comes into contact with the road surface, it is extremely important in tire development to know how the physical quantities acting on this tread portion are distributed. However, since such physical quantities require actual measurement using, for example, a tire testing device, obtaining the distribution of physical quantities requires a lot of effort. Therefore, there has been a demand for a device that can estimate distribution data of physical quantities using a trained model.

[0005] The present invention has been devised in consideration of the above-described circumstances, and its main object is to provide a tire performance prediction device that is capable of estimating distribution data of physical quantities acting on the tread portion of a target tire using a trained model. [Means for solving the problem]

[0006] The present invention is a tire performance prediction device that includes a derivation unit that inputs data including at least a pattern image of a target tire into a trained model that has been machine-learned so that, when data including at least a pattern image of the tread portion of a tire that is not in contact with the road surface is input, the trained model outputs distribution data of physical quantities that act on the tread portion when the tread portion is in contact with the road surface, and outputs distribution data of the physical quantities of the target tire. [Effects of the Invention]

[0007] By adopting the above-described configuration, the tire performance prediction device of the present invention is able to estimate distribution data of physical quantities acting on the tread portion of a target tire using a trained model. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram illustrating an example of a tire performance prediction device. [Figure 2] 1 is a development view showing an example of a tread portion of a tire. [Figure 3] 1 is a flowchart illustrating an example of a processing procedure for a method for generating a trained model. [Figure 4] 10 is a flowchart showing an example of a processing procedure of a preparation step. [Figure 5] FIG. 10 is a diagram illustrating an example of a training dataset. [Figure 6] FIG. 1 is a conceptual diagram illustrating an example of a machine learning model (trained model). [Figure 7] 1 is a flowchart showing an example of a processing procedure of a tire performance prediction method. [Figure 8] FIG. 1 is a partial cross-sectional view of a tire in contact with a road surface. [Figure 9] FIG. 10 is a diagram illustrating an example of a training dataset according to another embodiment of the present invention. [Figure 10] FIG. 10 is a diagram showing an example of a training dataset according to yet another embodiment of the present invention. [Figure 11]FIG. 1 is a conceptual diagram illustrating an example of a machine learning model (trained model) into which condition data is input. [Figure 12] (a) is a diagram showing distribution data estimated using a trained model, and (b) is a diagram showing distribution data calculated by simulation. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present invention will be described with reference to the drawings. It should be understood that the drawings include exaggerated representations and representations that differ from the dimensional ratios of actual structures in order to facilitate understanding of the contents of the invention. Furthermore, identical or common elements are designated by the same reference numerals throughout the embodiments, and redundant explanations will be omitted. Furthermore, the specific configurations shown in the embodiments and drawings are for the purpose of understanding the contents of the present invention, and the present invention is not limited to the specific configurations shown in the drawings.

[0010] [Tire performance prediction device] FIG. 1 is a block diagram showing an example of a tire performance prediction device 1A. The tire performance prediction device (hereinafter, sometimes referred to as a "prediction device") 1A is a device for predicting the performance of a target tire (a tire to be evaluated). The prediction device 1A of this embodiment is also configured as a device (hereinafter, sometimes referred to as a "generation device") 1B for generating a trained model. Note that the prediction device 1A and the generation device 1B may be configured separately.

[0011] The prediction device 1A (generation device 1B) of this embodiment is configured as a general-purpose computer 1. Examples of the computer 1 include a desktop personal computer, a laptop personal computer, a tablet, a smartphone, and a cloud server.

[0012] The prediction device 1A (generation device 1B) of this embodiment includes an input device 3, an output device 4, and a calculation processing device 5.

[0013] [Input section, output section, arithmetic processing unit] The input device 3 is configured as an input device. Examples of this input device include a keyboard or a mouse. The output device 4 is configured as an output device. Examples of this output device include a display device or a printer. The arithmetic processing device 5 is used to predict the performance of a target tire. The arithmetic processing device 5 of this embodiment is configured to include one or more processors (arithmetic units) 5A that perform various calculations, a storage unit 5B that stores data, programs, etc., and a working memory 5C.

[0014] [Processor] The processor 5A in this embodiment is configured as a central processing unit (CPU), but is not particularly limited thereto and may be configured as, for example, a microprocessor or other processing unit. In this embodiment, a mode is exemplified in which the tire performance prediction method (hereinafter sometimes referred to as the "prediction method") and the trained model generation method (hereinafter sometimes referred to as the "generation method") are executed by one processor 5A, but they may also be executed (parallel processed) by multiple processors (not shown).

[0015] [Storage] The storage unit 5B is a non-volatile information storage device formed of, for example, a magnetic disk, an optical disk, an SSD, etc. The storage unit 5B is provided with a data unit 7 and a program unit 8.

[0016] [Data section] The data unit 7 of this embodiment is for storing data and the like necessary for executing the prediction method and the generation method. The data unit 7 of this embodiment includes a target data storage unit 7A, a target distribution data storage unit 7B, a training dataset storage unit 7C, and a model storage unit 7D. Note that the data unit 7 is not limited to this configuration, and may include storage units for storing other data as needed, or some of these may be omitted. The data stored in each storage unit 5B will be explained in each step of the prediction method and generation method described below.

[0017] [Program section] The program unit 8 of this embodiment is a program (application) required to execute the prediction method and the generation method. When such program unit 8 is executed by the processor (calculation unit) 5A, the computer 1 (prediction device 1A and generation device 1B) can function as a specific means. The program unit 8 of this embodiment includes an input unit 8A, a derivation unit 8B, a preparation unit 8C, a learning unit 8D, and an evaluation unit 8E. Note that the program unit 8 is not limited to this configuration, and other programs may be included as needed, or some of these may be omitted. The functions of each program unit 8 will be explained in the respective steps of the prediction method and generation method described below.

[0018] [tire] 2 is a development view showing an example of the tread portion 12 of the tire 11. Although a pneumatic tire for a light truck is shown as an example of the tire 11 of this embodiment, the tire 11 is not limited to this example. The tire 11 may be a tire of another category, such as a tire for a passenger car, a tire for a heavy load such as a truck or bus, or an airless tire.

[0019] The outer surface 12o of the tread portion 12 of this embodiment includes a ground contact surface 14 that comes into contact with the road surface (not shown), and grooves 15 that are recessed from the ground contact surface 14 inward in the tire radial direction.

[0020] The groove 15 of this embodiment includes a longitudinal groove 15A and a lateral groove 15B. The longitudinal groove 15A of this embodiment extends continuously in a zigzag pattern in the tire circumferential direction, but may extend continuously in a straight line. In this embodiment, four longitudinal grooves 15A are provided to divide the land portion 16 into five portions, but this is not limited to this. The lateral groove 15B extends (crosses) the land portion 16 in the tire axial direction. This divides the land portion 16 into a plurality of blocks 17.

[0021] In the tread portion 12 of this embodiment, a tread pattern is formed in which a plurality of pitches P, which are pattern constituent units, are arranged in the tire circumferential direction by the longitudinal grooves 15A and the lateral grooves 15B. Each pitch P may be of only one type having the same circumferential length, or may be of multiple types having different circumferential lengths.

[0022] The outer surface 12o of the tread portion 12 may include sipes 18 that are narrower than the grooves 15 (for example, 1.5 mm or less). Furthermore, tie bars 19 that protrude radially outward in the tire may be provided at the bottom of the grooves 15. The longitudinal grooves 15A, lateral grooves 15B, sipes 18, tie bars 19, etc. form an uneven shape on the outer surface 12o of the tread portion 12.

[0023] Incidentally, since the tread portion 12 is the portion that comes into contact with the road surface (not shown), it is extremely important in developing the tire 11 to know how the physical quantities acting on the tread portion 12 are distributed. However, since such physical quantities require actual measurements using, for example, a testing device (not shown) for the tire 11, it takes a lot of time and effort to obtain the distribution of the physical quantities.

[0024] It is also known that the physical quantities vary significantly depending on the uneven shape (for example, longitudinal grooves 15A, lateral grooves 15B, sipes 18, tie bars 19, etc.) formed in the tread portion 12. Therefore, when estimating the distribution of physical quantities acting on the tread portion 12 using a machine learning model, for example, it is important to fully consider the characteristics of the uneven shape of the tread portion 12.

[0025] As a result of extensive research, the inventors discovered that it is effective to identify the uneven shape of the tread portion 12 using a pattern image (hereinafter sometimes referred to as a "pattern image") of the tread portion 12 of a tire 11 that is not in contact with the road surface, and to estimate the distribution of physical quantities acting on the tread portion 12.

