JP2021181934A - Rubber material physical property prediction system and rubber material physical property prediction method - Google Patents

Rubber material physical property prediction system and rubber material physical property prediction method Download PDF

Info

Publication number
JP2021181934A
JP2021181934A JP2020087528A JP2020087528A JP2021181934A JP 2021181934 A JP2021181934 A JP 2021181934A JP 2020087528 A JP2020087528 A JP 2020087528A JP 2020087528 A JP2020087528 A JP 2020087528A JP 2021181934 A JP2021181934 A JP 2021181934A
Authority
JP
Japan
Prior art keywords
rubber material
physical property
property prediction
scattered
image
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
JP2020087528A
Other languages
Japanese (ja)
Other versions
JP7453057B2 (en
Inventor
智 鷺谷
Satoshi Sagitani
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Toyo Tire Corp
Original Assignee
Toyo Tire Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Toyo Tire Corp filed Critical Toyo Tire Corp
Priority to JP2020087528A priority Critical patent/JP7453057B2/en
Publication of JP2021181934A publication Critical patent/JP2021181934A/en
Application granted granted Critical
Publication of JP7453057B2 publication Critical patent/JP7453057B2/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Analysing Materials By The Use Of Radiation (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)

Abstract

To provide a rubber material physical property prediction system and rubber material physical property prediction method capable of highly accurately predicting physical properties of a rubber material.SOLUTION: A rubber material physical property prediction system 100 comprises a scattering image acquisition section 21 and a physical property predicting section 30. The scattering image acquisition section 21 acquires a scattering image captured by an X-ray small angle scattering measurement applying an X ray to a rubber material. The physical property predicting section 30 includes a learning-type physical property prediction arithmetic model 31, where the scattering image acquired by the scattering image acquisition section 21 is input in an input layer, that outputs physical properties of the rubber member from an output layer and executes convolution operation in the halfway operation from the input layer toward the output layer to extract a feature quantity of the scattering image.SELECTED DRAWING: Figure 1

Description

本発明は、ゴム材料の物性を予測するシステム、およびゴム材料物性予測方法に関する。 The present invention relates to a system for predicting the physical properties of a rubber material and a method for predicting the physical properties of a rubber material.

例えば車両に装着されるタイヤ等に用いられるゴム材料は、主要原料であるポリマーに補強剤、各種薬剤が添加された複合材料である。ゴム材料は、用途に応じて様々な化学構造および物性を備えたものが開発されている。 For example, a rubber material used for a tire or the like mounted on a vehicle is a composite material in which a reinforcing agent and various chemicals are added to a polymer which is a main raw material. Rubber materials with various chemical structures and physical properties have been developed depending on the intended use.

特許文献１には従来のゴム材料の特性推定方法が開示されている。この特性推定方法は、ゴム材料を顕微鏡により撮像した画像を取得するステップと、取得した画像から、画像の特徴を示す指標を算出するステップと、算出した指標に基づいて、連続的な曲線で表されるゴム材料の特性を推定するステップと、を備える。この特性推定方法は、ゴム材料の特性として応力−ひずみ曲線を推定するものである。 Patent Document 1 discloses a conventional method for estimating the characteristics of a rubber material. This characteristic estimation method consists of a step of acquiring an image of a rubber material imaged with a microscope, a step of calculating an index indicating the characteristics of the image from the acquired image, and a continuous curve based on the calculated index. It comprises a step of estimating the properties of the rubber material to be made. This characteristic estimation method estimates the stress-strain curve as a characteristic of the rubber material.

特許第６６０９３８７号公報Japanese Patent No. 6609387

特許文献１に記載されたゴム材料の特性推定方法では、ゴム材料を顕微鏡によって撮影した画像に基づいて特性を推定しているが、顕微鏡画像ではゴム材料中の局所的な画像を得ることしかできないため、予測精度や再現性が低くなるという問題点があった。 In the method for estimating the characteristics of a rubber material described in Patent Document 1, the characteristics are estimated based on an image of the rubber material taken with a microscope, but the microscopic image can only obtain a local image in the rubber material. Therefore, there is a problem that the prediction accuracy and the reproducibility are lowered.

本発明は、斯かる事情に鑑みてなされたものであり、その目的とするところは、ゴム材料の物性を精度良く予測することができるゴム材料物性予測システムおよびゴム材料物性予測方法を提供することにある。 The present invention has been made in view of such circumstances, and an object of the present invention is to provide a rubber material property prediction system and a rubber material property prediction method capable of accurately predicting the physical properties of a rubber material. It is in.

本発明のある態様はゴム材料物性予測システムである。ゴム材料物性予測システムは、ゴム材料にＸ線を照射するＸ線小角散乱測定によって撮影された散乱像を取得する散乱像取得部と、前記散乱像取得部によって取得された前記散乱像が入力層に入力され、出力層から前記ゴム材料の物性を出力する学習型の物性予測演算モデルを有し、前記入力層から前記出力層へ向けての途中演算において畳み込み演算を実行して前記散乱像の特徴量を抽出する物性予測処理部と、を備えることを特徴とする。 One aspect of the present invention is a rubber material property prediction system. In the rubber material physical property prediction system, a scattering image acquisition unit that acquires a scattering image taken by X-ray small-angle scattering measurement that irradiates the rubber material with X-rays, and the scattering image acquired by the scattering image acquisition unit are input layers. It has a learning-type physical property prediction calculation model that is input to and outputs the physical properties of the rubber material from the output layer, and performs a convolution calculation in the intermediate calculation from the input layer to the output layer to execute the folding calculation of the scattered image. It is characterized by including a physical characteristic prediction processing unit for extracting a feature amount.

