Physical property estimation method and device for composite material

Abstract

To provide a physical property estimation method and device for a composite material capable of easily and accurately estimating the physical property of the composite material.SOLUTION: A method for estimating the physical property of a composite material obtained by kneading a polymer and a filler with a kneader 10 includes: a preprocessing step in which a vibration sensor 11 provided on the kneader 10 and a temperature sensor 13 for detecting a material temperature are used, the output of the corresponding vibration sensor 11 is extracted at predetermined temperature intervals of the material temperature detected by the temperature sensor 13, frequency analysis is performed on the extracted output of the vibration sensor 11 to obtain an intensity spectrum, and the obtained intensity spectrum is associated with the corresponding material temperature and stored as preprocessed data 32; and a physical property estimation step in which the physical property of the composite material is estimated on the basis of the preprocessed data 32 of the composite material being an estimation target by using a learned model created by using the preprocessed data 32 and the physical property of the composite material when kneading has been normally performed for learning.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a method and apparatus for estimating physical properties of a composite material.

Background Art

[0002] Kneading of a composite material containing a polymer and a compounding agent such as an inorganic filler is performed, for example, in a batch kneader such as a kneader. In order to obtain desired physical properties in the composite material, it is important that the compounding agent is uniformly dispersed in the polymer and that sufficient adhesive force is obtained at the interface between the polymer and the compounding agent by kneading.

[0003] Note that Patent Document 1 is available as prior art document information related to the invention of this application.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] By the way, when manufacturing a composite material, even if the same material is kneaded under the same conditions, variations in the physical properties of the composite material due to kneading occur due to variations in raw materials, seasonal factors, effects of disturbances, etc. Therefore, for quality assurance, after kneading, there is a demand to confirm whether a composite material with desired physical properties can be manufactured from the obtained kneaded product.

[0006] However, there is a problem that it takes time and labor to evaluate the physical properties of the composite material obtained from the kneaded product, and the manufacturing cost increases. For example, when measuring the tensile strength, it is necessary to perform a tensile test after sheet forming, crosslinking, and specimen processing of the kneaded product, which takes a lot of time and labor.

[0007] To solve such problems, the applicant has proposed a method for estimating the physical properties of composite materials using a vibration sensor, but further improvement in estimation accuracy is desired.

[0008] Therefore, an object of the present invention is to provide a method and an apparatus for estimating the physical properties of a composite material that can easily and accurately estimate the physical properties of the composite material.

Means for Solving the Problems

[0009] The present invention is a method for estimating the physical properties of a composite material obtained by kneading at least a polymer and a filler with a kneader, the method comprising: a vibration sensor provided in the kneader for detecting vibration; and temperature detection means for detecting the material temperature which is the temperature of the composite material being kneaded in the kneader. Using these, for each predetermined temperature interval of the material temperature detected by the temperature detection means, the output of the corresponding vibration sensor is extracted, frequency analysis is performed on the extracted output of the vibration sensor to obtain an intensity spectrum, and the obtained intensity spectrum is associated with the corresponding material temperature and stored as pre-processed data. A pre-processing step; and a physical property estimation step of estimating the physical properties of the composite material to be estimated based on the pre-processed data of the composite material to be estimated, using at least the pre-processed data obtained when kneading is normally performed and a learned model created using the physical properties of the composite material for learning. A method for estimating the physical properties of a composite material is provided.

[0010] The present invention also aims to solve the above problems, and provides an apparatus for estimating the physical properties of a composite material obtained by kneading at least a polymer and a filler with a kneader, the apparatus comprising: a vibration sensor provided on the kneader for detecting vibration; temperature detection means for detecting the material temperature which is the temperature of the composite material being kneaded by the kneader; for each predetermined temperature interval of the material temperature detected by the temperature detection means, extracting the output of the corresponding vibration sensor, performing frequency analysis on the extracted output of the vibration sensor to obtain an intensity spectrum, and associating the obtained intensity spectrum with the corresponding material temperature and storing it as pre-processed data; and a physical property estimation unit for estimating the physical properties of the composite material to be estimated based on the pre-processed data of the composite material to be estimated, using at least the pre-processed data when kneading is performed normally and a learned model created using the physical properties of the composite material for learning.

