Load stress estimation device and method corresponding to fatigue life

Abstract

To provide a device for estimating a relationship between a fatigue life in which a way of machine learning is improved for precise estimation, and a load stress.SOLUTION: A load stress estimation device includes a function portion (310) for reading designation of a type of fatigue limit estimation and designation of mechanical characteristics used in a decision tree of machine learning, a function portion (320) for reading fatigue data of a designated metal structure material from a fatigue data sheet, and a machine learning arithmetic section (330) for performing machine learning with a machine-learned teacher regarding a load stress corresponding to a fatigue life to read fatigue data of a metal structure material by using the designated mechanical characteristics, and an arithmetic result of load stress estimation corresponding to a fatigue life corresponding to a designated type of fatigue limit estimation is outputted by using mechanical characteristics designated by the machine learning arithmetic section that performs the machine learning with a teacher.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to an apparatus and method for estimating applied stress corresponding to a suitable fatigue life for a structure or a structural material, and in particular to a machine learning method and a method for estimating applied stress corresponding to a fatigue life of various structural materials. 7 The present invention relates to an apparatus and method for estimating load stress corresponding to fatigue life using measurement data such as cycle fatigue limit. [Background technology]

[0002] Steel materials are the main components of machines and structures, and fatigue properties are one of the important issues in terms of maintenance and management and response to damage and defects. 7 The fatigue limits of the specimens have been accumulated as NIMS Fatigue Data Sheets (FDS). From these FDS, it is empirically known that there is a correlation between the fatigue limit and other mechanical properties (e.g., Vickers hardness, tensile strength, etc.) (see Figures 19A and 19B). In addition to the fatigue limit, we are also attempting to estimate the fracture life (SN curve) by normalizing the stress amplitude. Patent Document 1 discloses a method for estimating fatigue properties of steel materials. Patent Document 2 discloses a method for analyzing rubber materials to improve the wear resistance of tires. Patent Document 3 discloses a highly accurate classification and analysis method for measurement data using machine learning software.

[0003] FIG. 20 shows the index property proposed by the applicant for the fatigue properties of materials (see Table 1 in Non-Patent Document 1). For simplicity, hereinafter it will be called the index property. An index property is defined as a value that "gives an approximate understanding of the fatigue properties of a material" by referring to the index property of fatigue, just as tensile strength is used as an index to evaluate material strength. In FIG. 20, fatigue is first divided into high cycle fatigue and low cycle fatigue according to the life range. High cycle fatigue life properties are generally expressed by the relationship curve σ between stress amplitude and life. a In this case, the index to be referred to is the strength characteristic, and the static index is the tensile strength σ B , and the dynamic index is the repeated yield stress σyc The reason for this will be explained later.

[0004] On the other hand, the low cycle fatigue life characteristic is shown by the strain-life relationship curve ε a -Nf. Therefore, the index to be referred to is deformation characteristics. In this case, the fracture ductility ε f The dynamic index is the exponent n' of the repeated stress-strain curve (see Non-Patent Document 2). Tensile strength σ B and fatigue limit σ w It is empirically known that there is a good correlation between the yield stress σ y (or 0.2% yield strength σ 0.2 ) and σ w The correlation between has also been investigated, but σ B -σ w This is not as strong as the relationship between σ y However, the repeated yield stress σ yc and σ w This is because a linear relationship exists between the cyclic yield stress σ yc This is thought to be because the internal structure has reached a steady state after repeated plastic strain. Thus, the tensile strength σ B , and the dynamic index is the repeated yield stress σ yc It is considered appropriate to adopt the above.

[0005] Since fatigue occurs due to repeated plastic strain, it is believed that a dynamic index should essentially be adopted. yc There are barriers to adoption. First, yc To measure this, it is necessary to perform strain control testing using the companion specimen method or the incremental step method, but there is a problem in that the number of measurements is not always sufficient. As shown in FIG. 21, the tensile strength σ B and the cyclic yield stress σ ycSince the two index characteristics are proportional to each other, it seems that there is no problem in practical use with the static index. B σ normalized by a / σ B The fatigue properties of the material were evaluated using the tensile strength σ B σ normalized by a / σ B The overall -Nf band was broad, and there was a problem in that it was not an accurate estimate. In addition, in the estimation of the fracture life (SN curve) of the steel structural materials S25C and S55C published in the FDS, the estimation by a regression model linking all decision tree models of Vickers hardness, tensile strength, fracture elongation, and fracture area achieved a high estimation accuracy of 92.0% in the case of training data in which the fracture data of S25C and S55C were distinguished. However, as shown in Figures 7 and 8 of Non-Patent Document 1, there was a problem that the estimation accuracy decreased, with the regression rate of the estimated value being 65.8% and the mean absolute error rate MAPE being 38.7%, because the fracture data of S25C and S55C were not distinguished in the randomly extracted test data. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2017-187408 A [Patent Document 2] Patent Publication No. 2021-071803 [Patent Document 3] WO2018-207524 publication [Non-patent literature]

[0007] [Non-Patent Document 1] Nobuo Nagashima, Masao Hayakawa, Hiroyuki Masuda, and Hisashi Nagai, "Estimation of SN curves using machine learning random forest method," Materials Science, Vol. 70, pp. 876-880 (2021) [Non-Patent Document 2] S. Matsuoka, N, Nagashima and S. Nishijima, “Index property for the fatigue of engineering alloys”, NIMS Materials Strength Data Sheet Technical Document, No.17(1997) Summary of the Invention [Problem to be solved by the invention]

[0008] In the method of estimating the fatigue SN diagram by machine learning as shown in Non-Patent Document 1, when estimating the fatigue life from the applied stress during machine learning, there was a problem that the accuracy decreased if the data on the long life side was also used for training. Also, while it was known that the accuracy of estimating the fatigue limit was improved by taking into account the effects of the loading type and stress ratio, the accuracy of estimating the fatigue life was poor, and the effects of the loading type and stress ratio were unknown. The present invention solves the problems of the conventional technology described above, and aims to provide an apparatus for estimating the relationship between fatigue life and applied stress by improving the machine learning method so as to obtain accurate estimations when estimating the SN curve (relationship between stress amplitude and fracture life) using machine learning. [Means for solving the problem]

[0009] The inventors of the present invention came up with the idea that the accuracy of estimating the relationship between fatigue life and applied stress may be improved by improving the method of machine learning when estimating the relationship between stress amplitude and rupture life using experimental data available from the NIMS fatigue data sheet. 7 The accuracy of estimating the fatigue limit and the estimated value of the SN curve are 6 This is to verify the probability that the difference between the estimated fracture life of less than 1000 times and the actual measured value falls within the acceptable range.