また本発明の別の態様はゴム材料物性予測方法である。ゴム材料物性予測方法は、ゴム材料にＸ線を照射するＸ線小角散乱測定によって撮影された散乱像を取得する散乱像取得ステップと、前記散乱像取得ステップによって取得された前記散乱像が入力層に入力され、出力層から前記ゴム材料の物性を出力する学習型の物性予測演算モデルを有し、前記入力層から前記出力層へ向けての途中演算において畳み込み演算を実行して前記散乱像の特徴量を抽出する物性予測処理ステップと、を備えることを特徴とする。 Another aspect of the present invention is a method for predicting physical properties of a rubber material. In the method for predicting the physical properties of the rubber material, a scattering image acquisition step of acquiring a scattering image taken by X-ray small-angle scattering measurement in which the rubber material is irradiated with X-rays and the scattering image acquired by the scattering image acquisition step are input layers. It has a learning-type physical property prediction calculation model that is input to and outputs the physical properties of the rubber material from the output layer, and performs a convolution calculation in the intermediate calculation from the input layer to the output layer to execute the folding calculation of the scattered image. It is characterized by including a physical characteristic prediction processing step for extracting a feature amount.

本発明によれば、ゴム材料の物性を精度良く予測することができる。 According to the present invention, the physical properties of the rubber material can be predicted with high accuracy.

実施形態１に係るゴム材料物性予測システムの機能構成を示すブロック図である。It is a block diagram which shows the functional structure of the rubber material property property prediction system which concerns on Embodiment 1. 小角散乱測定装置によるＸ線小角散乱測定について説明するための模式図である。It is a schematic diagram for demonstrating the X-ray small angle scattering measurement by a small angle scattering measuring apparatus. Ｘ線小角散乱測定によって撮影された散乱像の一例を示す図である。It is a figure which shows an example of the scattering image taken by the X-ray small angle scattering measurement. 物性予測演算モデルの構成を示す模式図である。It is a schematic diagram which shows the structure of the physical property prediction calculation calculation model. 物性予測演算モデルの学習処理の手順を示すフローチャートである。It is a flowchart which shows the procedure of the learning process of a physical property prediction calculation model. ゴム材料の散乱像から予測した応力を示すグラフである。It is a graph which shows the stress predicted from the scattering image of a rubber material. ゴム材料の散乱像から予測した伸長量を示すグラフである。It is a graph which shows the elongation amount predicted from the scattering image of a rubber material. 実施形態２に係るゴム材料物性予測システムの機能構成を示すブロック図である。It is a block diagram which shows the functional structure of the rubber material property property prediction system which concerns on Embodiment 2. 画像生成処理部により生成した画像の例を示す図である。It is a figure which shows the example of the image generated by the image generation processing unit.

以下、本発明を好適な実施の形態をもとに図１から図９を参照しながら説明する。各図面に示される同一または同等の構成要素、部材には、同一の符号を付するものとし、適宜重複した説明は省略する。また、各図面における部材の寸法は、理解を容易にするために適宜拡大、縮小して示される。また、各図面において実施の形態を説明する上で重要ではない部材の一部は省略して表示する。 Hereinafter, the present invention will be described with reference to FIGS. 1 to 9 based on a preferred embodiment. The same or equivalent components and members shown in the drawings shall be designated by the same reference numerals, and duplicate description thereof will be omitted as appropriate. Further, the dimensions of the members in each drawing are shown in an appropriately enlarged or reduced size for easy understanding. In addition, some of the members that are not important for explaining the embodiment in each drawing are omitted and displayed.

（実施形態１）
図１は、実施形態１に係るゴム材料物性予測システム１００の機能構成を示すブロック図である。ゴム材料物性予測システム１００は、測定部１０、データ結合部２０および物性予測処理部３０を備え、例えばタイヤ等に用いられるゴム材料の物性を予測する。ゴム材料物性予測システム１００におけるデータ結合部２０および物性予測処理部３０は、例えばＰＣ（パーソナルコンピュータ）等の情報処理装置である。ゴム材料物性予測システム１００における各部は、ハードウェア的には、コンピュータのＣＰＵをはじめとする電子素子や機械部品などで実現でき、ソフトウェア的にはコンピュータプログラムなどによって実現されるが、ここでは、それらの連携によって実現される機能ブロックを描いている。したがって、これらの機能ブロックはハードウェア、ソフトウェアの組合せによっていろいろな形態で実現できることは、当業者には理解されるところである。 (Embodiment 1)
FIG. 1 is a block diagram showing a functional configuration of the rubber material property property prediction system 100 according to the first embodiment. The rubber material property prediction system 100 includes a measurement unit 10, a data coupling unit 20, and a physical property prediction processing unit 30, and predicts the physical properties of a rubber material used for, for example, a tire. The data coupling unit 20 and the physical property prediction processing unit 30 in the rubber material physical property prediction system 100 are information processing devices such as a PC (personal computer). Each part of the rubber material property prediction system 100 can be realized by electronic elements such as a computer CPU and mechanical parts in terms of hardware, and by a computer program in terms of software. It depicts a functional block realized by the cooperation of. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms by combining hardware and software.