Advantages of the Invention

[0011] According to the present invention, it is possible to provide a method and an apparatus for estimating the physical properties of a composite material, which can easily and accurately estimate the physical properties of the composite material.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Mode for Carrying Out the Invention

[0013] [Embodiment] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0014] FIG. 1 is a schematic configuration diagram of a physical property estimation device 1 for a composite material according to the present embodiment. The physical property estimation device 1 for a composite material is a device for estimating the physical properties of a composite material obtained by kneading at least a polymer and a filler with a kneader 10. In other words, the physical property estimation device 1 for a composite material is a device for estimating the physical properties of a composite material when the composite material is manufactured using the kneaded material obtained by kneading. The composite material is obtained by kneading two or more materials including at least a polymer and a compounding agent such as a filler. In the present embodiment, the composite material contains 200 parts by mass or more of an inorganic filler as a compounding agent.

[0015] As shown in FIG. 1, the physical property estimation device 1 for a composite material includes a vibration sensor 11 that detects vibration provided in the kneader 10, an arithmetic device 12, and a temperature sensor 13 as temperature detection means for detecting the temperature of the composite material being kneaded by the kneader 10.

[0016] The kneader 10 is, for example, a batch-type twin-screw kneader or the like. The kneader 10 has a rotor 101 attached thereto for kneading, and kneading is performed by rotating the rotor 101. A clearance 103 of about several millimeters is provided between a part of the rotor 101 and the wall surface of the kneading tank 102, and the region of this clearance 103 becomes a region where a strong shear stress is applied to the material (hereinafter referred to as a high-shear region). Kneading progresses as the material passes through this high-shear region. At the end of kneading, the polymer and the filler are in an adhered state, but when the material passes through the high-shear region in this state, the adhesion interface between the polymer and the filler deforms or is temporarily peeled off, and it is considered that sound waves (elastic waves) of high frequency (several kHz to several hundred kHz) are generated. In the present embodiment, this sound wave (elastic wave) is measured by the vibration sensor 11. Note that the kneader 10 is not limited to a batch-type twin-screw kneader, and other kneaders 10 such as a continuous kneader such as a continuous fig kneader or a roll kneader may be used.

[0017] In the present embodiment, an acoustic emission sensor (hereinafter referred to as an AE sensor) 111 is used as the vibration sensor 11. The AE sensor 111 is a sensor that detects elastic waves generated by the release of elastic energy such as local fracture inside the material during kneading. In the present embodiment, the AE sensor 111 is used as the vibration sensor 11, but it is not limited thereto. For example, when sensing at a lower frequency is desired, an acceleration sensor can be used. Also, various sensors such as a combination of the AE sensor 111 and an acceleration sensor may be used in combination.

[0018] Also, when the material passes through the high-shear region, the material generates heat due to the shear stress, and the temperature of the material rises. The temperature sensor 13 as the temperature detection means directly detects the temperature of the material being kneaded in the kneader 10 (hereinafter referred to as the material temperature). Note that the temperature detection means may indirectly detect the material temperature. For example, it may be a current sensor that detects the current value of the motor driving the rotor 101, or a sensor that measures the wall temperature of the kneading tank 102, or these may be used in combination. Since the material softens as the temperature rises and the current value of the motor driving the rotor 101 decreases, the material temperature can be indirectly detected by the current value of the motor.

[0019] The arithmetic unit 12 has a control unit 2 and a storage unit 3. A display 4 is connected to the arithmetic unit 12, and it is configured to be able to display the estimation results of the physical properties of the composite material, various data, etc. on the display 4. Further, an input device 5 such as a keyboard is provided in the arithmetic unit 12, and various settings and operations of the display content of the display 4 can be performed by the input of the input device 5. Note that the display 4 may be configured as a touch panel display so that the display 4 also serves as the input device 5. Furthermore, the display 4 and the input device 5 may not be wired-connected to the arithmetic unit 12 and may be wirelessly connected. In this case, the display 4 and the input device 5 may be, for example, a smartphone or a tablet.