[0010] [1] The load stress estimation device for fatigue life according to the present invention includes, as shown in FIG. 3, a function unit (310) for reading the specification of the type of fatigue limit estimation and the specification of the mechanical properties used in the machine learning decision tree; A functional unit (320) for reading fatigue data of a specified metallic structural material from a fatigue data sheet; A machine learning calculation unit (330) that performs supervised machine learning of a load stress corresponding to a fatigue life on the fatigue data of the loaded metal structural material using the specified mechanical characteristics, The machine learning calculation unit that performed the supervised machine learning uses the specified mechanical characteristics to output a calculation result of a load stress estimation corresponding to a fatigue life according to a specified type of fatigue limit estimation. [2] In the load stress estimation device corresponding to fatigue life [1] of the present invention, it is preferable that the machine learning calculation unit (330) uses the specified mechanical properties to perform machine learning of the load stress corresponding to the fatigue life for the fatigue data of the loaded metal structural material, and has a pre-processing unit (325) that performs the supervised machine learning. [3] In the load stress estimation device [2] corresponding to fatigue life of the present invention, preferably, the model used for the machine learning includes a model using a random forest method, deep learning, a neural network, or LASSO regression.

[0011] [4] In the load stress estimation device for fatigue life according to the present invention [1] to [3], preferably, the type of fatigue limit estimation is 10 7 Estimation of fatigue limit, 10 6 This may include estimation of the fatigue limit, estimation of the SN curve, and estimation of the low cycle fatigue life. [5] In the load stress estimation device for fatigue life according to the present invention [1] to [4], the mechanical properties used in the machine learning decision tree preferably include Vickers hardness, tensile strength, breaking elongation, breaking reduction, stress ratio, or fatigue test method. [6] In the load stress estimation device for fatigue life according to the present invention [5], preferably, the fatigue testing method includes at least one of rotating bending test data, axial load test data, and torsion test data. [7] In the load stress estimation device corresponding to fatigue life of the present invention [1] to [6], preferably, the metal structural material includes at least one of carbon steel, nickel chromium molybdenum steel, manganese steel, stainless steel, aluminum alloy, or titanium alloy. [8] In the load stress estimation device [7] corresponding to fatigue life of the present invention, preferably, the metal structural material includes, for example, S25C, S35C, S45C, S55C for carbon steel materials, SNCM439 for nickel-chromium-molybdenum steel materials, SMn438, SMn443 for manganese steel materials, and SUS403, SUS304 for stainless steel materials.

[0012] [9] In the applied stress estimation device corresponding to fatigue life [1] to [8] of the present invention, it is preferable to have a machine learning evaluation unit (335) that evaluates the accuracy of the applied stress estimation corresponding to the fatigue life of the machine learning calculation unit (330) that performed supervised machine learning, using the calculation result of the applied stress estimation corresponding to the fatigue life according to the type of fatigue limit estimation used for the calibration target.

[10] In the load stress estimation device for fatigue life [9] of the present invention, it is preferable to have an assist function (340) that recommends, based on the evaluation results of the machine learning evaluation unit (335), the type of mechanical properties to be used in the machine learning decision tree and the type of fatigue data of the metal structural material to be read in the fatigue data sheet, so as to improve the accuracy of the load stress estimation for fatigue life by the machine learning calculation unit (330).

[0013]

[11] The method for estimating applied stress corresponding to fatigue life of the present invention includes, for example, a step (S400) of specifying a type of fatigue limit estimation and a mechanical characteristic to be used in a machine learning decision tree, as shown in FIG. A step (S405) of reading fatigue data of the specified metal structural material from a fatigue data sheet; The method further includes a step (S415) of outputting, using the specified mechanical characteristics, a calculation result of a load stress estimation corresponding to a fatigue life according to a specified type of fatigue limit estimation to a machine learning calculation unit (330) that has performed supervised machine learning of a load stress corresponding to a fatigue life for the fatigue data of the loaded metal structural material using the specified mechanical characteristics.

[12] In the method for estimating applied stress corresponding to fatigue life

[11] of the present invention, it is preferable to have a step (S410) of performing supervised machine learning by using specified mechanical properties to perform machine learning of applied stress corresponding to fatigue life on fatigue data of the loaded metal structural material.

[13] In the method for estimating load stress corresponding to fatigue life of the present invention

[12] , preferably, the model used for the machine learning includes a model using a random forest method, deep learning, a neural network, or LASSO regression.

[0014]

[14] In the method for estimating applied stress corresponding to fatigue life according to the present invention

[11] to

[13] , preferably, the type of fatigue limit estimation is 10 7 Estimation of fatigue limit, 10 6 This may include estimation of the fatigue limit, estimation of the SN curve, and estimation of the low cycle fatigue life.

[15] In the load stress estimation method

[11] to

[14] corresponding to the fatigue life of the present invention, preferably, the mechanical properties used in the machine learning decision tree include Vickers hardness, tensile strength, breaking elongation, breaking reduction, stress ratio, or fatigue test method.

[16] In the method for estimating applied stress corresponding to fatigue life according to the present invention

[15] , preferably, the fatigue testing method includes at least one of rotating bending test data, axial load test data, and torsion test data.

[17] In the load stress estimation method corresponding to fatigue life of the present invention

[11] to

[16] , preferably, the metal structural material includes at least one of carbon steel, nickel chromium molybdenum steel, manganese steel, stainless steel, aluminum alloy, or titanium alloy.

[18] In the method

[17] for estimating load stress corresponding to fatigue life of the present invention, preferably, the metal structural material includes, for example, S25C, S35C, S45C, and S55C for carbon steel materials, SNCM439 for nickel-chromium-molybdenum steel materials, SMn438 and SMn443 for manganese steel materials, and SUS403 and SUS304 for stainless steel materials.

[0015]

[19] In the method of estimating applied stress corresponding to fatigue life

[11] to

[18] of the present invention, it is preferable to have a step (S425) of evaluating the accuracy of the applied stress estimation corresponding to fatigue life of the machine learning calculation unit (330) that performed supervised machine learning, using the calculation result of the applied stress estimation corresponding to fatigue life according to the type of fatigue limit estimation used for the calibration object.