[0026] In the prediction method (prediction device 1A) of this embodiment, features of the uneven shape of the tread portion 12 are incorporated into a machine-learned trained model (machine learning model) using data including at least a pattern image of the target tire. This allows distribution data of physical quantities acting on the tread portion of the target tire (hereinafter, sometimes referred to as "distribution data") to be estimated.

[0027] In this embodiment, prior to the implementation of the prediction method, a trained model is generated based on the generation method. Note that, when the prediction method is implemented, if a prediction model has already been generated, the generation method may be omitted.

[0028] [How to generate a trained model] In the generation method of this embodiment, a trained model is generated. The trained model of this embodiment is a model that has been machine-learned so as to output distribution data of a target tire (a tire to be evaluated) when data including at least a pattern image of the tread portion 12 is input. In order to create such a trained model, it is effective to train a machine learning model using a training dataset that combines data and distribution data acquired from multiple types of tires 11.

[0029] 3 is a flowchart showing an example of a processing procedure of a method for generating a trained model. Each step of the generation method of this embodiment is executed by one or more processors 5A included in the generation device 1B (computer 1) shown in FIG.

[0030] [Prepare the training dataset (preparation process)] In the generation method of this embodiment, first, a training dataset is prepared (preparation step S1). In the preparation step S1 of this embodiment, first, a preparation unit 8C included in the program unit 8 shown in FIG. 1 is loaded into the working memory 5C. The preparation unit 8C is a program for preparing the training dataset. When this preparation unit 8C is executed by the processor 5A, the computer 1 (generation device 1B) can function as a means for preparing the training dataset.

[0031] As described above, the training data set is used to train the machine learning model. This training data set is a combination of data including at least the pattern image of the tread portion 12 shown in Fig. 2 and distribution data of physical quantities acting on the tread portion 12.

[0032] The data may be only a pattern image, or may further include data other than the pattern image. The data in this embodiment includes only a pattern image.

[0033] The distribution data is not particularly limited as long as it indicates the distribution of physical quantities acting on the tread portion 12. Examples of such physical quantities include physical quantities acting on the outer surface 12o of the tread portion 12 (hereinafter, sometimes referred to as "outer surface physical quantities") and physical quantities acting inside the tread radially inward of the outer surface 12o (hereinafter, sometimes referred to as "internal physical quantities").

[0034] The distribution data in this embodiment includes only distribution data of outer surface physical quantities (hereinafter, sometimes referred to as "outer surface distribution data"), but is not particularly limited thereto. The distribution data may include only distribution data of inner physical quantities (hereinafter, sometimes referred to as "inner distribution data"), or may include both outer surface distribution data and inner distribution data.

[0035] The training dataset is preferably acquired based on multiple types of tires 11. The categories of the multiple types of tires 11 are not particularly limited. For example, from the viewpoint of improving the prediction accuracy of the trained model (machine learning model), the multiple types of tires 11 may be classified as the same category, or from the viewpoint of improving the versatility of the trained model (machine learning model), they may be classified as different categories. It is preferable that the multiple types of tires differ from each other in at least one of, for example, category, tire size, tread pattern, and internal structure (including carcass and belt layer). FIG. 4 is a flowchart showing an example of the processing procedure of the preparation step S1.

[0036] [Get data including pattern images] In the preparation step S1 of this embodiment, first, data including at least a pattern image of the tread portion 12 shown in Fig. 2 is acquired (step S11). In step S11 of this embodiment, data including at least a pattern image is acquired for a plurality of types of tires 11.

[0037] The pattern image is identified based on the tire 11 that is not in contact with the road surface (not shown). Therefore, unlike the contact patch image of Patent Document 1, the pattern image does not require the tire 11 to be inflated with internal pressure and in contact with the road surface. Therefore, the pattern image can be acquired in a shorter time than the contact patch image.

[0038] The pattern image is an image for identifying a pattern (tread pattern) engraved on the outer surface 12o of the tread portion 12. Such a pattern is constituted by, for example, grooves 15 (longitudinal grooves 15A and lateral grooves 15B), sipes 18, tie bars 19, etc. provided on the outer surface 12o.

[0039] 2, in the case where a tread pattern is formed by arranging a plurality of pitches P in the tire circumferential direction, a pattern image may be acquired for a specific (one or more) pitches P. Alternatively, pattern images for all pitches P may be acquired.

[0040] In this embodiment, a pattern image at one pitch P is acquired. In this case, the pitch P at which the pattern image is acquired is appropriately selected according to the distribution data estimated by the machine learning model; for example, it is preferable to select the pitch P with the highest ground contact pressure from among all pitches P. This allows pattern images of multiple types of tires 11 to be acquired under the same conditions. Furthermore, by acquiring a pattern image at one pitch P, an increase in the amount of data can be suppressed compared to, for example, when pattern images of all pitches P are acquired. Note that the pitch P with the highest ground contact pressure can be easily identified, for example, by measuring or calculating the ground contact pressures of all pitches P.

[0041] In the pattern image of this embodiment, the contour of the tread pattern (grooves 15, sipes 18, tie bars 19, etc.) is identified for each position (pixel) on the outer surface 12o of one pitch P. Each position on the outer surface 12o can be identified, for example, by a coordinate value in the tire circumferential direction (x-axis direction) and a coordinate value in the tire axial direction (y-axis direction). The contour of the tread pattern is also identified, for example, based on color information or shading information. The color information is used to identify the color. The shading information is used to identify the grayscale gradation. Therefore, the pattern image can be identified as a matrix (image data) in which color information or shading information is two-dimensionally arranged for each position on the outer surface 12o.

[0042] If the color information includes an R channel, a G channel, and a B channel, the pattern image is configured as a color image. If the color information includes only two values ​​(e.g., white and black), the pattern image is configured as a monochrome image. If the outline of the tread pattern is identified based on shading information, the pattern image is configured as a grayscale image.

[0043] The pattern image of this embodiment is configured as a grayscale image. As a result, the contours of the tread pattern (i.e., the contours of the grooves 15, sipes 18, tie bars 19, etc.) can be distinguished by shading in the pattern image. Furthermore, unlike, for example, a black and white image, the depths of the grooves 15, sipes 18, tie bars 19, etc. can be identified by shading in the grayscale image. Furthermore, compared to a color image, the grayscale image can suppress an increase in the amount of data. Note that the pattern image is not limited to a grayscale image, and may be configured as, for example, a color image or a black and white image.

[0044] Fig. 5 is a diagram showing an example of the training data set 23. Fig. 5 shows the training data set 23 of one tire 11 as a representative. Fig. 5 shows a pattern image 21A at one pitch P (in this example, the pitch with the highest ground contact pressure). The pattern image 21A is configured as a grayscale image.

[0045] In the pattern image 21A of this embodiment, the contours including the contact patch 14, grooves 15 (including longitudinal grooves 15A and lateral grooves 15B), sipes 18, and tie bars 19 can be identified for one pitch P. Furthermore, the depths of the grooves 15, sipes 18, etc. can be identified based on the shading of the pattern image 21A. These contours and depths make it possible to identify the uneven shape of the tread portion 12 (longitudinal grooves 15A, lateral grooves 15B, sipes 18, and tie bars 19). Such uneven shape affects the rigidity of the tread portion 12, and therefore has a strong correlation with the distribution data 22 (outer surface distribution data 22A in this example) output from the trained model.

[0046] The pattern image 21A can be acquired as appropriate. For example, the pattern image 21A can be acquired by actually photographing the tread portion 12 of the tire 11, or can be acquired from tire design data (CAD data), or can be acquired from a tire model modeled based on the design data. The tire model can be modeled as appropriate based on, for example, a known method (for example, the method described in JP 2023-084539 A).

[0047] In step S11 of this embodiment, pattern images 21A are acquired for each of a plurality of types of tires 11. Data 21 including these pattern images 21A is input to the learning dataset storage unit 7C shown in FIG.

[0048] [Get distribution data] Next, in the preparation step S1 of this embodiment, distribution data 22 (shown in FIG. 5) of physical quantities acting on the tread portion 12 shown in FIG. 2 is acquired (step S12). In step S12 of this embodiment, the distribution data 22 is acquired for a plurality of types of tires 11. In this embodiment, outer surface distribution data 22A is acquired as the distribution data 22.

[0049] The outer surface distribution data 22A is used to identify a physical quantity acting on the outer surface 12o of the tread portion 12 when the tread portion 12 comes into contact with the road surface (not shown). This physical quantity is not particularly limited as long as it acts on the outer surface 12o of the tire that is in contact with the road surface (not shown). In this embodiment, the physical quantity is, for example, wear energy acting on the outer surface 12o. Note that the physical quantity is not limited to wear energy, and may be, for example, shear force or slip amount. These physical quantities change depending on the uneven shape (rigidity) of the tread portion 12. Therefore, the outer surface distribution data 22A has a strong correlation with the pattern image 21A from which the uneven shape can be identified.