測定部１０は、小角散乱測定装置１１および引張試験機１２を有する。小角散乱測定装置１１は、物性を予測する対象であるゴム材料に対してＸ線を照射するＸ線小角散乱（ＳＡＸＳ：Small Angle X-ray Scattering）測定を実施し、散乱像を撮影する装置である。図２は小角散乱測定装置１１によるＸ線小角散乱測定について説明するための模式図であり、図３はＸ線小角散乱測定によって撮影された散乱像の一例を示す図である。小角散乱測定装置１１は、ゴム材料のサンプルＳに対してＸ線を照射し、サンプルＳから所定のカメラ長の距離に配置された検出器１１ａによって散乱像を二次元画像として検出する。サンプルＳと検出器１１ａとの間には、散乱光が透過しないビームストッパー１１ｂが設けられている。 The measuring unit 10 has a small angle scattering measuring device 11 and a tensile tester 12. The small-angle scattering measuring device 11 is a device that performs X-ray small-angle scattering (SAXS: Small Angle X-ray Scattering) measurement by irradiating a rubber material whose physical properties are to be predicted with X-rays and captures a scattered image. be. FIG. 2 is a schematic diagram for explaining the small-angle X-ray scattering measurement by the small-angle scattering measuring device 11, and FIG. 3 is a diagram showing an example of a scattering image taken by the small-angle X-ray scattering measurement. The small-angle scattering measuring device 11 irradiates the sample S of the rubber material with X-rays, and detects the scattered image as a two-dimensional image by the detector 11a arranged at a distance of a predetermined camera length from the sample S. A beam stopper 11b that does not allow scattered light to pass through is provided between the sample S and the detector 11a.

検出器１１ａで撮像された散乱像は、ゴム材料の化学構造に応じて散乱光の輝度の高低が生じて所定の模様が現れ、ビームストッパー１１ｂによって中央に輝度０の部分が形成される。測定部１０では、引張試験機１２によってサンプルＳを任意の方向へ引張って変形させ、変形量に応じた複数の散乱像を小角散乱測定装置１１によって撮影することができる。小角散乱測定装置１１で撮影された散乱像はデータ結合部２０の散乱像取得部２１へ出力される。 In the scattered image captured by the detector 11a, the brightness of the scattered light varies depending on the chemical structure of the rubber material, a predetermined pattern appears, and a portion having a brightness of 0 is formed in the center by the beam stopper 11b. In the measuring unit 10, the sample S is pulled and deformed in an arbitrary direction by the tensile tester 12, and a plurality of scattered images according to the amount of deformation can be photographed by the small-angle scattering measuring device 11. The scattering image captured by the small-angle scattering measuring device 11 is output to the scattering image acquisition unit 21 of the data coupling unit 20.

引張試験機１２は、サンプルＳを変形させ、サンプルＳの変形量に対応して、応力および歪みの各データを計測し、データ結合部２０の変形データ取得部２２へ出力する。データ結合部２０は、散乱像取得部２１により取得した散乱像、並びに変形データ取得部２２により取得した変形量、応力および歪みの各データを対応付け、物性予測処理部３０へ出力する。 The tensile tester 12 deforms the sample S, measures each data of stress and strain corresponding to the amount of deformation of the sample S, and outputs the data to the deformation data acquisition unit 22 of the data coupling unit 20. The data coupling unit 20 associates the scattered image acquired by the scattered image acquisition unit 21 with the deformation amount, stress, and strain data acquired by the deformation data acquisition unit 22, and outputs the data to the physical property prediction processing unit 30.

物性予測処理部３０は、物性予測演算モデル３１および更新処理部３２を備え、既知のゴム材料の化学構造および物性に基づいて物性予測演算モデル３１を学習させ、作成される新たなゴム材料の物性を予測する。物性予測演算モデル３１は、ニューラルネットワーク等の学習型モデルを用いる。 The physical property prediction processing unit 30 includes a physical property prediction calculation model 31 and an update processing unit 32, and learns the physical property prediction calculation model 31 based on the chemical structure and physical properties of a known rubber material, and creates a new physical property of the rubber material. Predict. The physical property prediction calculation model 31 uses a learning model such as a neural network.

図４は、物性予測演算モデル３１の構成を示す模式図である。物性予測演算モデル３１は、ＣＮＮ（Convolutional Neural Network）型であり、その原型であるいわゆるＬｅＮｅｔで使用された畳み込み演算およびプーリング演算を備える学習型モデルである。物性予測演算モデル３１は、入力層４０、特徴抽出部４１、全結合部４２および出力層４３を備える。入力層４０には、散乱像取得部２１で取得した散乱像が入力される。特徴抽出部４１は、畳み込み演算４１ａおよびプーリング演算４１ｂを用いて特徴量を抽出し、全結合部４２へ伝達する。 FIG. 4 is a schematic diagram showing the configuration of the physical property prediction calculation model 31. The physical property prediction calculation model 31 is a CNN (Convolutional Neural Network) type, and is a learning type model including a convolutional calculation and a pooling calculation used in the so-called LeNet which is the prototype thereof. The physical property prediction calculation model 31 includes an input layer 40, a feature extraction unit 41, a fully connected unit 42, and an output layer 43. The scattered image acquired by the scattered image acquisition unit 21 is input to the input layer 40. The feature extraction unit 41 extracts the feature amount by using the convolution operation 41a and the pooling operation 41b, and transmits the feature amount to the fully connected unit 42.

特徴抽出部４１では、入力された散乱像に対して複数のフィルタを用いて１回目の畳み込み演算を実行する。特徴抽出部４１は、入力された散乱像に対してフィルタを移動させながら、畳み込み演算を実行する。尚、入力データの端に「０（ゼロ）」のデータを付加するゼロパティングを行って、畳み込み演算を実行するようにしてもよい。 The feature extraction unit 41 executes the first convolution operation on the input scattered image using a plurality of filters. The feature extraction unit 41 executes the convolution operation while moving the filter with respect to the input scattered image. It should be noted that zero putting may be performed in which "0 (zero)" data is added to the end of the input data to execute the convolution operation.