[20] In the method of estimating applied stress corresponding to fatigue life

[19] of the present invention, it is preferable to include a step (S430) of recommending, based on the evaluation results of the machine learning evaluation step (S425), the type of mechanical properties to be used in the machine learning decision tree and the type of fatigue data of the metal structural material to be read in the fatigue data sheet, so as to improve the accuracy of the applied stress estimation corresponding to the fatigue life of the machine learning calculation unit (330).

[21] In the load stress estimation method

[20] corresponding to a fatigue life of the present invention, it is preferable to have a step (S435) of revising the type of fatigue limit estimation specified for use in S400 of the load stress estimation method

[11] corresponding to the fatigue life and the mechanical properties used in the machine learning decision tree, taking into account the type of mechanical properties used in the recommended machine learning decision tree and the metal structural material read in the fatigue data sheet. Effect of the Invention

[0016] According to the present invention, the load stress estimation device and method for estimating fatigue life is used to estimate the load stress of various metal structural materials by random forest method using the data of the fatigue data sheet. 7 fatigue limit and 10 6They attempted to estimate the fracture life in cycles or less and also investigated the possibility of estimating the SN curve. They found that by using machine learning to predict the load stress corresponding to the fatigue life, the estimation error in fatigue limit estimation was significantly reduced compared to the conventional method of predicting fatigue life from the load stress. In addition, according to the device and method for estimating applied stress corresponding to fatigue life of the present invention, the applied stress corresponding to fatigue life is predicted using data on various metallic structural materials contained in a fatigue data sheet, so the number of fatigue data used for supervised machine learning can be much larger than when fatigue data on individual metallic structural materials is used, and the accuracy of the applied stress prediction corresponding to fatigue life using machine learning is significantly improved. To obtain fatigue data on individual metallic structural materials, for example, 10 7 At least 10 times the fatigue limit estimation 7 Since a reciprocating load must be applied to the test material multiple times, the cost of obtaining one piece of data is very high. [Brief description of the drawings]

[0017] [Figure 1] 1 is a block diagram showing an example of a schematic configuration of a fatigue limit estimation system according to an embodiment of the present invention. FIG. [Diagram 2] FIG. 2 is a block diagram showing an exemplary computing device 200 in the case where a fatigue life estimation calculation processing unit of the device shown in FIG. 1 is configured using a computer. [Diagram 3] FIG. 3 is a functional block diagram of software for a computer having the functional blocks shown in FIG. 2. [Figure 4] FIG. 4 is an explanatory diagram of a fatigue limit estimation algorithm of the device based on the functional block diagram of the software shown in FIG. 3. [Diagram 5] FIG. 1 is a graph showing a comparison between predicted and actually measured values ​​of applied stress corresponding to the fatigue life of S45C steel (rotating bending, R=-1) according to an embodiment of the present invention. [Figure 6] FIG. 1 is a graph showing a comparison between predicted and actually measured fatigue life values ​​corresponding to applied stress for S45C steel (rotating bending, R=−1) as a comparative example of the present invention. [Figure 7]FIG. 1 is a graph showing a comparison between predicted and actually measured values ​​of applied stress corresponding to the fatigue life of S45C steel (torsion, R=-1) according to an embodiment of the present invention. [Figure 8] FIG. 1 is a graph showing a comparison between predicted and actually measured values ​​of load stress corresponding to the fatigue life of SUS304 stainless steel (rotating bending, R=-1) according to an embodiment of the present invention. [Figure 9] FIG. 1 is a graph showing a comparison between predicted and actually measured values ​​of applied stress corresponding to the fatigue life of an aluminum alloy (7075) (axial load, R=-1) according to an embodiment of the present invention. [Figure 10] FIG. 1 is a graph showing a comparison between predicted and actually measured values ​​of applied stress corresponding to the fatigue life of a titanium alloy (Ti64ELI-900) (axial load, R=0.3) according to an embodiment of the present invention. [Figure 11] This figure compares the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel in one embodiment of the present invention, and shows a case in which the load type and stress ratio are taken into account in the decision tree model machine learning section, but tensile strength, hardness, elongation, and reduction in area are not taken into account. [Figure 12] FIG. 1 is a diagram showing a comparison between the load stress prediction and the actual measured value corresponding to the fatigue life of S45C steel according to one embodiment of the present invention, and shows a case where the elongation, reduction in area, loading type, and stress ratio are taken into account in the decision tree model machine learning section. [Figure 13] FIG. 1 is a diagram showing a comparison between the load stress prediction and the actual measured value corresponding to the fatigue life of S45C steel according to one embodiment of the present invention, and shows a case where the tensile strength, loading type, and stress ratio are taken into account in the decision tree model machine learning section. [Figure 14] FIG. 1 is a diagram showing a comparison between the load stress prediction and the actual measured value corresponding to the fatigue life of S45C steel according to one embodiment of the present invention, and shows a case where the tensile strength, hardness, loading type, and stress ratio are taken into account in the decision tree model machine learning section. [Figure 15] This figure compares the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel in one embodiment of the present invention, and shows a case in which the decision tree model machine learning section takes into account tensile strength, hardness, elongation, reduction in area, loading type, and stress ratio. [Figure 16]This figure compares the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, showing one embodiment of the present invention, and shows a case in which the decision tree model machine learning section takes into account tensile strength, hardness, elongation, and reduction in area, but does not take into account the loading type and stress ratio. [Figure 17] This figure compares the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, showing one embodiment of the present invention, and shows the case where the decision tree model machine learning section takes into account tensile strength, hardness, elongation, reduction in area, and stress ratio, but does not take into account the loading type. [Figure 18] This figure compares the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, showing one embodiment of the present invention, and shows the case where the decision tree model machine learning section takes into account tensile strength, hardness, elongation, reduction in area, and loading type (rotating bending, axial load, torsion), but does not take into account the stress ratio. [Figure 19A] FIG. 1 is a diagram showing the relationship between mechanical properties and fatigue limit, showing Vickers hardness. [Figure 19B] FIG. 1 is a diagram showing the relationship between mechanical properties and fatigue limit, showing tensile strength. [Figure 20] FIG. 1 is a diagram showing an index property proposed by the present applicant that serves as an index for the fatigue properties of a material. [Figure 21] FIG. 1 is a diagram showing the relationship between two index characteristics, tensile strength σB and repeated yield stress σyc. [Figure 22] This is a graph showing the SN curve normalized by tensile strength. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] The present invention will now be described with reference to the drawings. FIG. 1 is a block diagram showing an example of a schematic configuration of a fatigue limit estimation system according to an embodiment of the present disclosure.