[0050] As shown in Fig. 2, in a case where a tread pattern is formed by arranging a plurality of pitches P in the tire circumferential direction, outer surface distribution data 22A may be acquired for the contact patch 14 of a specific (one or more) pitches P when the tire is in contact with a road surface (not shown). Alternatively, outer surface distribution data 22A may be acquired for each of the contact patches 14 of all pitches P when the tire is in contact with a road surface. The outer surface distribution data 22A of each pitch P is preferably acquired, for example, when the center position (not shown) of the pitch P at the tire equator C in the tire circumferential direction coincides with the contact center position (not shown) of the tire 11 at the tire equator C. This allows outer surface distribution data 22A of a plurality of types of tires 11 to be acquired under the same conditions.

[0051] As shown in FIG. 5, in this embodiment, outer surface distribution data 22A is acquired at one pitch P. In this case, the pitch P at which the outer surface distribution data 22A is acquired is appropriately selected depending on the data 21 (in this example, the pattern image 21A) input to the trained model. For example, it is preferable to select the pitch P with the highest ground contact pressure from among all pitches P. This allows outer surface distribution data 22A for multiple types of tires 11 to be acquired under the same conditions. Furthermore, by acquiring outer surface distribution data 22A at one pitch P, an increase in the amount of data can be suppressed compared to, for example, acquiring outer surface distribution data (not shown) for all pitches P.

[0052] In the outer surface distribution data 22A of this embodiment, a physical quantity (wear energy in this example) is identified for each position (coordinate) on the outer surface 12o (including the contact surface 14) of the pitch P that is in contact with the road surface (not shown). Each coordinate on the outer surface 12o can be identified by a coordinate value in the tire circumferential direction (x-axis direction) and a coordinate value in the tire axial direction (y-axis direction). Therefore, the outer surface distribution data can be identified as a matrix in which the physical quantities are two-dimensionally arranged for each position on the outer surface 12o.

[0053] The physical quantity (wear energy in this example) in this embodiment is specified by color information or shading information. In this embodiment, the color information or shading information is defined for each of a plurality of predetermined gradations. These gradations divide the range of values ​​that the physical quantity can take into predetermined intervals. For example, if the range that the physical quantity can take is 0 to 500 and the interval range is 50, 10 gradations are set. Ranges of the physical quantity are assigned to these gradations, for example, 0 to 50, 51 to 100, and so on.

[0054] In this embodiment, a gradation that includes an actual physical quantity is identified from the range of physical quantities assigned to multiple gradations. The actual physical quantity is then converted into color information or shading information for the identified gradation. As a result, the outer surface distribution data 22A can be configured as a contour diagram expressed with color information or shading information for the number of gradations according to the magnitude of the physical quantity. In such outer surface distribution data 22A, the physical quantity (in this example, wear energy) can be color-coded according to its magnitude. In the outer surface distribution data 22A in FIG. 5, the darker the color, the smaller the wear energy. On the other hand, the lighter the color, the larger the wear energy.

[0055] In the outer surface distribution data 22A of this embodiment, physical quantities acting on the contact patch 14, grooves 15 (including longitudinal grooves 15A and lateral grooves 15B), sipes 18, and tie bars 19 are identified for one pitch P. In this case, the physical quantities (wear energy) of the grooves 15, sipes 18, and tie bars 19 that do not contact the road surface (not shown) are zero, and are displayed in dark colors. Such outer surface distribution data 22A has a strong correlation with the data 21 (pattern image 21A) input to the trained model.

[0056] The outer surface distribution data 22A is preferably determined under predetermined driving conditions, in which the tire 11 (shown in FIG. 2) mounted on a rim (not shown) and inflated to a predetermined internal pressure is brought into contact with a road surface (not shown) at a predetermined camber angle and with a predetermined load applied. This allows the outer surface distribution data 22A to be determined under the same conditions. The camber angle is set appropriately depending on the mounting conditions of the tire 11 on the vehicle, and can be set to 0 degrees, for example. The driving conditions include, for example, the driving speed.

[0057] The rim (not shown) is exemplified as a genuine rim. A "genuine rim" is a rim defined for each tire by a standard system that includes the standard on which the tire 11 is based. Therefore, a genuine rim is, for example, a "standard rim" for JATMA, a "design rim" for TRA, or a "measuring rim" for ETRTO.

[0058] The internal pressure is exemplified as a normal internal pressure. The "normal internal pressure" is the air pressure determined for each tire by each standard in a standard system including the standard on which the tire 11 is based. Therefore, the normal internal pressure is, for example, the "maximum air pressure" in the case of JATMA, the maximum value listed in the table "TIRE LOAD LIMITS AT VARIOUS COLD INFLATION PRESSURES" in the case of TRA, and the "INFLATION PRESSURE" in the case of ETRTO.

[0059] The load is exemplified as a normal load. The "normal load" is the load determined for each tire by each standard in a standard system including the standard on which the tire 11 is based. Therefore, the normal load is, for example, "maximum load capacity" in the case of JATMA, the maximum value listed in the table "TIRE LOAD LIMITS AT VARIOUS COLD INFLATION PRESSURES" in the case of TRA, or "LOAD CAPACITY" in the case of ETRTO.

[0060] The outer surface distribution data 22A may be measured using, for example, a known wear energy measuring device (not shown) or a stress measuring device (not shown), or may be determined by performing a simulation using a computer 1 (shown in FIG. 1). In the simulation, the outer surface distribution data 22A may be calculated by bringing a tire model (not shown) that models the tire 11 into contact with a road surface model (not shown) that models the road surface (not shown) and running the tire model. Such a simulation may be performed as appropriate based on a known method (for example, the method described in JP 2023-084539 A).

[0061] In step S12 of this embodiment, outer surface distribution data 22A is acquired for a plurality of types of tires 11. The outer surface distribution data 22A (distribution data 22) is input to the learning dataset storage unit 7C shown in FIG.

[0062] [Create a training dataset] Next, in the preparation step S1 of this embodiment, a learning dataset 23 (shown in FIG. 5) is created (step S13). In step S13 of this embodiment, the data 21 (in this example, the pattern image 21A) and the distribution data 22 (in this example, the outer surface distribution data 22A) are combined for each of the plurality of types of tires 11, thereby creating a plurality of learning datasets 23.

[0063] In this embodiment, in step S2 described below, training datasets 23 of multiple types of tires 11 are used to train the machine learning model. This makes it possible to generate a trained model that can output distribution data 22 (outer surface distribution data 22A in this example) of the target tire when target tire data 21 (pattern image 21A in this example) is input. The multiple training datasets 23 are input to the training dataset storage unit 7C shown in FIG. 1.

[0064] [Training the machine learning model] Next, in the generation method of this embodiment, learning of a machine learning model is performed using the learning dataset 23 shown in Fig. 5 (step S2). In step S2, the machine learning model is trained so that when data 21 including at least the pattern image 21A of the target tire (in this example, only the pattern image 21A) is input, distribution data 22 of the target tire (in this example, only the outer surface distribution data 22A) is output.

[0065] In step S2 of this embodiment, first, the training dataset 23 (shown in FIG. 5) stored in the training dataset storage unit 7C shown in FIG. 1 and a training unit 8D included in the program unit 8 are loaded into the working memory 5C. The training unit 8D is a program for training a machine learning model using the training dataset 23. When the training unit 8D is executed by the processor 5A, the computer 1 (generation device 1B) can function as a means for training the machine learning model.

[0066] The machine learning model is not particularly limited as long as it can perform machine learning so that when data 21 including at least a pattern image 21A of a target tire is input, distribution data 22 of the target tire is output. The machine learning model of this embodiment is preferably a deep learning model equipped with a neural network. This enables the machine learning model to extract the feature quantities of the input data 21 and estimate the distribution data 22.

[0067] As described above, the data 21 (in this example, the pattern image 21A) input to the machine learning model is configured as image data (in this example, a grayscale image). In this case, the machine learning model (deep learning model) is preferably configured as a convolutional neural network (CNN) capable of extracting features of the image data.

[0068] The convolutional neural network (CNN) can be appropriately selected from known ones. When a contour diagram of a physical quantity is output as distribution data 22 output from the machine learning model as in this embodiment, color information or shading information of the physical quantity can be treated as, for example, a class label for semantic segmentation. Therefore, it is preferable that the machine learning model be a model that can extract features from input image data and classify the features into class labels. An example of such a model is U-NET. FIG. 6 is a conceptual diagram showing an example of a machine learning model 24A (trained model 24B).