[0020] The control unit 2 is equipped with a data acquisition processing unit 21, a preprocessing unit 22, a relationship derivation processing unit 23, a physical property estimation processing unit 24, and an estimated physical property presentation processing unit 25. These data acquisition processing unit 21, preprocessing unit 22, relationship derivation processing unit 23, physical property estimation processing unit 24, and estimated physical property presentation processing unit 25 are realized by appropriately combining arithmetic elements such as a CPU, memories such as RAM and ROM, software, interfaces, storage devices, etc. Details of each unit will be described later. The storage unit 3 is realized by a predetermined storage area of a memory or a storage device.

[0021] (Data Acquisition Processing Unit 21) The data acquisition processing unit 21 performs data acquisition processing that, during kneading, receives the outputs from the vibration sensor 11 (AE sensor 111) and the temperature sensor 13, and stores the received data in the storage unit 3 as measurement data 31. It may be configured such that formulation information of the kneaded material can be input using the input device 5 or the like.

[0022] (Preprocessing unit 22) The preprocessing unit 22 performs preprocessing (preprocessing step) to convert the measurement data 31 into data for machine learning (or data serving as an estimation source). More specifically, in the preprocessing unit 22, first, for each predetermined temperature interval of the material temperature detected by the temperature detection means, the output of the corresponding vibration sensor 11 (AE sensor 111) is extracted for a predetermined period. In the present embodiment, the output of the AE sensor 111 for 16 seconds is extracted every 1°C from 100°C to 150°C. Note that the period for extracting the output of the AE sensor 111 may be appropriately set in consideration of, for example, the time for the rotor 101 to rotate a predetermined number of times, the calculation time required for preprocessing to be within a realistic range, and the like.

[0023] More specifically, as shown in FIG. 2, for example, the output of the AE sensor 111 for 16 seconds from the time when the temperature reaches 100°C is extracted as the output of the AE sensor 111 corresponding to 100°C. The same applies to other temperatures. Note that, for example, in the illustrated example, if the time taken for the temperature to rise from 100°C to 101°C is 16 seconds or less, a part of the output of the AE sensor 111 corresponding to 100°C and the output of the AE sensor 111 corresponding to 101°C will overlap. Therefore, the period for acquiring the output of the AE sensor 111 may be appropriately set in consideration of the temperature rising rate of the material temperature so that such overlap does not occur. Note that if the period for acquiring the output of the AE sensor 111 is too short, the periodicity will be impaired and the estimation accuracy will decrease, so it is necessary to set it to a certain length of time (a time period in which the periodicity is not impaired).

[0024] After that, the preprocessing unit 22 performs frequency analysis on each of the outputs of the AE sensor 111 for each extracted material temperature to obtain an intensity spectrum. More specifically, as shown in FIG. 3, the preprocessing unit 22 performs a fast Fourier transform (FFT) on the output of the extracted AE sensor 111 (upper left in FIG. 3) and converts it into data of intensity for each frequency (upper right in FIG. 3). Since the number of data is still too large only by performing the fast Fourier transform, the preprocessing unit 22 divides the frequency into predetermined intervals, integrates the intensity distribution of the output of the AE sensor 111 for each interval to form a histogram, and obtains an intensity spectrum (lower left in FIG. 3). In the present embodiment, further, only a predetermined frequency range is extracted from the obtained intensity spectrum, data of unnecessary frequencies are deleted, and the number of data is reduced (lower right in FIG. 3). The number of data after reduction is preferably about 20 or more and 100 or less, for example. Regarding the frequency range to be used, experiments should be performed in advance and set to a frequency range in which it is easy to determine whether the kneading state is normal or abnormal, and it can be appropriately set according to the material to be kneaded and the type of the kneader 10, etc. The preprocessing unit 22 stores the obtained intensity spectrum in the storage unit 3 as preprocessed data 32 in association with the corresponding material temperature.

[0025] FIG. 4 shows an example of the preprocessed data 32 obtained by one kneading. As shown in FIG. 4, in the preprocessed data 32, the material temperature and the intensity spectrum obtained from the output of the AE sensor 111 corresponding to the material temperature are associated and stored. In the illustrated example, a total of 51 data are stored at intervals of 1°C from 100°C to 150°C.

[0026] As a result of measuring the physical properties of the composite material after kneading, when the physical properties of the composite material satisfy a preset pass criterion, the preprocessed data 32 is added to the learning data 33. Also, when the preprocessing of the measurement data 31 to be estimated is performed, the obtained preprocessed data 32 is stored in the storage unit 3 as the estimation source data 34.