[0019] The fatigue limit estimation system 10 includes a fatigue testing machine 20 and a fatigue limit estimation device 30. The fatigue testing machine 20 includes a testing machine main body 22, a power supply device 24, and a controller (not shown). The fatigue limit estimation device 30 includes an estimation calculation processing unit 32, and an extensometer 34. The controller of the fatigue testing machine 20 and the extensometer 34 are connected to the estimation calculation processing unit 32 of the fatigue limit estimation device 30.

[0020] The fatigue testing machine 20 is fitted with a test piece SP for testing the fatigue limit, and performs a fatigue test by repeatedly applying a predetermined load. The fatigue limit estimation device 30 uses an extensometer 34 to measure the elongation of the test piece SP to which a load is applied by the fatigue testing machine 20, and estimates the fatigue limit.

[0021] Next, an example of the hardware configuration of the estimation calculation processing unit 32 constituting the fatigue limit estimation device 30 will be described. Fig. 2 is a block diagram showing an exemplary computing device 200 in the case where the fatigue life estimation processing unit of the device shown in Fig. 1 is configured using a computer. The fatigue life estimation processing unit 32 of Fig. 1 can be implemented using all or a part of the computing device 200. In a very basic configuration 201, a computing device 200 typically includes one or more processors 210 and a system memory 220. A memory bus 230 may be used for communication between the processor 210 and the system memory 220.

[0022] Depending on the desired configuration, the processor 210 may be of any type, including but not limited to a microprocessor (μP), a microcontroller (μC), a digital signal processor (DSP), or a combination thereof. The processor 210 may include another level of caching, such as a level 1 cache 211 and a level 2 cache 212, a processor core 213, and registers 214. An exemplary processor core 213 may include an arithmetic logic unit (ALU), a floating point unit (FPU), a digital signal processing core (DSP core), or any combination thereof, etc. An exemplary memory controller 215 may also be used with the processor 210, or in some implementations, the memory controller 215 may be an internal part of the processor 210.

[0023] Depending on the desired configuration, the system memory 220 can be of any type, including but not limited to, volatile memory (such as RAM), non-volatile memory (such as ROM, flash memory, etc.), or any combination thereof. The system memory 220 can include an operating system 221, one or more applications 222, and a decision tree model machine learning unit 232. The applications 222 may run on a 10 7 Estimation of fatigue limit 223, 10 6 The process may include a cycle fatigue limit estimation unit 224, an SN curve estimation unit 225, and a low cycle fatigue life estimation unit 226.

[0024] The decision tree model machine learning unit 232 may include, as a decision tree model used in a random forest method, which is a type of machine learning, Vickers hardness 233, tensile strength 234, breaking elongation 235, breaking reduction 236, stress ratio 237, and fatigue test method 238. The fatigue test method 238 includes, for example, a rotating bending test method, an axial load test method, and a torsion test method.

[0025] The computing device 200 may have additional features or functionality and additional interfaces to facilitate communication between the basic configuration 201 and any necessary devices and interfaces. For example, a bus / interface control 240 may be used to facilitate communication between the basic configuration 201 and one or more data storage devices 250 via a storage interface bus 241. The data storage device 250 may be a removable storage device 251, a non-removable storage device 252, or a combination thereof. Examples of removable and non-removable storage devices include magnetic disk drives such as floppy disk drives and hard disk drives (HDDs), optical disk drives such as compact disk (CD) drives or digital versatile disk (DVD) drives, solid state drives (SSDs), and tape drives. Exemplary computer storage media may include volatile and non-volatile, removable and fixed media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data.

[0026] System memory 220, removable storage 251, and non-removable storage 252 are all examples of computer storage media. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices. Any such computer storage media that can be used to store desired information and that can be accessed by computing device 200 may be part of another computing device 290.

[0027] The computing device 200 may also include an interface bus 242 to facilitate communication from various interface devices (eg, output interfaces, peripheral interfaces, and communications interfaces) to the basic configuration 201 via a bus / interface control unit 240 . In the output device 260, the image processing unit 261 and the audio processing unit 262 may be configured to communicate with various external devices, such as a display device 291 or speakers, via one or more AV ports 263.

[0028] The exemplary peripheral interface 270 includes a serial interface control unit 271 or a parallel interface control unit 272 that may be configured to communicate with external devices such as input devices (e.g., keyboard, mouse, pen, voice input device, touch input device, etc.) The peripheral interface 270 may be configured to communicate with a fatigue testing machine 292 and an external database device storing fatigue data sheets 293 via an I / O port 273. The exemplary communications device 280 includes a network controller 281, which may be configured to facilitate communications with one or more other computing devices 290 over a network communications link via one or more communications ports 282. The fatigue data sheets 293 include a data sheet 294 for carbon steel materials, a data sheet 295 for nickel-chromium-molybdenum steel materials, a data sheet 296 for manganese steel materials, a data sheet 297 for stainless steel materials, a data sheet 298 for aluminum alloys, and a data sheet 299 for titanium alloys. For carbon steel materials, there are fatigue data sheets for various metal structural materials such as S25C, S35C, S45C, and S55C, which are carbon steel materials for machine structures specified in JIS G 4051, and these are available as FDS-No. 1 to 16, etc. provided by the present applicant.

[0029] A network communication link may be an example of a communication medium. A communication medium may typically be embodied by computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transport mechanism, and may include any information delivery media. A "modulated data signal" may be a signal that has one or more of its characteristics set or changed in such a manner as to encode information in the signal. By way of example, and not limitation, communication media may include wired media, such as a wired network or direct-wired connection, and wireless media, such as acoustic, radio frequency (RF), microwave, infrared (IR) and other wireless media. As used herein, the term computer-readable media may include both storage media and communication media.

[0030] Computing device 200 may be implemented as part of a small form factor portable (or mobile) electronic device such as a mobile telephone, a personal data assistant (PDA), a personal media player device, a wireless web watch device, a personal computer, a headset device including any of the above functionality, a special purpose device, or a hybrid device. Computing device 200 may also be implemented as a personal computer, including both laptop and non-laptop computer configurations.

[0031] Figure 3 is a functional block diagram of software for a computer having the functional blocks shown in Figure 2, illustrating an exemplary computer program product 300. The program bearing medium 302, which may be implemented as a computer readable medium 306, a recordable medium 308, a communication medium 309, or a combination thereof, has program instruction storage 304 that may be configured to perform all or a portion of the processing of a processing unit.