１回目の畳み込み演算後のデータに対して、１回目の最大値プーリング演算を実行する。特徴抽出部４１は、さらに２回目の畳み込み演算を実行して特徴量データを得て、全結合部４２へ出力する。 The first maximum value pooling operation is executed for the data after the first convolution operation. The feature extraction unit 41 further executes a second convolution operation to obtain feature amount data, and outputs the feature amount data to the fully connected unit 42.

全結合部４２は、重みづけを用いた線形演算等を実行する全結合のパスによって出力層４３へ結び付ける。全結合部４２では、線形演算に加えて、活性化関数などを用いて非線形演算を実行するようにしてもよい。出力層４３の各ノードには、例えば応力および歪み等のゴム材料の物性が出力される。 The fully connected portion 42 is connected to the output layer 43 by a fully connected path that executes a linear operation or the like using weighting. In the fully connected unit 42, in addition to the linear operation, a non-linear operation may be executed by using an activation function or the like. The physical characteristics of the rubber material such as stress and strain are output to each node of the output layer 43.

物性予測演算モデル３１は、ゴム材料に対して計測した散乱像と当該ゴム材料の物性に基づいて学習させることができる。更新処理部３２は、散乱像に基づいて物性予測演算モデル３１により算出した物性と、教師データとして与えられる物性とを比較し、例えば逆拡散演算により、各ノード間の重みづけを修正して物性予測演算モデル３１の学習を繰り返す。学習の際に物性予測演算モデル３１に入力される散乱像は、１つのゴム材料の引張りによる変形量を変えて計測した散乱像であってもよいし、複数のゴム材料に対して計測した散乱像であってもよい。 The physical property prediction calculation model 31 can be trained based on the scattering image measured for the rubber material and the physical properties of the rubber material. The update processing unit 32 compares the physical properties calculated by the physical property prediction calculation model 31 based on the scattered image with the physical properties given as the teacher data, and corrects the weighting between the nodes by, for example, a back diffusion calculation. The learning of the prediction calculation model 31 is repeated. The scattering image input to the physical property prediction calculation model 31 at the time of learning may be a scattering image measured by changing the amount of deformation due to tension of one rubber material, or scattering measured for a plurality of rubber materials. It may be an image.

また物性予測演算モデル３１の学習では、適宜、物性予測演算モデル３１に入力されるデータをトレーニング用データ（例えば９０％のデータ）と、検証用データ（残りの１０％のデータ）とに分けて、交差検証を実行する。物性予測処理部３０は、交差検証により平均的に良い結果を予測する物性予測演算モデル３１を選択することになる。 Further, in the training of the physical property prediction calculation model 31, the data input to the physical property prediction calculation model 31 is appropriately divided into training data (for example, 90% data) and verification data (remaining 10% data). , Perform cross-validation. The physical property prediction processing unit 30 selects the physical property prediction calculation model 31 that predicts good results on average by cross-validation.

物性予測処理部３０は、散乱像取得部２１から入力される散乱像に対して、学習済みの物性予測演算モデル３１を用いて演算を実行し、ゴム材料の物性を新たに予測することができる。 The physical property prediction processing unit 30 can perform a calculation on the scattered image input from the scattered image acquisition unit 21 using the trained physical property prediction calculation model 31 to newly predict the physical properties of the rubber material. ..

次にゴム材料物性予測システム１００の動作について説明する。図５は、物性予測演算モデル３１の学習処理の手順を示すフローチャートである。データ結合部２０の散乱像取得部２１は小角散乱測定装置１１からゴム材料の散乱像を取得し、変形データ取得部２２は応力および歪み等の物性を取得する（Ｓ１）。物性予測処理部３０は、データ結合部２０において取得した散乱像を物性予測演算モデル３１の入力層４０へ入力し、物性予測演算モデル３１によりゴム材料の物性を予測する（Ｓ２）。ステップＳ２では、上述のように、適宜、交差検証等の手法を用いる。 Next, the operation of the rubber material property prediction system 100 will be described. FIG. 5 is a flowchart showing the procedure of the learning process of the physical property prediction calculation model 31. The scattering image acquisition unit 21 of the data coupling unit 20 acquires a scattering image of the rubber material from the small angle scattering measuring device 11, and the deformation data acquisition unit 22 acquires physical properties such as stress and strain (S1). The physical property prediction processing unit 30 inputs the scattered image acquired in the data coupling unit 20 to the input layer 40 of the physical property prediction calculation calculation model 31, and predicts the physical properties of the rubber material by the physical property prediction calculation model 31 (S2). In step S2, as described above, a method such as cross-validation is appropriately used.

物性予測処理部３０は、既知の物性値に対して所定範囲内を予め定めて目標値（目標範囲）とし、交差検証によって平均的に良い結果を出力する物性予測演算モデル３１によって予測した物性が、目標値を満たすか否かを判定する（Ｓ３）。 The physical property prediction processing unit 30 sets a predetermined range in advance with respect to the known physical property value and sets it as a target value (target range), and the physical property predicted by the physical property prediction calculation model 31 that outputs good results on average by cross-validation is obtained. , It is determined whether or not the target value is satisfied (S3).

ステップＳ３によって、物性予測演算モデル３１によって予測した物性が、目標値を満たす場合（Ｓ３：ＹＥＳ）、処理を終了する。一方、物性予測演算モデル３１によって予測した物性が、目標値を満たさない場合（Ｓ３：ＮＯ）、ステップＳ１に戻って処理を繰り返す。 When the physical characteristics predicted by the physical property prediction calculation model 31 in step S3 satisfy the target value (S3: YES), the process ends. On the other hand, when the physical properties predicted by the physical property prediction calculation model 31 do not satisfy the target value (S3: NO), the process returns to step S1 and the process is repeated.