[0027] (Relationship derivation processing unit 23) In the relationship derivation processing unit 23, a relationship derivation process (relationship derivation step) is performed to create a learned model by performing machine learning using at least the pre-processed data 32 when kneading is performed normally and the physical properties of the composite material. In the present embodiment, unsupervised learning using a one-class classifier is performed to create a learned model.

[0028] More specifically, in the present embodiment, a one-class support vector machine (1classSVM) is used as the one-class classifier, and the learning data 33, which is the pre-processed data 32 in the past when kneading is performed normally, that is, when the physical properties of the composite material satisfy the preset acceptance criteria, is repeatedly learned to create a learned model. In the learned model created in this way, when data to be estimated (that is, the estimation source data 34) is input, it is possible to estimate whether or not the physical properties of the manufactured composite material satisfy the preset acceptance criteria. Since the physical properties vary greatly depending on the composition of the composite material, it is preferable that the learned model be created for each composition. That is, the learned model is preferably created for each composition information and physical property to be estimated.

[0029] Since the one-class classifier can perform estimation by learning only normal data, it is particularly effective when the defect rate is low and it is difficult to obtain defective data (here, data when the physical properties of the composite material do not satisfy the acceptance criteria). Note that the machine learning algorithm is not limited to the one-class classifier and can be set as appropriate. For example, it may be one that obtains the value of the physical properties of the composite material.

[0030] (Physical property estimation processing unit 24) The physical property estimation processing unit 24 performs a physical property estimation process (physical property estimation step) to estimate the physical properties of the composite material based on the pre-processed data 32 of the composite material to be estimated using the learned model created in advance by the relationship derivation processing unit 23. In the present embodiment, the physical property estimation processing unit 24 estimates whether or not the physical properties of the composite material satisfy the acceptance criteria based on the estimation source data 34, which is the pre-processed data 32 of the composite material to be estimated, using the learned model.

[0031] As described with reference to FIG. 4, a plurality of pieces of data (51 pieces in FIG. 4) are obtained by preprocessing in one kneading and stored in the storage unit 3 as estimation source data 34. Therefore, in the physical property estimation process, for each piece of data included in the estimation source data 34 (each piece of data of the strength spectrum for each material temperature), it is estimated whether the physical properties of the composite material meet the acceptance criteria. Then, based on the number or ratio of the data estimated to meet or not meet the acceptance criteria, it is estimated whether the physical properties of the composite material meet the acceptance criteria. In the present embodiment, when it is estimated that more than half of each piece of data included in the estimation source data 34 meets the acceptance criteria, it is determined that the physical properties of the composite material meet the acceptance criteria. For example, in the example of FIG. 4, when it is estimated that 26 or more of the 51 pieces of data meet the acceptance criteria, it is determined that the physical properties of the composite material meet the acceptance criteria. On the other hand, when the number of data estimated to meet the acceptance criteria is 25 or less out of the 51 pieces of data, it is determined that the physical properties of the composite material do not meet the acceptance criteria. The estimation result is stored in the storage unit 3 as estimation data 35.

[0032] (Estimated Physical Property Presentation Processing Unit 25) The estimated physical property presentation processing unit 25 performs an estimated physical property presentation process for presenting the estimation data 35. In the estimated physical property presentation process, for example, the estimation data 35 is displayed on the display 4. Note that in the estimated physical property presentation process, appropriate information other than the estimation data 35, such as the estimation source data 34, may also be presented together.

[0033] (Method for Estimating Physical Properties of Composite Material) In the method for estimating the physical properties of the composite material according to the present embodiment, prior to the estimation, it is necessary to first collect the learning data 33 by performing test manufacturing or collecting past manufacturing data, and create a learned model. The flow at this time is shown in FIG. 5.

[0034] As shown in FIG. 5, first, in step S11, raw materials are input into the kneader 10 for kneading, and measurements are performed by the AE sensor 111 and the temperature sensor 13. The measurement results by the AE sensor 111 and the temperature sensor 13 are stored in the storage unit 3 as measurement data 31.