[0032] The program instructions stored in the program instruction storage unit 304 include, for example, a function (310) for reading the specification of the type of fatigue limit estimation and the specification of the mechanical properties used in the machine learning decision tree, a function (320) for reading fatigue data of a specified metal structural material from a fatigue data sheet, and a pre-processing unit (325) for performing machine learning by the random forest method on the fatigue data of the loaded metal structural material using the specified mechanical properties, and performing supervised machine learning of the load stress corresponding to the fatigue life.Furthermore, the program instruction includes a machine learning calculation unit (330) for outputting the calculation result of the estimation of the load stress corresponding to the fatigue life according to the specified type of fatigue limit estimation using the specified mechanical properties to the machine learning calculation unit (330) that performed the supervised machine learning. In addition, the random forest method in machine learning may be a model using the well-known random forest method, which uses multiple decision trees to perform "classification" or "regression". In addition, the model used in machine learning is not limited to a model using the random forest method, but may include, for example, a model using a neural network, such as the well-known deep learning, and a model using LASSO regression. In addition, the regression equation for fatigue limit estimation is not limited to a linear function, but may also be a nonlinear function using a polynomial, kriging, or RBF (Radial Base Function).

[0033] The program instructions stored in the program instruction storage unit 304 also have a machine learning evaluation unit (335) that evaluates the accuracy of the fatigue limit estimation by the machine learning calculation unit (330) that performed supervised machine learning, using the calculation result of the fatigue limit estimation according to the type of fatigue limit estimation used for the calibration target, for example. Furthermore, the program instructions have an assist function (340) that recommends the type of mechanical properties to be used in the machine learning decision tree and the type of fatigue data of the metal structural material to be read in the fatigue data sheet, based on the evaluation result of the machine learning evaluation unit (335), so as to improve the accuracy of the fatigue limit estimation by the machine learning calculation unit (330).

[0034] FIG. 4 is an explanatory diagram of the fatigue limit estimation algorithm of the device based on the functional block diagram of the software shown in FIG. 3, where (A) is the basic algorithm and (B) is an additional algorithm. First, in the basic algorithm, the type of fatigue limit estimation and the mechanical properties to be used in the machine learning decision tree are specified (S400). 7 Estimation of fatigue limit, 10 6 These include estimation of cycle fatigue limit, estimation of SN curve, and estimation of low cycle fatigue life. The mechanical properties used in the machine learning decision tree include Vickers hardness, tensile strength, breaking elongation, breaking reduction, stress ratio, and fatigue test method.

[0035] Next, the fatigue data of the specified metal structural material is read from the fatigue data sheet (S405). The metal structural materials include data sheets for carbon steel, nickel chrome molybdenum steel, manganese steel, stainless steel, aluminum alloy, and titanium alloy, and among the metal structural materials, the steel types of the subdivided low alloy steels include, for example, S25C, S35C, S45C, and S55C for carbon steel, SNCM439 for nickel chrome molybdenum steel, SMn438 and SMn443 for manganese steel, and SUS403 and SUS304 for stainless steel. Using the specified mechanical properties, machine learning is performed by the random forest method on the fatigue data of the loaded metal structural material, and supervised machine learning of the load stress corresponding to the fatigue life is performed (S410). Note that if supervised machine learning has already been performed, this step may be omitted. Next, the specified mechanical characteristics are used to output the calculation result of the load stress estimation corresponding to the fatigue life according to the specified type of fatigue limit estimation to the machine learning calculation unit (330) that performed supervised machine learning (S415).

[0036] The additional algorithm added to the basic algorithm includes the following steps: Using the calculation results of the fatigue limit estimation according to the type of fatigue limit estimation used for the calibration target, the accuracy of the load stress estimation corresponding to the fatigue life of the machine learning calculation unit (330) that performed the supervised machine learning is evaluated (S425). In addition, based on the evaluation results of the machine learning evaluation unit (335), the type of mechanical properties to be used in the machine learning decision tree and the type of fatigue data of the metal structural material to be read in the fatigue data sheet are recommended (S430) so as to increase the accuracy of the load stress estimation corresponding to the fatigue life of the machine learning calculation unit (330). Taking into account the type of mechanical properties used in the recommended machine learning decision tree and the metal structural material loaded in the fatigue data sheet, the type of fatigue limit estimation specified for use in S400 and the mechanical properties used in the machine learning decision tree are modified (S435).

[0037] Next, we will explain the supervised machine learning used in the fatigue limit estimation algorithm mentioned above. The random forest method is one of the machine learning algorithms, and is an ensemble learning algorithm that improves generalization ability by integrating weak learners of multiple decision tree models, and is mainly used for classification (discrimination) and regression (estimation). What is important here is that (i) more accurate data can be sampled for the target data group, and (ii) a decision tree model is created for each learning element. Previously, regression in mathematical models was performed using least-squares approximation to determine the correlation between the two target data groups, but machine learning can create regression models that link decision tree models of multiple learning elements, making it possible to make even more accurate estimations by appropriately selecting multiple learning elements for the decision tree model.

[0038] 10 7The data set used to estimate the cycle fatigue limit was based on experimental data from rotating bending fatigue tests on S45C (FDS-No.3), a carbon steel material for machine structures specified in JIS G 4051. The effect of each learning element was examined using the Random Forest method. Next, fatigue limit data from torsional fatigue tests was added to the data set to examine the effects of differences in fatigue test methods. Furthermore, fatigue test data for stress ratios R=-1 and R=0 were added to examine the effects of stress ratio.

[0039] Low alloy steel 10 7 The data set used to estimate the cycle fatigue limit includes low alloy steel S45C (FDS-No.3), carbon steel for machine construction S25C (FDS-No.1), S35C (FDS-No.2), S55C (FDS-No.4), nickel chrome molybdenum steel SNCM439 (FDS-No.25), manganese steel SMn438 (FDS-No.16), and manganese steel SMn443 (FDS-No.17).Fatigue test data for metal materials other than low alloy steel includes fatigue data for various stainless steel materials, such as martensitic stainless steel SUS403 (FDS-No.30) for stainless steel pipes for machine construction specified in JIS G 3446, and austenitic stainless steel SUS304 (FDS-No.33). The number of fatigue data points for S45C (FDS-No. 3) is 999, but by expanding this to include carbon steel materials for mechanical structures S25C (FDS-No. 1), S35C (FDS-No. 2), and S55C (FDS-No. 4), the number of data points increases by approximately 3.4 times to 3,471, by expanding this to low alloy steels, the number of data points increases by approximately 6.8 times to 6,865, and by expanding this to include fatigue test data for metal materials other than low alloy steels, the number of data points increases by approximately 7.6 times to 7,616, which is favorable for machine learning.