ゴム材料物性予測システム１００は、学習によって物性予測演算モデル３１を構築し、ゴム材料の新たな散乱像に基づいてゴム材料の物性を予測することができる。図６は、ゴム材料の散乱像から予測した応力を示すグラフである。図６に示すグラフでは、横軸に実際に発生している応力を、縦軸に予測した応力をとり、実際に発生している応力と予測した応力とが一致している場合を破線で示している。図６に示すように、ゴム材料物性予測システム１００によって予測したゴム材料の応力は、破線の付近に分布し、破線に添って良好な相関関係を示している。 The rubber material physical property prediction system 100 can construct a physical property prediction calculation model 31 by learning and predict the physical properties of the rubber material based on a new scattered image of the rubber material. FIG. 6 is a graph showing the stress predicted from the scattered image of the rubber material. In the graph shown in FIG. 6, the stress actually generated on the horizontal axis is taken as the predicted stress on the vertical axis, and the case where the actually generated stress and the predicted stress match is shown by a broken line. ing. As shown in FIG. 6, the stress of the rubber material predicted by the rubber material property prediction system 100 is distributed in the vicinity of the broken line, and shows a good correlation along the broken line.

また図７は、ゴム材料の散乱像から予測した伸長量を示すグラフである。図７に示すグラフでは、横軸に実際に発生している伸長量を、縦軸に予測した伸長量をとり、実際に発生している伸長量と予測した伸長量とが一致している場合を破線で示している。 Further, FIG. 7 is a graph showing the amount of elongation predicted from the scattered image of the rubber material. In the graph shown in FIG. 7, when the amount of elongation actually occurring is taken on the horizontal axis and the amount of elongation predicted is taken on the vertical axis, and the amount of elongation actually occurring and the predicted amount of elongation match. Is indicated by a broken line.

図７に示すように、ゴム材料物性予測システム１００によって予測したゴム材料の伸長量は、破線の付近に分布し、破線に添って良好な相関関係を示している。図６および図７に示されるように、ゴム材料物性予測システム１００は、小角散乱測定によって計測した散乱像を用いることで、精度良くゴム材料の物性を予測することができる。 As shown in FIG. 7, the elongation amount of the rubber material predicted by the rubber material property property prediction system 100 is distributed in the vicinity of the broken line, and shows a good correlation along the broken line. As shown in FIGS. 6 and 7, the rubber material property prediction system 100 can accurately predict the physical properties of the rubber material by using the scattering image measured by the small-angle scattering measurement.

またゴム材料物性予測システム１００は、ゴム材料を引張って変形させて散乱像を取得することができ、変形量に応じた複数の散乱像と物性（応力や歪み、伸長量等）とを用いて物性予測演算モデル３１を学習させることができる。 Further, the rubber material physical property prediction system 100 can acquire a scattered image by pulling and deforming the rubber material, and uses a plurality of scattered images according to the amount of deformation and physical properties (stress, strain, elongation amount, etc.). The physical property prediction calculation model 31 can be trained.

さらにゴム材料物性予測システム１００は、ゴム材料を引張って変形させて散乱像を取得し、変形量に応じた複数の散乱像を入力として用い、ゴム材料の物性（応力や歪み、伸長量等）を予測し、例えば応力−歪み曲線などを予測することもできる。 Further, the rubber material property prediction system 100 acquires a scattered image by pulling and deforming the rubber material, and uses a plurality of scattered images according to the amount of deformation as inputs, and the physical properties of the rubber material (stress, strain, elongation amount, etc.). It is also possible to predict, for example, a stress-strain curve.

（実施形態２）
図８は実施形態２に係るゴム材料物性予測システム１００の機能構成を示すブロック図である。ゴム材料物性予測システム１００は、画像生成処理部５０を有するほか、実施形態１と同等の構成を有する。実施形態２では、ゴム材料を繰り返し複数往復の変形をさせる。ゴム材料の変形量に関して一つ若しくは複数の変形量区間に分割し、各区間において、各区間の応力の中央値より応力が大きい場合の散乱像と小さい場合の散乱像に分ける。ゴム材料物性予測システム１００は、区間内の中央値より応力が大きい場合の散乱像および小さい場合の散乱像を物性予測演算モデル３１に入力してゴム材料の物性を予測する。 (Embodiment 2)
FIG. 8 is a block diagram showing a functional configuration of the rubber material property property prediction system 100 according to the second embodiment. The rubber material property prediction system 100 has an image generation processing unit 50 and has the same configuration as that of the first embodiment. In the second embodiment, the rubber material is repeatedly deformed in a plurality of reciprocations. The deformation amount of the rubber material is divided into one or a plurality of deformation amount sections, and each section is divided into a scattered image when the stress is larger than the median stress of each section and a scattered image when the stress is smaller than the median. The rubber material property prediction system 100 predicts the physical properties of the rubber material by inputting the scattering image when the stress is larger than the median in the section and the scattering image when the stress is smaller than the median in the section into the physical property prediction calculation model 31.