[0035] Then, in step S12, a 1-mm-thick sheet sample is produced by press-forming the kneaded material, and in step S13, the sheet sample of the kneaded material is crosslinked by irradiating it with an electron beam or the like to produce a sheet sample of the composite material. Then, in step S14, a tensile test is performed on the sheet sample of the composite material, and the tensile strength is measured as a physical property of the composite material. Then, the process proceeds to step S16.

[0036] On the other hand, after step S11, in step S15, preprocessing is performed on the measurement data 31. In the preprocessing, as shown in FIG. 6, in step S21, the material temperature is sampled at each constant temperature (for example, every 1° C.), and time data synchronized with the sampling is acquired. Then, in step S22, the outputs of the AE sensor 111 for a predetermined period (for example, 16 seconds) are respectively extracted from the acquired time data. Then, in step S23, the outputs of the extracted AE sensor 111 are frequency-analyzed, and in step S24, an intensity spectrum is obtained from the results of the frequency analysis. Then, in step S25, the data of the material temperature and the intensity spectrum are combined and stored in the storage unit 3 as the post-preprocessing data 32. Then, the process returns and proceeds to step S16 in FIG. 5.

[0037] In step S16, it is determined whether the tensile strength obtained in step S14 satisfies the acceptance criteria. If it is determined as NO (N) in step S16, the process returns (returns to step S11). If it is determined as YES (Y) in step S16, in step S17, after adding the post-preprocessing data 32 to the learning data 33, a relationship derivation process is performed in step S18. In the relationship derivation process, the learning data 33 is input to a one-class classifier for learning. Then, the process returns and returns to step S11. The flow in FIG. 5 may be repeated an appropriate number of times to create a learned model.

[0038] (Main Routine) In the method for estimating physical properties of a composite material according to this embodiment, the control flow shown in FIG. 7 is executed. As shown in FIG. 8, first, in step S1, data acquisition processing is performed. In the data acquisition processing, the data acquisition processing unit 21 acquires data from the vibration sensor 11 (AE sensor 111) and the temperature sensor 13, and stores it in the storage unit 3 as measurement data 31.

[0039] After that, in step S2, preprocessing (see FIG. 6) is performed, and in step S26, the preprocessed data 32 obtained by the preprocessing is stored in the storage unit 3 as the estimation source data 34. Note that the preprocessing in step S2 may be performed in real time during kneading or after kneading is completed.

[0040] After that, in step S3, physical property estimation processing is performed. In the physical property estimation processing, as shown in FIG. 8, first, in step S31, each data included in the estimation source data 34 is input into the learned model, and it is determined whether a preset pass criterion is satisfied. After that, in step S32, it is determined whether the number of data determined to satisfy the pass criterion is greater than or equal to the number of data determined not to satisfy the pass criterion. If it is determined YES (Y) in step S32, it is estimated that the pass criterion is satisfied in step S33, and the process proceeds to step S35. If it is determined NO (N) in step S32, it is estimated that the pass criterion is not satisfied in step S34, and the process proceeds to step S35. In step S35, the estimation result is stored in the storage unit 3 as the estimation data 35. After that, the process returns and proceeds to step S4 in FIG. 7.

[0041] In step S4, estimated physical property presentation processing is performed. In the estimated physical property presentation processing, the estimated physical property presentation unit 25 displays the estimation data 35 obtained in step S3 on the display 4. After that, the process ends.

[0042] (Explanation of experimental results) Two types of EVA (ethylene-vinyl acetate copolymer) and a modified polymer were used as the polymers, a metal hydroxide (particle size of about several μm) which is an inorganic filler was used as the compounding agent, and a crosslinking aid was used to knead the composite material. The kneading was performed using a batch kneader, a kneader. During the kneading, first, a mixing process was carried out. In the mixing process, the kneader 10 was preheated, only the polymer was charged and semi-melted, then the powdery compounding agent was charged in two portions, and stirred until a predetermined temperature or a predetermined time to integrate the polymer and the compounding agent. Then, a dispersion process of kneading the material until it became uniform was carried out. In the dispersion process, the compounding agent was finely crushed and dispersed in the polymer, and the compounding agent was uniformly distributed in the polymer. At this time, it is considered that the polymer and the compounding agent adhere as the kneading progresses.