[0040] For the machine learning, a commercially available personal computer was used as the computing device 200, and a system having software functional blocks including the fatigue limit estimation algorithm shown in Figs. 2 to 4 was constructed using Python (registered trademark) 3.6.1 and an external library, Anaconda (registered trademark).

[0041] The data used were fatigue test results listed on the fatigue data sheet. To ensure fairness in the analysis, 80% was used as training data and the remaining 20% ​​was used as test data, which were randomly selected each time. Therefore, it is not possible to determine which data corresponds to which test data. In addition, as one way of evaluating the analysis results, the mean absolute percentage error (MAPE) was calculated from the analysis results using the test data.

number

number

[0042] Indices for evaluating the goodness of fit of a model obtained by regression analysis include the root mean square error (RMSE), mean squared error (RSE), and coefficient of determination R 2 However, when calculated using the error functions of RMSE and MSE, the positive and negative data are added together, resulting in an average error that is offset. On the other hand, since MAPE is an absolute value, it can localize discrepancies in the predicted data. Problems with MAPE include cases where the actual measured value is 0 and the predicted value is too small, but the actual measured value and predicted value targeted by this invention are fatigue limits, and the above cases do not apply. In addition, it is thought that a biased conclusion will be obtained if cross-validation and grid search are not performed, but in all predictions, by creating and visually inspecting the relationship between the experimental value and the predicted value, as shown in Figures 5 and 6, it is believed that this serves as a substitute for cross-validation and grid search. For these reasons, it was considered appropriate to use MAPE as an error function in this invention rather than RMSE and MSE. EXAMPLES

[0043] FIG. 5 is a diagram showing a comparison between the load stress prediction and the actual measured value corresponding to the fatigue life of S45C (S45C-550F) steel (rotating bending, R=-1) according to one embodiment of the present invention, in which (A) shows the relationship between the load stress by AI prediction and the load stress by experiment, (B) shows the relationship between the fracture life and the stress amplitude by the actual measured value, and (C) shows the relationship between the fracture life and the stress amplitude by AI prediction and the actual measured value. In FIG. 5(A), training data is indicated by "+" and test data is indicated by "▲". The same applies to the following FIG. 6(A), FIG. 7(A), FIG. 8(A), FIG. 9(A), and FIG. 10(A). In addition, the AI ​​prediction here refers to the predicted value by the supervised machine learning of the present invention. In addition, in Fig. 5(B), the actual measured data is represented by "◆", and in Fig. 5(C), the AI ​​predicted data is represented by "■" and the actual measured data is represented by "◆". The same applies to the following Figs. 6(B), 6(C), 7(B), 7(C), 8(B), 8(C), 9(B), 9(C), 10(B), and 10(C). Comparative Example 1

[0044] FIG. 6 is a graph comparing fatigue life predictions and actual measured values ​​corresponding to applied stress for S45C (S45C-550F) steel (rotating bending, R=-1), which is a comparative example of the present invention. FIG. 6A is a graph showing the relationship between fatigue life predictions based on AI predictions and actual measured fatigue life values ​​from experiments. FIG. 6B is a graph showing the relationship between fracture life and stress amplitude based on actual measured values. FIG. 6C is a graph showing the relationship between fracture life and stress amplitude based on AI predictions and actual measured values. In Comparative Example 1, the fatigue life corresponding to the applied stress is used as the AI ​​prediction, which is equivalent to the technical content disclosed in Non-Patent Document 1.

[0045] In Comparative Example 1, which estimates fatigue life from applied stress, and Example 1, which estimates applied stress from fatigue life, the average error accuracy of Comparative Example 1 shown in Fig. 6(A) is 72.9%, and the average error accuracy of Example 1 shown in Fig. 5(A) is 2.3%, demonstrating that the accuracy is significantly improved by using the prediction method that assumes the applied stress corresponding to the fatigue life used in Example 1. It can be seen that the average error accuracy of Comparative Example 1 varies greatly. In addition, in the estimation of the fatigue life from the load stress of Comparative Example 1 shown in FIG. 6(C), for example, the number of repetitions N until fracture is 4×10 6 The predicted stress amplitude for the test was 510MPa, 2x10 6 In contrast, the measured values ​​at a stress amplitude of 510 MPa shown in FIG. 6(B) are 0.4, 0.5, and 1.0x10 6 The AI ​​predicted value of Comparative Example 1 predicts a fatigue life that is about 4 to 10 times longer than the actual measured value. In contrast, in the load stress prediction corresponding to the fatigue life shown in Fig. 5(C), for example, the number of cycles N until fracture is 1x10 6 The predicted stress amplitude for the test was 514MPa, 0.6x10 6 In contrast, the measured values ​​at a stress amplitude of 510 MPa shown in Figure 5(B) are 0.4, 0.5, and 1.0x10 6 The AI ​​predicted value of Example 1 predicts the load stress corresponding to the fatigue life with a higher degree of agreement with the actual measured value. EXAMPLES

[0046] FIG. 7 is a graph comparing the load stress prediction and actual measurement values ​​corresponding to the fatigue life of S45C steel (torsion, R=-1) showing one embodiment of the present invention, in which (A) is a graph showing the relationship between the AI-predicted load stress and the experimental load stress, (B) is a graph showing the relationship between the fracture life and stress amplitude based on the actual measured values, and (C) is a graph showing the relationship between the AI-predicted load stress and the actual measured values ​​and stress amplitude. In the load stress prediction corresponding to the fatigue life shown in Fig. 7(C), for example, the number of cycles N until fracture is 2.0x10 7 The predicted stress amplitude for the test was 318MPa, 1.1x10 7 In contrast, the measured values ​​at a stress amplitude of 310 MPa shown in FIG. 7(B) are 0.6, 0.8, and 2.0x10 7 The actual measured value at a stress amplitude of 320 MPa was 1.0x10 7 The AI ​​predicted value in Example 2 predicts the load stress corresponding to the fatigue life with a high degree of agreement with the actual measured value. EXAMPLES

[0047] FIG. 8 is a graph comparing the load stress prediction and actual measurement values ​​corresponding to the fatigue life of SUS304 stainless steel (rotating bending, R=-1) showing one embodiment of the present invention, where (A) is a graph showing the relationship between the load stress predicted by AI and the load stress obtained by experiment, (B) is a graph showing the relationship between the fracture life and stress amplitude based on actual measurements, and (C) is a graph showing the relationship between the fracture life and stress amplitude based on AI prediction and actual measurements. In the load stress prediction corresponding to the fatigue life shown in Fig. 8(C), for example, the number of cycles N until fracture is 4.8x10 5 The predicted stress amplitude for the test was 297MPa, 2.0x10 5 In contrast, the measured values ​​at a stress amplitude of 300 MPa shown in Figure 8(B) are 0.8, 1.8, and 4.8x10 5 The actual measured values ​​at a stress amplitude of 310 MPa were 0.9 and 1.2x10 5 The AI ​​predicted value in Example 3 predicts the load stress corresponding to the fatigue life with a high degree of agreement with the actual measured value. EXAMPLES