[0069] The machine learning model 24A of this embodiment is configured based on the well-known U-NET. This machine learning model 24A includes an encoder 25 and a decoder 26. The encoder 25 convolves data 21 (in this example, a pattern image 21A) input as image data across multiple layers. This allows features of the data 21 to be extracted. Meanwhile, the decoder 26 deconvolves the extracted features (the reverse process of convolution). This outputs a probability map classified into class labels of physical quantities from the input data 21. This probability map allows the physical quantity (in this example, wear energy) of the gradation defined by the color information or shading information to be identified based on the color information or shading information identified by the class label. Therefore, the probability map can be configured as distribution data 22 (in this example, outer surface distribution data 22A).

[0070] In deconvolution, the feature map 27 of the encoder 25 is linked to the feature map 27 of the decoder 26 within the same layer of the U-NET. This allows the data (in this example, the pattern image 21A) to be properly understood, improving the classification accuracy (in this example, the wear energy prediction accuracy) of the distribution data 22 (in this example, the outer surface distribution data 22A) into class labels.

[0071] In step S2 of this embodiment, first, data 21 (in this example, pattern image 21A) included in one training dataset 23 shown in FIG. 5 among multiple training datasets 23 (not shown) is input to the machine learning model 24A. As a result, distribution data 22 (in this example, outer surface distribution data 22A) estimated by the machine learning model 24A can be output. Next, in step S2 of this embodiment, an error between the output distribution data (estimated data) 22 and the distribution data (ground truth data) 22 included in the training dataset 23 shown in FIG. 5 is calculated. Then, weights in each layer of the machine learning model (U-NET) 24A are adjusted to minimize this error. In this way, a series of steps from inputting the training dataset 23 to adjusting the function is performed for each of the multiple training datasets 23, thereby optimizing the machine learning model 24A.

[0072] In step S2 of this embodiment, a machine learning model 24A is optimized using a plurality of learning datasets 23 (not shown). As a result, when data 21 including at least a pattern image 21A of the target tire is input, a trained model 24B capable of outputting distribution data 22 of the target tire (in this example, outer surface distribution data 22A) is generated. Such trained model 24B makes it possible to estimate the distribution data 22 of the target tire without requiring the experience or intuition of an expert. The trained model 24B is input to the model storage unit 7D.

[0073] As described above, the pattern image 21A shown in FIG. 5 can identify the uneven shape (e.g., longitudinal grooves 15A, lateral grooves 15B, sipes 18, tie bars 19, etc.) formed in the tread portion 12 shown in FIG. 2. Therefore, the pattern image 21A has a strong correlation with the distribution data 22 (in this example, outer surface distribution data 22A) shown in FIG. 5. By inputting data 21 including such pattern image 21A into the machine learning model 24A shown in FIG. 6, feature quantities for estimating the distribution data 22 can be effectively extracted. Therefore, the generation method (generation device 1B) of this embodiment makes it possible to generate a trained model 24B that can accurately estimate the distribution data 22.

[0074] In this embodiment, the pattern image 21A is configured as image data (in this example, a grayscale image), and the outer surface distribution data 22A is configured as a contour diagram. The gradation of the physical quantities in this contour diagram can be treated as class labels for semantic segmentation, so a trained model 24B can be easily generated based on a conventional U-NET. This allows the trained model 24B to be created in a short time.

[0075] [Tire performance prediction method (first embodiment)] Next, an example of a processing procedure for a tire performance prediction method will be described. The prediction method of this embodiment uses a trained model 24B (shown in FIG. 6) generated based on the generation method shown in FIGS. 3 and 4.

[0076] 7 is a flowchart showing an example of a processing procedure of a tire performance prediction method. Each step of the prediction method of this embodiment is executed by one or more processors 5A included in the prediction device 1A (computer 1) shown in FIG.

[0077] [Enter data including the pattern image of the target tire] In the prediction method of this embodiment, first, data 21 including at least a pattern image 21A of a target tire is input to the trained model 24B shown in Fig. 6 (step S3). This trained model 24B has been trained by machine learning so that when data (in this example, the pattern image 21A) is input, the trained model 24B outputs distribution data 22 (in this example, outer surface distribution data 22A).

[0078] In step S3 of this embodiment, first, the trained model 24B (shown in FIG. 6) input to the model storage unit 7D shown in FIG. 1 and the input unit 8A included in the program unit 8 are loaded into the working memory 5C. The input unit 8A is a program for inputting data 21 including at least a pattern image 21A of a target tire to the trained model 24B. When the input unit 8A is executed by the processor 5A, the computer 1 (prediction device 1A) can function as a means for inputting the data 21 to the trained model 24B.

[0079] In step S3 of this embodiment, first, a pattern image 21A of the target tire is acquired. In this embodiment, the pattern image 21A of the target tire shown in FIG. 6 is acquired based on the processing procedure in step S11 of the preparation step S1 shown in FIG. 4. The acquired pattern image 21A of the target tire is input to the target data storage unit 7A (shown in FIG. 1). Furthermore, in step S3 of this embodiment, the pattern image 21A of the target tire is input as image data to the trained model 24B. This makes it possible to output distribution data 22 of the target tire (in this example, outer surface distribution data 22A) from the trained model 24B.

[0080] [Output distribution data for target tires] Next, in the prediction method of this embodiment, the distribution data 22 of the target tire is output from the trained model 24B (step S4).

[0081] In step S4 of this embodiment, first, the derivation unit 8B included in the program unit 8 shown in Fig. 1 is loaded into the working memory 5C. The derivation unit 8B is a program for outputting distribution data of the target tire from the trained model 24B. Execution of this derivation unit 8B by the processor 5A causes the computer 1 to function as a means for outputting distribution data 22 of the target tire.

[0082] In this embodiment, in step S3, as shown in Fig. 6, data 21 including at least a pattern image 21A of the target tire (in this example, only the pattern image 21A) is input to the trained model 24B. As a result, in step S4, distribution data 22 of the target tire (in this example, outer surface distribution data 22A) can be output from the trained model 24B. Therefore, the prediction method of this embodiment makes it possible to estimate the distribution data 22 of the target tire without requiring the experience or intuition of a skilled person.

[0083] The output distribution data 22 (in this example, outer surface distribution data 22A) is stored in the target distribution data storage unit 7B shown in Fig. 1. In addition, in step S4, the output distribution data 22 may be output (displayed) to an output device 4 including, for example, a display device. This allows an operator or the like to understand the distribution data 22 and further enables evaluation of the performance of the target tire.

[0084] [Evaluate distribution data of target tires] Next, in the prediction method of this embodiment, the quality of the distribution data 22 (shown in FIG. 6) of the target tire is evaluated (step S5). The quality of the distribution data 22 may be evaluated by the prediction device 1A (computer 1) shown in FIG. 1 or by an operator or the like.

[0085] In step S5 of this embodiment, first, the distribution data 22 (in this example, outer surface distribution data 22A) input to the target distribution data storage unit 7B shown in Fig. 1 and an evaluation unit 8E included in the program unit 8 are read into the working memory 5C. This evaluation unit 8E is a program for determining whether the distribution data 22 of the target tire is good or not based on the output distribution data 22. Execution of this evaluation unit 8E by the processor 5A allows the computer 1 to function as a means for evaluating the distribution data 22.

[0086] The quality of the distribution data 22 can be evaluated as appropriate. In the case where the distribution data 22 of this embodiment is outer surface distribution data 22A of wear energy, for example, if the maximum value of the wear energy is less than a predetermined threshold, it is determined to be good. The threshold can be set as appropriate, for example, depending on the wear resistance performance required of the target tire.

[0087] If the distribution data 22 of the target tire is determined to be good ("Yes" in step S5), the target tire is manufactured based on the data 21 (pattern image 21A) of the target tire input to the trained model 24B (step S6). On the other hand, if the distribution data 22 of the target tire is determined to be bad ("No" in step S5), at least a part of the tread pattern of the tread portion 12 of the target tire is changed (step S7), and steps S3 to S5 are performed again. When the outer surface distribution data 22A of wear energy is output as in this embodiment, the portion with relatively high wear energy may be changed in step S7. Such changes may be performed by an operator or the like, or may be performed by the computer 1 based on a known optimization method or the like. This allows the tire 11 with reduced wear energy to be reliably designed and manufactured.

[0088] [Method for generating trained model (second embodiment)] In the above embodiments, as shown in Fig. 6, only the outer surface distribution data 22A has been exemplified as the distribution data 22 output from the trained model 24B (machine learning model 24A), but this is not limited to such an embodiment. For example, only the inner distribution data (not shown) may be output as the distribution data 22, or both the outer surface distribution data 22A and the inner distribution data may be output. In this embodiment, an embodiment in which only the inner distribution data is output as the distribution data 22 is exemplified.