[0043] Here, the kneading was carried out by changing the temperature of the kneader 10 in the mixing process. As a result, even with the same formulation, the physical properties such as the tensile strength change. The temperature of the kneader 10 was set at two levels, conditions 1 and 2. Kneading was carried out 9 times under condition 1 (referred to as conditions 1-1 to 1-9), and 4 times under condition 2 (referred to as conditions 2-1 to 2-4). Condition 1 was set as the condition where the physical properties (here, the tensile strength) of the composite material satisfy the acceptance criteria, and condition 2 was set as the condition that does not satisfy the acceptance criteria.

[0044] The material temperature was measured by a temperature sensor 13 provided at the center of the kneading tank 102 of the kneader 10. Also, the AE sensor 111 measured at 2 MHz and acquired data. As the head of the AE sensor 111, AE-900S-WB manufactured by NF Circuit Design Block Co., Ltd. was used, and as the signal processor for the AE sensor 111, AE9702 manufactured by the same company was used. Also, the tensile strength was measured as a physical property of the composite material. When measuring the tensile strength, the kneaded composite material was press-molded to produce a 1-mm-thick sheet, irradiated with an electron beam for irradiation crosslinking, and then punched into a No. 6 dumbbell shape to obtain a test piece. Then, a tensile test was performed on the test piece using a tensile testing machine (STA-1225 manufactured by A&D Company, Limited) under the condition of 250 mm / min, and the tensile strength was measured. The tensile test results are shown in Fig. 9. Condition 1 (Conditions 1-1 to 1-9) had a high tensile strength and met the passing standard, while Condition 2 (Conditions 2-1 to 2-4) had a low tensile strength and did not meet the passing standard.

[0045] A learned model was created using the measurement data 31 of Conditions 1-5 to 1-9. The measurement data 31 under each condition was preprocessed, and during the preprocessing, the output of the AE sensor 111 for 8 seconds was extracted every 1°C of the material temperature. The frequency width at the time of signal strength calculation (the frequency width of one point when obtaining the intensity spectrum) was set to 10 kHz. Also, for machine learning, a one-class support vector machine was used, and learning was performed in a matlab environment to create a learned model.

[0046] In an embodiment of the present invention, a learned model was created using the material temperature and the strength spectrum for learning. The estimation results for Conditions 1-1 to 1-4 and Conditions 2-1 to 2-4 are collectively shown in FIG. 10(a). Further, the detailed estimation results for Conditions 1-4 and 2-1 (the estimation results of each data included in the estimation source data 34) are shown in FIGS. 10(b) and 10(c). In FIGS. 10(a) to 10(c), for the physical properties of the composite material, those that meet the acceptance criteria are indicated by ○, and those that do not meet the acceptance criteria are indicated by ×. In Condition 1-4 of FIG. 10(b), since the number of data estimated to be ○ (meeting the acceptance criteria) is large, it is estimated to be ○ (meeting the acceptance criteria) as a whole. Similarly, in Condition 2-1 of FIG. 10(c), since the number of data estimated to be × (not meeting the acceptance criteria) is large, it is estimated to be × (not meeting the acceptance criteria) as a whole. As shown in FIG. 10(a), in the embodiment, although there was one incorrect answer in Condition 1, the correct answer rate was very high at 0.875.

[0047] Next, for comparison with the present invention, a comparative example in which a learned model was created using only the strength spectrum for learning without using the material temperature was evaluated in the same manner as in the embodiment. The results are collectively shown in FIG. 11(a). Further, the detailed estimation results for Conditions 1-4 and 2-1 (the estimation results of each data included in the estimation source data 34) are shown in FIGS. 11(b) and 11(c). As shown in FIG. 11(a), it can be seen that in the comparative example, the correct answer rate is 0.625, which is lower than that in the embodiment. From the above results, it was confirmed that using both the material temperature and the output of the AE sensor 111 for learning can improve the estimation accuracy of the physical properties of the composite material compared to using only the output of the AE sensor 111 for learning. This is presumably because the way and frequency of the fracture phenomenon during kneading change depending on the material temperature. Although the recall rate was the same, the precision rate and the f-value were also better in the embodiment.