[0048] FIG. 9 is a graph comparing the load stress prediction and actual measurement values ​​corresponding to the fatigue life of an aluminum alloy (7075) (axial load, R=-1) showing one embodiment of the present invention, where (A) is a graph showing the relationship between the load stress predicted by AI and the load stress obtained by experiment, (B) is a graph showing the relationship between the fracture life and stress amplitude based on actual measurements, and (C) is a graph showing the relationship between the fracture life and stress amplitude based on AI prediction and actual measurements. In the load stress prediction corresponding to the fatigue life shown in Fig. 9(C), for example, the number of cycles N until fracture is 2.6, 3.0, and 3.6x10 7 The predicted stress amplitudes for the times were 200MPa, 4.0, and 5.0x10 6 In contrast, the measured values ​​at a stress amplitude of 200 MPa shown in Figure 9(B) are 2.6, 3.0, and 3.6x10 7 The actual measured values ​​at a stress amplitude of 220 MPa were 4.0 and 5.0x10 6 The AI ​​predicted value of Example 4 predicts the load stress corresponding to the fatigue life with a very high degree of agreement with the actual measured value. EXAMPLES

[0049] FIG. 10 is a graph comparing the load stress prediction and actual measured values ​​corresponding to the fatigue life of a titanium alloy (Ti64ELI-900) (axial load, R=0.3) showing one embodiment of the present invention, where (A) is a graph showing the relationship between the load stress predicted by AI and the load stress obtained by experiment, (B) is a graph showing the relationship between the fracture life and stress amplitude based on actual measurements, and (C) is a graph showing the relationship between the fracture life and stress amplitude based on AI prediction and actual measurements. In the load stress prediction corresponding to the fatigue life shown in Fig. 10(C), for example, the number of cycles N until fracture is 8.5x10 7 The predicted stress amplitude for the test was 201MPa, 6.2x10 7 In contrast, the actual measured value at a stress amplitude of 200 MPa shown in Figure 10(B) is 8.5x10 7 The measured values ​​at a stress amplitude of 220 MPa were 4.4 and 6.2x10 7 The AI ​​predicted value of Example 5 predicts the load stress corresponding to the fatigue life with a high degree of agreement with the actual measured value.

[0050] In low alloy steel, the fatigue life estimation method from the load stress in Comparative Example 1 was 10 6 Although the error became large for the fatigue life after the 1st test, it was found that the load stress prediction method corresponding to the fatigue life shown in Examples 1 to 5 can predict the SN diagram of low alloy steel, stainless steel, aluminum alloy, and titanium alloy with high accuracy regardless of the load type or stress ratio. In particular, for aluminum alloy, which has little variation, the observed results and the predicted results almost coincided. The data analysis took into account all factors such as elongation and reduction in area.

[0051] Material strength factors used as learning elements in the decision tree model and conditions during fatigue data acquisition experiments (1) The conditions for fatigue data acquisition experiments include the loading type and stress ratio. In Example 6, material strength factors are examined. EXAMPLES

[0052] FIG. 11 is a diagram showing one embodiment of the present invention, comparing the load stress prediction corresponding to the fatigue life of S45C steel with the actual measured values, and shows a case in which the load type and stress ratio are taken into account in the decision tree model machine learning section, but the tensile strength, hardness, elongation, and reduction in area are not taken into account. FIG. 12 is a diagram showing one embodiment of the present invention, comparing the load stress prediction corresponding to the fatigue life of S45C steel with the actual measured values, and shows a case where the elongation, reduction in area, loading type, and stress ratio are taken into consideration in the decision tree model machine learning section. FIG. 13 is a diagram showing one embodiment of the present invention, comparing the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, and shows a case where the tensile strength, loading type, and stress ratio are taken into account in the decision tree model machine learning section. FIG. 14 is a diagram showing one embodiment of the present invention, comparing the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, and shows a case where the tensile strength, hardness, loading type, and stress ratio are taken into consideration in the decision tree model machine learning section. FIG. 15 is a diagram showing one embodiment of the present invention, comparing the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, and shows a case where the decision tree model machine learning section takes into account tensile strength, hardness, elongation, reduction in area, loading type, and stress ratio.

[0053] Material strength factors include tensile strength, hardness, elongation, and reduction in area. If none of these factors are taken into consideration, the accuracy is 20.3%, if only tensile strength is taken into consideration, it is 3.0%, if tensile strength and hardness are taken into consideration, it is 2.5%, and if all of them are taken into consideration, it is 2.3%. It is desirable to take all of these factors into consideration, but it was found that sufficient accuracy can be obtained by taking tensile strength and hardness into consideration.

[0054] Material strength factors used as learning elements in the decision tree model and conditions during fatigue data acquisition experiments (2) In Example 7, the conditions for the fatigue data acquisition experiment are examined. EXAMPLES

[0055] FIG. 16 is a diagram showing one embodiment of the present invention, comparing the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, showing a case where the decision tree model machine learning section takes into account tensile strength, hardness, elongation, and reduction in area, but does not take into account the loading type and stress ratio. FIG. 17 is a diagram showing one embodiment of the present invention, comparing the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, showing a case in which the decision tree model machine learning section takes into account tensile strength, hardness, elongation, reduction in area, and stress ratio, but does not take into account the loading type. FIG. 18 is a diagram comparing the load stress predictions and actual measured values ​​corresponding to the fatigue life of S45C steel, showing one embodiment of the present invention, and shows a case in which the decision tree model machine learning section takes into account tensile strength, hardness, elongation, reduction in area, and loading type (rotating bending, axial load, torsion), but does not take into account the stress ratio.