画像生成処理部５０は、物性予測演算モデル３１での演算に基づきＣＮＮの判断根拠を散乱像にハイライトで可視化した画像を生成する。ＣＮＮの判断根拠を算出する方法は公知の深層学習ライブラリを用いて実現できる。例えば、ＳＡＳ Ｉｎｓｔｉｔｕｔｅ社が提供するＤＬＰｙや、オープンソースであるＫｅｒａｓにあるＧｒａｄ−ＣＡＭ（Gradient-weighted Class Activation Mapping）手法などがある。図９は、画像生成処理部５０により生成した画像の例を示す図である。図９に示す画像は、ＤＬＰｙのヒートマップ解析手法を用いて可視化したものである。図９ではゴム材料の変形量について区間Ａ〜Ｃの３区間において区間内の応力の中央値より応力が小さい場合と、応力が大きい場合とで物性予測演算モデル３１により予測演算し、画像生成処理部５０により画像を生成している。画像生成処理部５０は、上述のようにゴム材料を繰り返し複数往復の変形をさせる過程において取得される散乱像を選択し、物性予測演算モデル３１により予測演算して画像を生成している。 The image generation processing unit 50 generates an image in which the judgment basis of CNN is visualized as a scattered image with highlights based on the calculation in the physical property prediction calculation model 31. The method of calculating the judgment basis of CNN can be realized by using a known deep learning library. For example, there are DLPy provided by SAS Institute and a Grad-CAM (Gradient-weighted Class Activation Mapping) method in Keras, which is open source. FIG. 9 is a diagram showing an example of an image generated by the image generation processing unit 50. The image shown in FIG. 9 is visualized using the heat map analysis method of DLPy. In FIG. 9, the deformation amount of the rubber material is predicted and calculated by the physical property prediction calculation model 31 in the three sections A to C when the stress is smaller than the median stress in the section and when the stress is large, and image generation processing is performed. The image is generated by the unit 50. The image generation processing unit 50 selects a scattered image acquired in the process of repeatedly deforming the rubber material in a plurality of reciprocations as described above, and performs a prediction calculation by the physical property prediction calculation calculation model 31 to generate an image.

図９に示すように、各区間で応力が小さい場合と大きい場合で画像に表われる濃淡が変化しており、ヒートマップ解析等の解析に供することができる。画像生成処理部５０によって生成された画像から、例えば画像中のどの箇所に着目すれば、応力を高く（または低く）することができるかがわかる。また散乱像はゴム材料中のフィラー分散構造を示しているため、画像生成処理部５０によって生成された画像から、散乱像のどの箇所がフィラー分散に寄与して応力が高く（または低く）することができるかがわかる。 As shown in FIG. 9, the shading that appears in the image changes depending on whether the stress is small or large in each section, and can be used for analysis such as heat map analysis. From the image generated by the image generation processing unit 50, it is possible to know, for example, which part of the image the stress can be increased (or decreased). Further, since the scattered image shows the filler dispersion structure in the rubber material, which part of the scattered image contributes to the filler dispersion from the image generated by the image generation processing unit 50 to increase (or decrease) the stress. I know if I can do it.

また、画像生成処理部５０によって生成された画像に対応する箇所の散乱像から散乱曲線を作成しギニエ解析（凝集体サイズやフラクタル構造解析など）することで、応力に影響するフィラー分散構造中の要因分析などを行うこともできる。 Further, by creating a scattering curve from the scattering image of the portion corresponding to the image generated by the image generation processing unit 50 and performing Guinier analysis (aggregate size, fractal structure analysis, etc.), the filler dispersion structure that affects stress is formed. It is also possible to perform factor analysis.

次に実施形態に係るゴム材料物性予測システム１００およびゴム材料物性予測方法の特徴について説明する。
実施形態に係るゴム材料物性予測システム１００は、散乱像取得部２１および物性予測処理部３０を備える。散乱像取得部２１は、ゴム材料にＸ線を照射するＸ線小角散乱測定によって撮影された散乱像を取得する。物性予測処理部３０は、散乱像取得部２１によって取得された散乱像が入力層４０に入力され、出力層４３からゴム材料の物性を出力する学習型の物性予測演算モデル３１を有し、入力層４０から出力層４３へ向けての途中演算において畳み込み演算を実行して散乱像の特徴量を抽出する。これにより、ゴム材料物性予測システム１００は、ゴム材料の散乱像に基づいてゴム材料の物性を精度良く予測することができる。 Next, the features of the rubber material property prediction system 100 and the rubber material property prediction method according to the embodiment will be described.
The rubber material physical property prediction system 100 according to the embodiment includes a scattering image acquisition unit 21 and a physical property prediction processing unit 30. The scattering image acquisition unit 21 acquires a scattering image captured by the small-angle X-ray scattering measurement in which the rubber material is irradiated with X-rays. The physical property prediction processing unit 30 has a learning-type physical property prediction calculation model 31 in which the scattered image acquired by the scattered image acquisition unit 21 is input to the input layer 40 and the physical properties of the rubber material are output from the output layer 43, and is input. The convolution operation is executed in the intermediate calculation from the layer 40 to the output layer 43 to extract the feature amount of the scattered image. As a result, the rubber material property prediction system 100 can accurately predict the physical properties of the rubber material based on the scattered image of the rubber material.

また散乱像取得部２１は、ゴム材料を変形させて撮影した複数の散乱像を取得する。物性予測演算モデル３１は、散乱像取得部２１によって取得された複数の散乱像に基づいて学習されている。これにより、ゴム材料物性予測システム１００は、ゴム材料を変形させて散乱像を取得し物性予測演算モデル３１を学習させることができる。 Further, the scattered image acquisition unit 21 acquires a plurality of scattered images taken by deforming the rubber material. The physical property prediction calculation model 31 is learned based on a plurality of scattered images acquired by the scattered image acquisition unit 21. As a result, the rubber material property prediction system 100 can deform the rubber material to acquire a scattered image and train the physical property prediction calculation model 31.