[0048] (Operations and Effects of the Embodiment) As described above, in the method for estimating the physical properties of the composite material according to the present embodiment, the vibration sensor 11 (AE sensor 111) provided in the kneader 10 for detecting vibration and the temperature sensor 13 as the temperature detection means for directly or indirectly detecting the material temperature, which is the temperature of the composite material being kneaded in the kneader 10, are used. For each predetermined temperature interval of the material temperature detected by the temperature sensor 13, the output of the corresponding AE sensor 111 is extracted, and frequency analysis is performed on the output of the extracted AE sensor 111 to obtain an intensity spectrum. The obtained intensity spectrum is associated with the corresponding material temperature and stored as pre-processed data 32 in a pre-processing step, and using a learned model created in advance using at least the pre-processed data 32 when kneading is performed normally and the physical properties of the composite material (learning data 33) for learning, based on the pre-processed data 32 (estimated source data 34) of the composite material to be estimated, a physical property estimation step of estimating the physical properties of the composite material is provided.

[0049] When a material that becomes plastic at a high temperature, such as a polymer, is used for kneading, the adhesive force at the interface between the polymer and the filler during kneading changes depending on the temperature, and it is considered that the output of the AE sensor 111 also changes. Therefore, by estimating the physical properties of the composite material in consideration of the material temperature in addition to the output of the vibration sensor 11 (AE sensor 111), it becomes possible to further improve the estimation accuracy. By accurately estimating the physical properties of the composite material, an improvement in the quality of the product obtained in the next process, for example, after extrusion processing, can be expected.

[0050] (Modification example) In the above embodiment, a one-class support vector machine is used as the machine learning algorithm, but it is not limited to this. For example, a decision tree, a neural network, or the like may be used. Also, in the above embodiment, unsupervised machine learning is used, but it is not limited to this, and it is also possible to use a classifier of two or more classes as supervised machine learning. Further, the physical properties of the composite material to be estimated are not limited to the tensile strength, and may be, for example, elongation, viscosity, elastic modulus, dynamic viscoelasticity, or the like.

[0051] (Summary of the embodiment) Next, the technical idea grasped from the embodiments described above will be described by referring to the reference numerals and the like in the embodiments. However, each reference numeral and the like in the following description are not limited to the members and the like that specifically show the components in the claims in the embodiments.

[0052] [1] A method for estimating the physical properties of a composite material obtained by kneading at least a polymer and a filler with a kneader (10), comprising: a vibration sensor (11) provided in the kneader (10) for detecting vibration; and temperature detection means (temperature sensor 13) for detecting the material temperature, which is the temperature of the composite material being kneaded in the kneader (10). Using these, for each predetermined temperature interval of the material temperature detected by the temperature detection means (temperature sensor 13), the output of the corresponding vibration sensor (11) is extracted, and frequency analysis is performed on the extracted output of the vibration sensor (11) to obtain an intensity spectrum. The obtained intensity spectrum is associated with the corresponding material temperature and stored as pre-processed data (32). A pre-processing step; and a physical property estimation step of estimating the physical properties of the composite material to be estimated based on the pre-processed data (32) of the composite material to be estimated, using at least the pre-processed data (32) when kneading is normally performed and a learned model created using the physical properties of the composite material for learning. A method for estimating the physical properties of a composite material.

[0053] [2] The learned model is a one-class classifier that has learned the pre-processed data (32) in which the physical properties of the composite material satisfy a preset acceptance criterion. In the physical property estimation step, using the learned model, based on the pre-processed data (32) of the composite material to be estimated, it is estimated whether the physical properties of the composite material satisfy the acceptance criterion. The method for estimating the physical properties of a composite material according to [1].

[0054] [3] In the physical property estimation step, for each data of the intensity spectrum for each material temperature included in the pre-processed data (32), it is estimated whether the physical properties of the composite material satisfy the acceptance criterion. Based on the number or ratio of the data estimated to satisfy or not satisfy the acceptance criterion, it is estimated whether the physical properties of the composite material satisfy the acceptance criterion. The method for estimating the physical properties of a composite material according to [2].

[0055] [4] The method for estimating physical properties of a composite material according to [1], wherein the vibration sensor (11) is an acoustic emission sensor (111) or an acceleration sensor.