[0089] In this embodiment, in step S12 (shown in FIG. 4) of acquiring distribution data 22, internal distribution data is acquired for each of a plurality of types of tires 11. The internal distribution data is used to identify physical quantities acting inside the tread. Such physical quantities are not particularly limited as long as they act inside the tread. The physical quantities in this embodiment include strains and / or stresses acting inside the tread. Such strains and / or stresses change depending on the uneven shape (rigidity) of the tread portion 12. Therefore, the internal distribution data has a strong correlation with a pattern image 21A (shown in FIG. 6) from which the uneven shape can be identified.

[0090] In this embodiment, similar to the outer surface distribution data 22A shown in Fig. 6, internal distribution data may be obtained for the contact patch 14 of specific (one or more) pitches P shown in Fig. 2 when the pitches P are in contact with the road surface (not shown). Alternatively, internal distribution data may be obtained for each of the contact patches 14 of all pitches P when the pitches P are in contact with the road surface. In this case, it is preferable that the internal distribution data of each pitch P is obtained, for example, when the center position (not shown) of the pitch P at the tire equator C in the tire circumferential direction coincides with the contact center position (not shown) of the tire 11 at the tire equator C.

[0091] In this embodiment, internal distribution data (not shown) can be acquired for one pitch P. In this case, the pitch P for which internal distribution data is acquired is appropriately selected according to the data 21 (in this example, the pattern image 21A) input to the trained model 24B shown in FIG. 6, and, for example, it is preferable to select the pitch P with the highest ground contact pressure from among all pitches P. This allows internal distribution data for multiple types of tires 11 to be acquired under the same conditions, and can prevent an increase in the amount of data.

[0092] In the internal distribution data of this embodiment, physical quantities (in this example, strain and / or stress) are identified for each position (coordinate) inside the tread radially inward of the outer surface 12o of the pitch P that is in contact with the road surface (not shown). Each coordinate inside the tread can be identified, for example, by a coordinate value in the tire circumferential direction (x-axis direction) and a coordinate value in the tire axial direction (y-axis direction). Therefore, the internal distribution data can be identified as a matrix in which physical quantities are two-dimensionally arranged for each position inside the tread.

[0093] In this embodiment, the physical quantity (in this example, strain and / or stress) is converted into color information or shading information, similarly to the outer surface distribution data 22A. As a result, the internal distribution data can be configured as a contour diagram expressed by color information or shading information for the number of gradations, depending on the magnitude of the physical quantity.

[0094] The internal distribution data of this embodiment identifies physical quantities acting within the tread on the radially inner side of the contact patch 14, grooves 15 (including longitudinal grooves 15A and lateral grooves 15B), and sipes 18 for one pitch P. Such internal distribution data has a strong correlation with the data 21 (pattern image 21A shown in FIG. 6) input to the trained model 24B.

[0095] As with the outer surface distribution data 22A, the internal distribution data is preferably determined under predetermined driving conditions, in which the tire 11 is mounted on a rim, inflated to a predetermined internal pressure, and placed in contact with the road surface (not shown) at a predetermined camber angle and with a predetermined load applied. This allows the internal distribution data to be determined under the same conditions. Details of the rim, internal pressure, camber angle, load, etc. are as described above.

[0096] The internal distribution data may be determined based on the calculation results of a simulation using a computer 1 (shown in FIG. 1), for example. In the simulation, the internal distribution data can be calculated by bringing a tire model (not shown) that models the tire 11 into contact with a road surface model (not shown) that models the road surface (not shown). Such a simulation can be appropriately performed based on a known method (for example, the method described in JP 2022-135085 A).

[0097] In step S12 of this embodiment, internal distribution data (not shown) is acquired for multiple types of tires 11. The internal distribution data (distribution data 22) is input to the learning dataset storage unit 7C shown in FIG.

[0098] Next, in the generation method of this embodiment, as in the previous embodiments, a training dataset 23 (shown in FIG. 5) is created in step S13 shown in FIG. 4. In this step S13, a training dataset 23 is created for each of a plurality of types of tires 11, combining data 21 (in this example, pattern image 21A) and distribution data 22 (in this example, internal distribution data (not shown)). Then, in step S2 shown in FIG. 3, a machine learning model 24A is optimized (trained) using the plurality of training datasets 23. As a result, when data 21 (pattern image 21A) of a target tire is input, a trained model 24B capable of outputting internal distribution data of the target tire is generated. Such a trained model 24B makes it possible to estimate the internal distribution data of the target tire without the need for the experience or intuition of an expert.

[0099] As described above, the pattern image 21A shown in FIG. 5 can identify the uneven shape of the tread portion 12, and therefore has a strong correlation with the internal distribution data (not shown). This allows the generation of a trained model 24B that can accurately estimate the internal distribution data. Furthermore, in this embodiment, the internal distribution data is configured as a contour diagram, so the trained model 24B can be easily generated based on a conventional U-NET. Therefore, the trained model 24B can be created in a short time.

[0100] [Tire performance prediction method (second embodiment)] In the prediction method of this embodiment, as in the previous embodiments, a trained model 24B (shown in FIG. 6) generated based on the generation method shown in FIG. 4 is used.

[0101] In the prediction method of this embodiment, as in the previous embodiments, in step S3 shown in Fig. 7, data 21 including at least a pattern image 21A of the target tire is input to the trained model 24B shown in Fig. 6. This makes it possible to output distribution data 22 of the target tire (in this example, internal distribution data (not shown)) from the trained model 24B.

[0102] Next, in the prediction method of this embodiment, as in the previous embodiments, in step S4 shown in FIG. 7, distribution data 22 of the target tire (in this example, internal distribution data (not shown)) is output from the trained model 24B shown in FIG. 6. As a result, the prediction method of this embodiment makes it possible to estimate the internal distribution data of the target tire without requiring the experience or intuition of a skilled person. Then, if it is determined that the internal distribution data is not good ("No" in step S5), at least a part of the tread pattern of the tread portion 12 of the target tire is changed in step S7 based on the physical quantities (strain and / or stress) of the internal distribution data. This makes it possible to reliably design and manufacture a tire 11 with improved durability performance, rolling resistance performance, etc.

[0103] [Method for generating trained model (third embodiment)] In the above embodiments, as shown in Fig. 6, only the pattern image 21A has been exemplified as the data 21 input to the trained model 24B (machine learning model 24A), but the present invention is not limited to this. In addition to the pattern image 21A, the data 21 may further include, for example, a distance distribution.

[0104] In the generation method of this embodiment, a pattern image 21A of the tread portion 12 and a distance distribution are acquired in step S11 (shown in FIG. 4) of acquiring data 21. The pattern image 21A can be acquired in the same procedure as in the previous embodiments.

[0105] The distance distribution is intended to identify the distance from the road surface (not shown) on which the tire 11 shown in Fig. 2 is in contact with the ground to the outer surface 12o of the tread portion 12 (including the contact surface 14, grooves 15, and tie bars 19). Such a distance distribution changes depending on the uneven shape of the tread portion 12, and therefore has a strong correlation with the distribution data 22 (outer surface distribution data 22A and internal distribution data) output from the trained model 24B shown in Fig. 6.

[0106] In the case of a pattern in which a plurality of pitches P are arranged in the tire circumferential direction, such as the tread portion 12 shown in Fig. 2, the distribution of distances when the tire comes into contact with the road surface (not shown) may be acquired for specific (one or more) pitches P. Alternatively, the distribution of distances when the tire comes into contact with the road surface may be acquired for each of all pitches P. In this case, the distribution of distances for each pitch P is preferably acquired, for example, in the tire circumferential direction, in a state in which the center position (not shown) of the pitch P at the tire equator C coincides with the contact center position (not shown) of the tire 11 at the tire equator C.

[0107] In this embodiment, the distance distribution is obtained for one pitch P. In this case, the pitch P for which the distance distribution is obtained is appropriately selected according to the distribution data 22 estimated by the machine learning model 24A shown in Fig. 6, and, for example, it is preferable to select the pitch P with the highest ground contact pressure from among all the pitches P. This makes it possible to obtain the distance distributions of multiple types of tires 11 under the same conditions and to suppress an increase in the amount of data.