また散乱像取得部２１は、ゴム材料の変形量に関する区間内の応力の中央値より応力が小さい場合と大きい場合とで撮影された複数の散乱像を取得する。画像生成処理部５０は、散乱像取得部２１によって取得された複数の散乱像による物性予測演算モデル３１での演算に基づき前記散乱像上に判断根拠を可視化した画像を生成する。これにより、ゴム材料物性予測システム１００は、生成した画像を、ゴム材料の応力等の物性に寄与する要因の分析に供することができる。 Further, the scattered image acquisition unit 21 acquires a plurality of scattered images taken when the stress is smaller and larger than the median stress in the section regarding the amount of deformation of the rubber material. The image generation processing unit 50 generates an image in which the judgment basis is visualized on the scattered image based on the calculation in the physical property prediction calculation model 31 using the plurality of scattered images acquired by the scattered image acquisition unit 21. Thereby, the rubber material physical property prediction system 100 can use the generated image for analysis of factors contributing to physical properties such as stress of the rubber material.

ゴム材料物性予測方法は、散乱像取得ステップおよび物性予測処理ステップを備える。散乱像取得ステップは、ゴム材料にＸ線を照射するＸ線小角散乱測定によって撮影された散乱像を取得する。物性予測処理ステップは、散乱像取得ステップによって取得された散乱像が入力層４０に入力され、出力層４３からゴム材料の物性を出力する学習型の物性予測演算モデル３１を有し、入力層４０から出力層４３へ向けての途中演算において畳み込み演算を実行して散乱像の特徴量を抽出する。このゴム材料物性予測方法によれば、ゴム材料の散乱像に基づいてゴム材料の物性を精度良く予測することができる。 The rubber material physical property prediction method includes a scattering image acquisition step and a physical property prediction processing step. The scattering image acquisition step acquires a scattering image captured by the small-angle X-ray scattering measurement in which the rubber material is irradiated with X-rays. The physical property prediction processing step has a learning type physical property prediction calculation model 31 in which the scattered image acquired by the scattered image acquisition step is input to the input layer 40 and the physical properties of the rubber material are output from the output layer 43, and the input layer 40. The feature amount of the scattered image is extracted by executing the convolution operation in the intermediate calculation toward the output layer 43. According to this method for predicting the physical characteristics of the rubber material, the physical properties of the rubber material can be accurately predicted based on the scattered image of the rubber material.

以上、本発明の実施の形態をもとに説明した。これらの実施の形態は例示であり、いろいろな変形および変更が本発明の特許請求範囲内で可能なこと、またそうした変形例および変更も本発明の特許請求の範囲にあることは当業者に理解されるところである。従って、本明細書での記述および図面は限定的ではなく例証的に扱われるべきものである。 The above description has been made based on the embodiment of the present invention. It will be appreciated by those skilled in the art that these embodiments are exemplary and that various modifications and modifications are possible within the claims of the invention and that such modifications and modifications are also within the claims of the present invention. It is about to be done. Therefore, the descriptions and drawings herein should be treated as exemplary rather than limiting.

２１ 散乱像取得部、 ３０ 物性予測処理部、 ３１ 物性予測演算モデル、
４０ 入力層、 ４３ 出力層、 ５０ 画像生成処理部、
１００ ゴム材料物性予測システム。 21 Scattered image acquisition unit, 30 Physical property prediction processing unit, 31 Physical property prediction calculation model,
40 input layer, 43 output layer, 50 image generation processing unit,
100 Rubber material physical property prediction system.

Claims (4)

ゴム材料にＸ線を照射するＸ線小角散乱測定によって撮影された散乱像を取得する散乱像取得部と、
前記散乱像取得部によって取得された前記散乱像が入力層に入力され、出力層から前記ゴム材料の物性を出力する学習型の物性予測演算モデルを有し、前記入力層から前記出力層へ向けての途中演算において畳み込み演算を実行して前記散乱像の特徴量を抽出する物性予測処理部と、
を備えることを特徴とするゴム材料物性予測システム。 A scattering image acquisition unit that acquires a scattering image taken by X-ray small-angle scattering measurement that irradiates a rubber material with X-rays, and a scattering image acquisition unit.
It has a learning-type physical property prediction calculation model in which the scattered image acquired by the scattered image acquisition unit is input to the input layer and outputs the physical properties of the rubber material from the output layer, and the input layer is directed toward the output layer. A physical characteristic prediction processing unit that extracts the feature amount of the scattered image by executing a convolution operation in the intermediate calculation.
A rubber material property prediction system characterized by being equipped with. 前記散乱像取得部は、前記ゴム材料を変形させて撮影した複数の前記散乱像を取得し、
前記物性予測演算モデルは、前記散乱像取得部によって取得された複数の前記散乱像に基づいて学習されていることを特徴とする請求項１に記載のゴム材料物性予測システム。 The scattered image acquisition unit acquires a plurality of the scattered images taken by deforming the rubber material.
The rubber material property prediction system according to claim 1, wherein the physical property prediction calculation model is learned based on a plurality of the scattered images acquired by the scattered image acquisition unit. 前記散乱像取得部は、前記ゴム材料の変形量に関する区間内の応力の中央値より応力が小さい場合と大きい場合とで撮影された複数の前記散乱像を取得し、
前記散乱像取得部によって取得された複数の前記散乱像による前記物性予測演算モデルでの演算に基づき前記散乱像上に判断根拠を可視化した画像を生成する画像生成処理部を更に備えることを特徴とする請求項１に記載のゴム材料物性予測システム。 The scattered image acquisition unit acquires a plurality of the scattered images taken when the stress is smaller and larger than the median stress in the section regarding the amount of deformation of the rubber material.
It is further provided with an image generation processing unit that generates an image that visualizes the judgment basis on the scattered image based on the calculation in the physical property prediction calculation model using the plurality of scattered images acquired by the scattered image acquisition unit. The rubber material physical property prediction system according to claim 1. ゴム材料にＸ線を照射するＸ線小角散乱測定によって撮影された散乱像を取得する散乱像取得ステップと、
前記散乱像取得ステップによって取得された前記散乱像が入力層に入力され、出力層から前記ゴム材料の物性を出力する学習型の物性予測演算モデルを有し、前記入力層から前記出力層へ向けての途中演算において畳み込み演算を実行して前記散乱像の特徴量を抽出する物性予測処理ステップと、
を備えることを特徴とするゴム材料物性予測方法。 A scattering image acquisition step to acquire a scattering image taken by X-ray small-angle scattering measurement that irradiates a rubber material with X-rays, and
The scattered image acquired by the scattered image acquisition step is input to the input layer, has a learning type physical property prediction calculation model that outputs the physical properties of the rubber material from the output layer, and is directed from the input layer to the output layer. A physical characteristic prediction processing step for extracting the feature amount of the scattered image by executing a convolution operation in the intermediate calculation, and
A method for predicting physical properties of a rubber material, which comprises.
JP2020087528A 2020-05-19 2020-05-19 Rubber material property prediction system and rubber material property prediction method Active JP7453057B2 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2020087528A JP7453057B2 (en) 2020-05-19 2020-05-19 Rubber material property prediction system and rubber material property prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP2020087528A JP7453057B2 (en) 2020-05-19 2020-05-19 Rubber material property prediction system and rubber material property prediction method