[0056] [5] An apparatus for estimating physical properties of a composite material obtained by kneading at least a polymer and a filler with a kneader (10), the apparatus including: a vibration sensor (11) provided in the kneader (10) for detecting vibration; a temperature detection means (temperature sensor 13) for detecting a material temperature which is the temperature of the composite material being kneaded in the kneader (10); a preprocessing unit (22) which, for each predetermined temperature interval of the material temperature detected by the temperature detection means (temperature sensor 13), extracts an output of the corresponding vibration sensor (11), performs frequency analysis on the extracted output of the vibration sensor (11) to obtain an intensity spectrum, and stores the obtained intensity spectrum in association with the corresponding material temperature as preprocessed data (32); and a physical property estimation processing unit (24) which estimates the physical properties of the composite material to be estimated based on the preprocessed data (32) of the composite material to be estimated, using at least the preprocessed data (32) when kneading is performed normally and a learned model created using the physical properties of the composite material for learning. The apparatus for estimating physical properties of a composite material (1) is provided with the above components.

[0057] (Appended Note) The embodiments of the present invention have been described above. However, the embodiments described above do not limit the invention according to the claims. Also, it should be noted that not all combinations of features described in the embodiments are essential means for solving the problems of the invention. Further, the present invention can be appropriately modified and implemented without departing from its gist.

Explanation of Reference Numerals

[0058] 1... Apparatus for estimating physical properties of a composite material 2... Control unit 21... Data acquisition processing unit 22... Preprocessing unit 23... Relationship derivation processing unit 24... Physical property estimation processing unit 25… Presumed physical property presentation processing unit 3… Memory unit 31… Measurement data 32… Data after preprocessing 33… Learning data 34… Data for estimation 35… Estimated data 10… Kneader 11… Vibration sensor 111… AE sensor 12… Arithmetic unit 13… Temperature sensor (temperature detection means)

Claims

1. A method for estimating the physical properties of a composite material obtained by kneading at least a polymer and a filler in a kneader, comprising: a vibration sensor provided in the kneader for detecting vibration; temperature detection means for detecting the material temperature, which is the temperature of the composite material being kneaded in the kneader, and using the same; a preprocessing step of extracting the output of the corresponding vibration sensor at each predetermined temperature interval of the material temperature detected by the temperature detection means, performing frequency analysis on the extracted output of the vibration sensor to obtain an intensity spectrum, and associating the obtained intensity spectrum with the corresponding material temperature and storing it as preprocessed data; a physical property estimation step of estimating the physical properties of the composite material to be estimated based on the preprocessed data of the composite material to be estimated, using at least the preprocessed data when kneading is performed normally and a learned model created using the physical properties of the composite material for learning; A method for estimating the physical properties of a composite material.

2. The learned model is a one-class classifier that has learned the preprocessed data in which the physical properties of the composite material satisfy a preset pass criterion, and in the physical property estimation step, using the learned model, it is estimated whether or not the physical properties of the composite material satisfy the pass criterion based on the preprocessed data of the composite material to be estimated. The method for estimating the physical properties of a composite material according to Claim 1.

3. In the physical property estimation step, for each data of the intensity spectrum for each material temperature included in the preprocessed data, it is estimated whether or not the physical properties of the composite material satisfy the pass criterion, and based on the number or ratio of the data estimated to satisfy or not satisfy the pass criterion, it is estimated whether or not the physical properties of the composite material satisfy the pass criterion. The method for estimating the physical properties of a composite material according to Claim 2.

4. The vibration sensor is an acoustic emission sensor or an acceleration sensor. The method for estimating the physical properties of a composite material according to Claim 1.

5. An apparatus for estimating the physical properties of a composite material obtained by kneading at least a polymer and a filler in a kneader, comprising: a vibration sensor provided in the kneader for detecting vibration; temperature detection means for detecting the material temperature, which is the temperature of the composite material being kneaded in the kneader; For each predetermined temperature interval of the material temperature detected by the temperature detection means, the output of the corresponding vibration sensor is extracted, frequency analysis is performed on the extracted output of the vibration sensor to obtain an intensity spectrum, and the obtained intensity spectrum is associated with the corresponding material temperature and stored as pre-processed data in a pre-processing unit. A physical property estimation processing unit that estimates the physical properties of the composite material based on the pre-processed data of the composite material to be estimated, using at least the pre-processed data when kneading is performed normally and a learned model created using the physical properties of the composite material for learning. A physical property estimation device for a composite material.

Images