[0108] FIG. 8 is a partial cross-sectional view of a tire 11 in contact with a road surface 13. In the distance distribution of this embodiment, for each position (coordinate) of the outer surface 12o of the pitch P in contact with the road surface 13, a distance D from the road surface 13 to the outer surface 12o is specified. Each coordinate of the outer surface 12o can be specified, for example, by a coordinate value in the tire circumferential direction (x-axis direction) and a coordinate value in the tire axial direction (y-axis direction). Furthermore, the distance D from the road surface 13 to the outer surface 12o can be specified by a coordinate value in a direction perpendicular to the road surface 13 (z-axis direction) with the road surface 13 set to zero. Therefore, the distance distribution can be specified as a matrix in which the distance D from the road surface 13 to the outer surface 12o is two-dimensionally arranged for each position on the outer surface 12o. In such a distance distribution, the distance D at the contact patch 14 in contact with the road surface 13 is zero. With such a distance distribution, the depth of grooves 15, sipes 18, etc. can be identified with higher accuracy compared to pattern image 21A (shown in Figure 5) in which the depth of grooves 15, sipes 18, etc. (shown in Figure 2) is identified by shades of gray.

[0109] In this embodiment, the distance D is converted into a combination of the values ​​of the R channel, G channel, and B channel (i.e., in the range of 0 to 16777215) used in a color image. This allows the distance distribution to be configured as a color image. Note that the distance distribution is not limited to a color image, and may be configured as, for example, a grayscale image.

[0110] FIG. 9 is a diagram showing an example of a training data set 23 according to another embodiment of the present invention. In FIG. 9, the training data set 23 for one tire 11 is shown as a representative example. In this training data set 23, a distance distribution 21B for one pitch P (in this example, the pitch with the highest ground contact pressure) is shown. This distance distribution 21B is configured as a color image, and is color-coded according to the magnitude of the distance D from the road surface 13 to the outer surface 12o of the tread portion 12 shown in FIG. 8. In the distance distribution 21B in FIG. 9, the darker the color (the bluer the portion), the smaller the distance D. On the other hand, the lighter the color (the redder the portion), the larger the distance D.

[0111] In the distance distribution 21B of this embodiment, for one pitch P, the distance D from the road surface 13 to the outer surface 12o shown in FIG. 8 can be identified for the contact patch 14, the grooves 15 (including the longitudinal grooves 15A and lateral grooves 15B), the sipes 18, and the tie bars 19. This makes it possible to identify the uneven shape (longitudinal grooves 15A, lateral grooves 15B, sipes 18, tie bars 19, etc.) of the tread portion 12 that has been deformed by contact with the road surface 13. Such uneven shape affects the rigidity of the tread portion 12, and therefore has a strong correlation with the distribution data 22 (outer surface distribution data 22A and internal distribution data) output from the trained model 24B.

[0112] Similar to the outer surface distribution data 22A, the distance distribution 21B is preferably determined when the tire 11 is mounted on a rim, inflated to a predetermined (predetermined) internal pressure, and placed in contact with the road surface 13 (shown in FIG. 8) at a predetermined camber angle and with a predetermined load applied. This allows the distance distribution 21B shown in FIG. 9 to be determined under the same conditions. The details of the rim, internal pressure, camber angle, and load are as described above.

[0113] Distance distribution 21B may be determined, for example, by CT scanning an image of tire 11 (shown in FIG. 8) that is actually placed in contact with road surface 13, or by performing a simulation using computer 1 (shown in FIG. 1). In the simulation, distance distribution 21B may be calculated by bringing a tire model (not shown) that models tire 11 into contact with a road surface model (not shown) that models road surface 13. Such a simulation may be performed as appropriate based on a known method (for example, the method described in JP 2023-084539 A).

[0114] In step S11 of this embodiment, distance distributions 21B are obtained for a plurality of types of tires 11. These distance distributions 21B are input to the learning data set storage unit 7C shown in FIG.

[0115] Next, in the generation method of this embodiment, in step S13 shown in FIG. 4, a training dataset 23 is created as shown in FIG. 9. In this step S13, a training dataset 23 is created for each of a plurality of types of tires 11, combining data 21 (in this example, a pattern image 21A and a distance distribution 21B) and distribution data 22 (in this example, outer surface distribution data 22A). Then, in step S2 shown in FIG. 3, a machine learning model 24A is optimized (trained) using the plurality of training datasets 23. As a result, when the data 21 (pattern image 21A and a distance distribution 21B) of the target tire is input, a trained model 24B capable of outputting outer surface distribution data 22A of the target tire is generated. Such a trained model 24B makes it possible to estimate the outer surface distribution data 22A of the target tire without the need for the experience or intuition of a skilled person.

[0116] In this embodiment, as shown in FIG. 9, in addition to the pattern image 21A, a distance distribution 21B is input to the machine learning model 24A (shown in FIG. 6). The pattern image 21A and the distance distribution 21B are related to each other in that they can identify the uneven shape of the tread portion 12, and furthermore, they are strongly correlated with the distribution data 22 (external surface distribution data 22A and internal distribution data) output from the trained model 24B. Therefore, the trained model 24B that can accurately estimate the distribution data 22 can be generated using the pattern image 21A and the distance distribution 21B. Furthermore, in this embodiment, since the distance distribution 21B is configured as a color image, the trained model 24B can be easily generated based on a conventional U-NET. Therefore, the trained model 24B can be created in a short time.

[0117] [Tire performance prediction method (third embodiment)] In the prediction method of this embodiment, as in the previous embodiments, a trained model 24B (shown in FIG. 6) generated based on the generation method shown in FIG. 4 is used.

[0118] In the prediction method of this embodiment, as in the previous embodiments, in step S3 shown in Fig. 7, data 21 including a pattern image 21A of the target tire and a distance distribution 21B shown in Fig. 9 is input to the trained model 24B shown in Fig. 6. This makes it possible to output distribution data 22 of the target tire (in this example, outer surface distribution data 22A) from the trained model 24B.

[0119] Next, in the prediction method of this embodiment, as in the previous embodiments, in step S4 shown in FIG. 7, distribution data 22 of the target tire (in this example, outer surface distribution data 22A) is output from the trained model 24B shown in FIG. 6. As a result, the prediction method of this embodiment makes it possible to estimate the distribution data 22 of the target tire without requiring the experience or intuition of a skilled person. Then, if it is determined that the distribution data 22 is not satisfactory ("No" in step S5), at least a part of the tread pattern of the tread portion 12 of the target tire is changed in step S7 based on the physical quantities of the distribution data 22. As a result, a tire 11 having desired performance (for example, wear resistance performance, durability performance, and rolling resistance performance) can be reliably designed and manufactured.

[0120] [Method for generating trained model (fourth embodiment)] In the embodiments described above, the pattern image 21A and the distance distribution 21B shown in FIG. 9 have been exemplified as data 21 input to the trained model 24B (machine learning model 24A) shown in FIG. 6, but the present invention is not limited to such an embodiment. In addition to the pattern image 21A, the data 21 may further include, for example, a ground pressure distribution. Note that the data in this embodiment may also include the distance distribution 21B of the embodiments described above.

[0121] In the generation method of this embodiment, a pattern image 21A of the tread portion 12 and the distribution of the contact patch are acquired in step S11 (shown in FIG. 4) of acquiring data 21. The pattern image 21A can be acquired in the same procedure as in the previous embodiments.

[0122] The distribution of contact pressure is used to identify the contact pressure of the contact surface 14 that comes into contact with the road surface 13 shown in Fig. 8. Such a distribution of contact pressure changes depending on the uneven shape of the tread portion 12, and therefore has a strong correlation with the distribution data 22 (external surface distribution data 22A and internal distribution data) output from the trained model 24B shown in Fig. 6.

[0123] In the case of a tread portion 12 having a pattern in which a plurality of pitches P are arranged in the tire circumferential direction, as in the case of the tread portion 12 shown in Fig. 2, the distribution of contact pressure when the tire is brought into contact with the road surface 13 shown in Fig. 8 may be acquired for a specific (one or more) pitches P. Alternatively, the distribution of contact pressure when the tire is brought into contact with the road surface 13 may be acquired for each of all pitches P. In this case, it is preferable that the distribution of contact pressure for each pitch P is acquired in the tire circumferential direction, for example, in a state in which the center position (not shown) of the pitch P at the tire equator C coincides with the contact center position (not shown) of the tire 11 at the tire equator C.

[0124] In this embodiment, it is preferable to acquire the distribution of ground contact pressure at two or more pitches P. This is because the distribution of ground contact pressure at each pitch P in the contact patch may differ depending on the profile of the tire 11, and by considering the distribution of ground contact pressure at these pitches P, it is possible to take the profile of the tire 11 into consideration. In this case, the pitches P from which the distribution of ground contact pressure is acquired are appropriately selected according to the distribution data 22 estimated by the machine learning model 24A shown in FIG. 6. For example, it is preferable to select the pitch P with the highest ground contact pressure and the pitch P with the second highest ground contact pressure from all the pitches P. This makes it possible to acquire the distribution of ground contact pressure for multiple types of tires 11 under the same conditions, and to suppress an increase in the amount of data.