Publications (2)

Publication Number Publication Date
JP2021181934A true JP2021181934A (en) 2021-11-25
JP7453057B2 JP7453057B2 (en) 2024-03-19

Family

ID=78606455

Family Applications (1)

Application Number Title Priority Date Filing Date
JP2020087528A Active JP7453057B2 (en) 2020-05-19 2020-05-19 Rubber material property prediction system and rubber material property prediction method

Country Status (1)

Country Link
JP (1) JP7453057B2 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115372200A (en) * 2022-07-12 2022-11-22 蓝冰河(常州)精密测量技术有限责任公司 Online surface density measuring method and device based on X-ray

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115372200A (en) * 2022-07-12 2022-11-22 蓝冰河(常州)精密测量技术有限责任公司 Online surface density measuring method and device based on X-ray

Also Published As

Publication number Publication date
JP7453057B2 (en) 2024-03-19

Similar Documents

Publication Publication Date Title
Liu et al. Computer vision-based concrete crack detection using U-net fully convolutional networks
JP6240206B2 (en) Displacement field and strain field measurement method and material testing machine
Ling et al. Integration of structural health monitoring and fatigue damage prognosis
KR101232010B1 (en) Method for calculating brightness level in the area of interest of the digital x-ray image for medical applications
JP6707920B2 (en) Image processing apparatus, image processing method, and program
Wu et al. From SEM images to elastic responses: A stochastic multiscale analysis of UD fiber reinforced composites
CN111310841A (en) Medical image classification method, apparatus, device, computer device and storage medium
Lampeas et al. On the validation of solid mechanics models using optical measurements and data decomposition
Ozcan A deep learning framework for coronavirus disease (covid-19) detection in x-ray images
Hanhan et al. ModLayer: a matlab GUI Drawing Segmentation tool for visualizing and Classifying 3D data
Das et al. Application of deep convolutional neural networks for automated and rapid identification and computation of crack statistics of thin cracks in strain hardening cementitious composites (SHCCs)
CN114596266A (en) Concrete crack detection method based on ConcreteCrackSegNet model
JP2021181934A (en) Rubber material physical property prediction system and rubber material physical property prediction method
Anton et al. Physics-informed neural networks for material model calibration from full-field displacement data
Papadimas et al. A hierarchical Bayesian approach for calibration of stochastic material models
Seemab et al. Automated detection of propagating cracks in RC beams without shear reinforcement based on DIC-controlled modeling of damage localization
Zhang et al. Identification of concrete surface damage based on probabilistic deep learning of images
Thakre et al. Quantification of similarity and physical awareness of microstructures generated via generative models
Dawood et al. Data fusion of multiple machine intelligent systems for the condition assessment of subway structures
Jung et al. Inverse identification of creep of concrete from in situ load–displacement monitoring
Mu et al. A radiographic image quality assessment algorithm based on network topology analysis
Shahid et al. A neural network-based method for coverage measurement of shot-peened panels
JP6789460B1 (en) Fracture surface analysis device, trained model generator, fracture surface analysis method, fracture surface analysis program, and trained model
Anton et al. Deterministic and statistical calibration of constitutive models from full-field data with parametric physics-informed neural networks
Bhatia et al. A proposed stratification approach for MRI images

Legal Events

Date Code Title Description
A621 Written request for application examination

Free format text: JAPANESE INTERMEDIATE CODE: A621

Effective date: 20230309

A131 Notification of reasons for refusal

Free format text: JAPANESE INTERMEDIATE CODE: A131

Effective date: 20231212

A977 Report on retrieval

Free format text: JAPANESE INTERMEDIATE CODE: A971007

Effective date: 20231213

A521 Request for written amendment filed

Free format text: JAPANESE INTERMEDIATE CODE: A523

Effective date: 20240208

TRDD Decision of grant or rejection written
A01 Written decision to grant a patent or to grant a registration (utility model)

Free format text: JAPANESE INTERMEDIATE CODE: A01

Effective date: 20240220

A61 First payment of annual fees (during grant procedure)

Free format text: JAPANESE INTERMEDIATE CODE: A61

Effective date: 20240307

R150 Certificate of patent or registration of utility model

Ref document number: 7453057

Country of ref document: JP

Free format text: JAPANESE INTERMEDIATE CODE: R150