[0125] In the distribution of contact pressure in this embodiment, as shown in Fig. 8, the contact pressure is specified for each position (coordinate) on the outer surface 12o (including the contact surface 14) of the pitch P that is in contact with the road surface 13. Each coordinate on the outer surface 12o can be specified by, for example, a coordinate value in the tire circumferential direction (x-axis direction) and a coordinate value in the tire axial direction (y-axis direction). Therefore, the distribution of contact pressure can be specified as a matrix in which the contact pressures are two-dimensionally arranged for each position on the outer surface 12o.

[0126] In this embodiment, the ground pressure is converted into a combination of the values ​​of the R channel, G channel, and B channel (i.e., in the range of 0 to 16777215) that are handled as a color image. As a result, the ground pressure distribution is configured as a color image. Note that the ground pressure distribution is not limited to a color image, and may be configured as, for example, a grayscale image.

[0127] FIG. 10 is a diagram showing an example of a training data set 23 according to yet another embodiment of the present invention. FIG. 10 shows a representative training data set 23 for one tire 11. This training data set 23 shows contact pressure distributions 21C, 21C at two pitches P (in this example, the pitch with the highest contact pressure and the pitch with the second highest contact pressure). These contact pressure distributions 21C, 21C are configured as color images, and are color-coded according to the magnitude of the contact pressure. In the contact pressure distributions 21C, 21C in FIG. 9, the darker the color (blue), the lower the contact pressure. On the other hand, the lighter the color (red), the higher the contact pressure.

[0128] In the ground contact pressure distribution 21C of this embodiment, the ground contact pressure can be determined for the ground contact patch 14, the grooves 15 (including the longitudinal grooves 15A and lateral grooves 15B), the sipes 18, and the tie bars 19 at one pitch P. In this case, the ground contact pressure of the grooves 15, sipes 18, and tie bars 19 that do not contact the road surface 13 is zero. Such ground contact pressure distribution 21C changes depending on the uneven shape of the tread portion 12, and therefore has a strong correlation with the distribution data 22 (external surface distribution data 22A and internal distribution data (not shown)) output from the trained model 24B shown in FIG. 6.

[0129] Similar to the outer surface distribution data 22A, the contact pressure distribution 21C is preferably determined when the tire 11, mounted on a rim and inflated to a predetermined (predetermined) internal pressure, is brought into contact with the road surface 13 (shown in FIG. 8) at a predetermined camber angle and under a predetermined load. This allows the contact pressure distribution 21C shown in FIG. 10 to be determined under the same conditions. The details of the rim, internal pressure, camber angle, and load are as described above.

[0130] The contact pressure distribution 21C may be measured using a known contact pressure distribution measuring device, or may be determined by performing a simulation using a computer 1 (shown in FIG. 1). In the simulation, the contact pressure distribution may be calculated by bringing a tire model (not shown) that models the tire 11 into contact with a road surface model (not shown) that models the road surface 13 shown in FIG. 8. Such a simulation may be performed as appropriate based on a known method (for example, the method described in JP 2014-149792 A).

[0131] In step S11 of this embodiment, contact pressure distributions 21C are obtained for a plurality of types of tires 11. These contact pressure distributions 21C are input to the learning data set storage unit 7C shown in FIG.

[0132] Next, in the generation method of this embodiment, in step S13 shown in FIG. 4, a training dataset 23 is created as shown in FIG. 10. In this step S13, a training dataset 23 is created for each of a plurality of types of tires 11, combining data 21 (in this example, a pattern image 21A and a contact pressure distribution 21C) and distribution data 22 (in this example, outer surface distribution data 22A). Then, in step S2 shown in FIG. 3, a machine learning model 24A is optimized (trained) using the plurality of training datasets 23. As a result, when data 21 (in this example, the pattern image 21A and the contact pressure distribution 21C) of a target tire is input, a trained model 24B capable of outputting outer surface distribution data 22A of the target tire is generated. Using such a trained model 24B, a trained model 24B capable of estimating outer surface distribution data 22A of the target tire can be generated without the need for the experience or intuition of a skilled person.

[0133] In this embodiment, as shown in FIG. 10, in addition to the pattern image 21A, a ground contact pressure distribution 21C is input to the machine learning model 24A (shown in FIG. 6). The pattern image 21A and the ground contact pressure distribution 21C are related to each other in that they can identify the uneven shape of the tread portion 12, and furthermore, they are strongly correlated with the distribution data 22 (external surface distribution data 22A and internal distribution data (not shown)) output from the trained model 24B. Therefore, the trained model 24B that can accurately estimate the distribution data 22 can be generated from the pattern image 21A and the ground contact pressure distribution 21C. Furthermore, in this embodiment, because the ground contact pressure distribution 21C is configured as a color image, the trained model 24B can be easily generated based on a conventional U-NET. Therefore, the trained model 24B can be created in a short time.

[0134] [Tire performance prediction method (fourth embodiment)] In the prediction method of this embodiment, as in the previous embodiments, a trained model 24B (shown in FIG. 6) generated based on the generation method shown in FIG. 4 is used.

[0135] In the prediction method of this embodiment, as in the previous embodiments, in step S3 shown in Fig. 7, data 21 including a pattern image 21A of the target tire and a contact pressure distribution 21C shown in Fig. 10 is input to the trained model 24B shown in Fig. 6. This makes it possible to output distribution data 22 of the target tire (in this example, outer surface distribution data 22A) from the trained model 24B.

[0136] Next, in the prediction method of this embodiment, as in the previous embodiments, in step S4 shown in FIG. 7, distribution data 22 of the target tire (in this example, outer surface distribution data 22A) is output from the trained model 24B shown in FIG. 6. As a result, the prediction method of this embodiment makes it possible to estimate the distribution data 22 of the target tire without requiring the experience or intuition of a skilled person. Then, if it is determined that the distribution data 22 is not satisfactory ("No" in step S5), at least a part of the tread pattern of the tread portion 12 of the target tire is changed in step S7 based on the physical quantities of the distribution data 22. As a result, a tire 11 having desired performance (for example, wear resistance performance, durability performance, and rolling resistance performance) can be reliably designed and manufactured.

[0137] [Method for generating trained model (fifth embodiment)] In the embodiments described above, the pattern image 21A and distance distribution 21B shown in FIG. 9 and the ground contact pressure distribution 21C shown in FIG. 10 have been exemplified as data 21 input to the trained model 24B (machine learning model 24A), but the present invention is not limited to these examples. In addition to the pattern image 21A, the data 21 may further include, for example, condition data. Note that the data 21 of this embodiment may further include the distance distribution 21B and the ground contact pressure distribution 21C of the embodiments described above.

[0138] In the generation method of this embodiment, a pattern image 21A and condition data of the tread portion 12 are acquired in step S11 (shown in FIG. 4) of acquiring data 21. The pattern image 21A can be acquired in the same procedure as in the previous embodiments.

[0139] The condition data relates to the design conditions and / or evaluation conditions of the tire. The condition data in this embodiment includes, for example, at least one of the tread gauge, category, load, internal pressure, aspect ratio, tire width, outer diameter, complex modulus of elasticity E* of the tread portion, rim width, running speed, and camber angle. These design conditions and / or evaluation conditions affect the physical quantities acting on the outer surface 12o of the tread portion 12 and the inside of the tread, and therefore have a strong correlation with the distribution data 22 (the outer surface distribution data 22A and the internal distribution data) output from the trained model 24B shown in FIG. 6.

[0140] 11 is a conceptual diagram showing an example of a machine learning model 24A (trained model 24B) to which condition data 21D is input. The condition data 21D in FIG. 11 is composed of numerical data specifying the tread gauge, category, load, internal pressure, aspect ratio, tire width, outer diameter, and complex modulus of elasticity E* of the tread portion for one tire 11. In step S11 of this embodiment, the condition data 21D is acquired for multiple types of tires 11. The condition data 21D is input to the training dataset storage unit 7C shown in FIG. 1.

[0141] Next, in the generation method of this embodiment, a training dataset 23 is created in step S13 shown in FIG. 4. In this step S13, a training dataset is created for each of a plurality of types of tires 11, combining data 21 (in this example, a pattern image 21A and condition data 21D) and distribution data 22 (in this example, outer surface distribution data 22A), as shown in FIG. 11. Then, in step S2 shown in FIG. 3, a machine learning model 24A is optimized (trained) using the plurality of training datasets 23. As a result, when data 21 (in this example, the pattern image 21A and condition data 21D) of a target tire is input, a trained model 24B capable of outputting outer surface distribution data 22A of the target tire is generated. Such a trained model 24B makes it possible to estimate the outer surface distribution data 22A of the target tire without the need for the experience or intuition of a skilled person.

[0142] As described above, the condition data 21D is numerical data specifying the tread gauge, etc., and therefore differs from the image data of the pattern image 21A and distance distribution 21B shown in FIG. 9 and the ground contact pressure distribution 21C shown in FIG. 10. Therefore, the condition data 21D cannot be directly input as image data to the machine learning model 24A (U-NET) shown in FIG. 11. In this embodiment, the condition data 21D is input to the lowest-level feature map 27 of the U-NET encoder 25. This lowest-level feature map 27 contains features extracted by convolving the pattern image 21A. The features (feature map 27) obtained by adding the condition data 21D to the features of the pattern image 21A are deconvolved by the decoder 26. As a result, a probability map (distribution data 22) classified into class labels of physical quantities can be output based on the pattern image 21A and the condition data 21D. In this embodiment, the condition data 21D is connected to the U-NET via a multilayer perceptron (not shown). This makes it possible to realize a multimodal model that can handle different data (image data and numerical data) together.

[0143] In this embodiment, in addition to the pattern image 21A, condition data 21D are input to the machine learning model 24A. The pattern image 21A and the condition data 21D are related to each other in that they affect the physical quantities acting on the outer surface and the inside of the tread portion 12. Furthermore, the pattern image 21A and the condition data 21D are strongly correlated with the distribution data 22 (the outer surface distribution data 22A and the internal distribution data (not shown)) output from the trained model 24B. Therefore, the trained model 24B capable of accurately estimating the distribution data 22 can be generated using the pattern image 21A and the condition data 21D. Furthermore, in this embodiment, the condition data 21D can be input to the feature map 27 at the lowest layer of the U-NET encoder 25, so that the trained model 24B can be easily generated based on a conventional U-NET. Therefore, the trained model 24B can be created in a short time.

[0144] [Tire performance prediction method (fifth embodiment)] In the prediction method of this embodiment, as in the previous embodiments, a trained model 24B (shown in FIG. 11) generated based on the generation method shown in FIG. 4 is used.

[0145] In the prediction method of this embodiment, as in the previous embodiments, in step S3 shown in Fig. 7, data 21 including a pattern image 21A of a target tire and condition data 21D is input to the trained model 24B shown in Fig. 11. This makes it possible to output distribution data 22 of the target tire (in this example, outer surface distribution data 22A) from the trained model 24B.

[0146] Next, in the prediction method of this embodiment, as in the previous embodiments, in step S4 shown in FIG. 7, distribution data 22 of the target tire (in this example, outer surface distribution data 22A) is output from the trained model 24B shown in FIG. 11. As a result, the prediction method of this embodiment makes it possible to estimate the distribution data 22 of the target tire without requiring the experience or intuition of a skilled person. Then, if it is determined that the distribution data 22 is not satisfactory ("No" in step S5), at least a part of the tread pattern of the tread portion 12 of the target tire is changed in step S7 based on the physical quantities of the distribution data 22. As a result, a tire 11 having desired performance (for example, wear resistance performance, durability performance, and rolling resistance performance) can be reliably designed and manufactured.

[0147] Although a particularly preferred embodiment of the present invention has been described in detail above, the present invention is not limited to the illustrated embodiment and can be modified and implemented in various ways. [Example]

[0148] A trained model was generated (Example) based on the processing procedures shown in Figures 3 and 4. In the Example, a training dataset was prepared for multiple types of tires, combining data consisting of pattern images of the tread portions of tires not in contact with the road surface with distribution data of wear energy, as shown in Figure 5.

[0149] The training data was created based on 393 types of tires, including passenger car tires, light truck tires, and heavy-duty tires. Using these training datasets, a machine learning model (U-NET) was trained so that when data consisting of pattern images of target tires is input, distribution data of the target tires is output.

[0150] Next, in the embodiment, based on the processing procedure shown in Figure 7, data consisting of a pattern image of the target tire was input into the trained model, and distribution data of the target tire estimated by the trained model was output.

[0151] In order to evaluate the accuracy of the estimated distribution data, a simulation was performed in which a tire model, which modeled the target tire, was brought into contact with a road surface model, which modeled the road surface, and the wear energy distribution data was calculated. The processing procedure for this simulation was as described above.

[0152] The estimated distribution data was then compared with the calculated distribution data (ground truth data). Figure 12(a) shows the distribution data estimated using the trained model. Figure 12(b) shows the distribution data calculated by simulation.

[0153] As a result of the test, the distribution data estimated by the trained model was able to predict with 90% accuracy compared to the calculated distribution data (ground truth data). Note that a prediction with an accuracy of 80% or more is considered satisfactory. Therefore, the example was able to estimate the distribution data of physical quantities acting on the tread portion of the target tire using the trained model without performing time-consuming actual measurements or simulations.

[0154] [Note] The present invention includes the following aspects.

[0155] [Invention 1] a derivation unit that inputs data including at least a pattern image of a target tire into a trained model that has been machine-learned so as to, when inputting data including at least a pattern image of a tread portion of a tire that is not in contact with a road surface, output distribution data of physical quantities acting on the tread portion when the tread portion is in contact with the road surface, and outputs distribution data of physical quantities of the target tire; Tire performance prediction device. [Invention 2] The tire performance prediction device according to the first aspect of the present invention, wherein the pattern image is a grayscale image. [Invention 3] 3. The tire performance prediction device according to claim 1 or 2, wherein the physical quantity is a physical quantity acting on the outer surface of the tread portion or on the inside of the tread radially inward of the outer surface. [Invention 4] The tire performance prediction device according to aspect 3, wherein the physical quantity includes abrasion energy acting on the outer surface. [Invention 5] 5. The tire performance prediction device according to claim 3 or 4, wherein the physical quantity includes strain and / or stress acting inside the tread. [Invention 6] 1. A method for predicting tire performance executed by one or more processors, comprising: a step of inputting data including at least a pattern image of a target tire into a trained model that has been machine-learned so that, when data including at least a pattern image of a tread portion of a tire not in contact with a road surface is input, the trained model outputs distribution data of physical quantities acting on the tread portion when the tread portion is in contact with the road surface; outputting distribution data of the physical quantities of the target tire from the trained model, A method for predicting tire performance. [Invention 7] preparing a learning dataset that combines data including at least a pattern image of a tread portion of a tire that is not in contact with a road surface and distribution data of physical quantities that act on the tread portion when the tread portion is in contact with a road surface; and training a machine learning model using the learning dataset so that when data including at least the pattern image of a target tire is input, distribution data of physical quantities of the target tire is output. How to generate a trained model. [Explanation of symbols]

[0156] 21 Data 21A Pattern Image 22 Distribution Data 24B trained model

Claims

1. a derivation unit that inputs data including at least a pattern image of a target tire into a trained model that has been machine-learned so as to, when inputting data including at least a pattern image of a tread portion of a tire that is not in contact with a road surface, output distribution data of physical quantities acting on the tread portion when the tread portion is in contact with the road surface, and outputs distribution data of physical quantities of the target tire; Tire performance prediction device.

2. The tire performance prediction device according to claim 1 , wherein the pattern image is a grayscale image.

3. 2. The tire performance prediction device according to claim 1, wherein the physical quantity is a physical quantity acting on an outer surface of the tread portion or on an inside of the tread radially inward of the outer surface.

4. The tire performance prediction device according to claim 3 , wherein the physical quantity includes abrasion energy acting on the outer surface.

5. The tire performance prediction device according to claim 3 , wherein the physical quantity includes strain and / or stress acting inside the tread.

6. 1. A method for predicting tire performance executed by one or more processors, comprising: a step of inputting data including at least a pattern image of a target tire into a trained model that has been machine-learned so that, when data including at least a pattern image of a tread portion of a tire not in contact with a road surface is input, the trained model outputs distribution data of physical quantities acting on the tread portion when the tread portion is in contact with the road surface; outputting distribution data of the physical quantities of the target tire from the trained model. A method for predicting tire performance.

7. preparing a learning dataset that combines data including at least a pattern image of a tread portion of a tire that is not in contact with a road surface and distribution data of physical quantities that act on the tread portion when the tread portion is in contact with a road surface; and training a machine learning model using the learning dataset so that when data including at least the pattern image of a target tire is input, distribution data of physical quantities of the target tire is output. How to generate a trained model.

Images