[0056] In the conventional method of estimating fatigue life from applied stress, the effects of stress ratio and loading type were not well understood. However, in the method of estimating applied stress from fatigue life, the accuracy was 5.6% when the stress ratio and loading type were not taken into account, 4.9% when only the stress ratio was taken into account, 3.5% when only the loading type was taken into account, and 2.3% when both were taken into account. In other words, it was found that the accuracy of estimating applied stress from fatigue life is significantly improved by taking into account both the stress ratio and loading type in the method of estimating applied stress from fatigue life. [Industrial Applicability]

[0057] According to the present invention, the load stress estimation device and method for fatigue life can use the method of estimating the load stress from the fatigue life, and by estimating the SN curve (stress amplitude-fracture life relationship) by machine learning, the correspondence between the fatigue life and a specific load stress of the experimental value can be obtained with high accuracy. Therefore, it is possible to evaluate the long-term reliability of various machines and structures using the corresponding metal material or various composite materials, and it can also be used for the design of safety factors, and can be used for the design and reliability evaluation of various machines and structures. [Explanation of symbols]

[0058] 300 Computer Program Products 302 Program-carrying media 304 Program instruction storage section 306 Computer-readable medium 308 Recordable media 309 Communication media 310 Specified reading function unit 320 Fatigue data sheet reading function 325 Preprocessing section for supervised machine learning 330 Machine Learning Calculation Unit 335 Machine Learning Evaluation Department 340 Assist function part

Claims

1. A functional unit that reads the specification of the type of fatigue limit estimation and the mechanical characteristics used in the decision tree of machine learning, A functional unit that reads the fatigue data of the specified metal structural material from the fatigue data sheet, A machine learning operation unit that performs supervised machine learning of the load stress corresponding to the fatigue life on the fatigue data of the read metal structural material using the specified mechanical characteristics, A load stress estimation device corresponding to the fatigue life, configured to output an operation result of estimating the load stress corresponding to the fatigue life according to the specified type of fatigue limit estimation using the specified mechanical characteristics by the machine learning operation unit that has performed the supervised machine learning.

2. The load stress estimation device corresponding to the fatigue life according to claim 1, further comprising a preprocessing unit that performs machine learning of the load stress corresponding to the fatigue life on the fatigue data of the read metal structural material using the specified mechanical characteristics, and performs the supervised machine learning on the machine learning operation unit.

3. The load stress estimation device corresponding to the fatigue life according to claim 2, wherein the model used for the machine learning includes any one of a random forest method, deep learning, a neural network, or a model using LASSO regression.

4. As the type of the fatigue limit estimation, 10 ７ times fatigue limit estimation, 10 ６ times fatigue limit estimation, S-N curve estimation, and at least one of low cycle fatigue life estimation, and a load stress estimation device corresponding to the fatigue life according to any one of claims 1 to 3.

5. The load stress estimation device corresponding to the fatigue life according to claim 1, wherein the mechanical characteristics used in the decision tree of the machine learning include at least one of Vickers hardness, tensile strength, elongation at break, reduction of area, stress ratio, or fatigue test method.

6. The load stress estimation device corresponding to the fatigue life according to claim 5, wherein the fatigue test method includes at least one of rotating bending test data, axial load test data, or torsion test data.

7. The load stress estimation device corresponding to the fatigue life according to claim 1, wherein the metal structural material includes at least one of carbon steel, nickel-chromium-molybdenum steel, manganese steel, stainless steel, aluminum alloy, or titanium alloy.

8. The load stress estimation device corresponding to the fatigue life according to claim 7, wherein for the carbon steel, there are S25C, S35C, S45C, S55C; for the nickel-chromium-molybdenum steel, there is SNCM439; for the manganese steel, there are SMn438, SMn443; and for the stainless steel, there are at least one of SUS403 and SUS304.

9. Furthermore, a machine learning evaluation unit is provided that evaluates the accuracy of the load stress estimation corresponding to the fatigue life of the machine learning calculation unit that has performed supervised machine learning using the calculation result of the load stress estimation corresponding to the fatigue life according to the type of fatigue limit estimation used for calibration. The load stress estimation device corresponding to the fatigue life according to claim 1.

10. Furthermore, from the evaluation result of the machine learning evaluation unit, there is provided an assist function unit that recommends the type of mechanical characteristics used in the decision tree of the machine learning and the type of fatigue data of the metal structure material read from the fatigue data sheet so as to improve the accuracy of the load stress estimation corresponding to the fatigue life of the machine learning calculation unit. The load stress estimation device corresponding to the fatigue life according to claim 9.

11. A step of designating the type of fatigue limit estimation and designating the mechanical characteristics used in the decision tree of the machine learning; A step of reading the fatigue data of the designated metal structure material from the fatigue data sheet; Using the designated mechanical characteristics, for the fatigue data of the read metal structure material, for the machine learning calculation unit that has performed supervised machine learning of the load stress corresponding to the fatigue life, using the designated mechanical characteristics, a step of outputting a calculation result of the load stress corresponding to the fatigue life according to the designated type of fatigue limit estimation. A method for estimating the load stress corresponding to the fatigue life.

12. Furthermore, for the machine learning calculation unit, a step of performing machine learning of the load stress corresponding to the fatigue life on the fatigue data of the read metal structure material using the designated mechanical characteristics and performing supervised machine learning. The method for estimating the load stress corresponding to the fatigue life according to claim 11.

13. The model used for the machine learning includes any one of a random forest method, deep learning, a neural network, or a model using LASSO regression. The method for estimating the load stress corresponding to the fatigue life according to claim 12.

14. As the types of the fatigue limit estimation, any one of the estimation of the 10 ７ times fatigue limit, the estimation of the 10 ６ times fatigue limit, the estimation of the S-N curve, or the estimation of the low cycle fatigue life is included. The method for estimating the load stress corresponding to the fatigue life according to any one of claims 11 to 13.

15. The mechanical characteristics used in the decision tree of the machine learning include at least one of Vickers hardness, tensile strength, elongation at break, reduction of area, stress ratio, or fatigue test method. The method for estimating the load stress corresponding to the fatigue life according to claim 11.

16. The fatigue test method includes a load stress estimation method corresponding to the fatigue life according to claim 15, including at least one of rotational bending test data, axial load test data, or torsional test data.

17. The load stress estimation method corresponding to the fatigue life according to claim 11, wherein the metal structural material includes at least one of carbon steel, nickel chromium molybdenum steel, manganese steel, stainless steel, aluminum alloy, or titanium alloy.

18. The load stress estimation method corresponding to the fatigue life according to claim 17, wherein for the carbon steel, there are S25C, S35C, S45C, S55C; for the nickel chromium molybdenum steel, there is SNCM439; for the manganese steel, there are SMn438, SMn443; and for the stainless steel, there is at least one of SUS403, SUS304.

19. Furthermore, a machine learning evaluation step is provided to evaluate the accuracy of the load stress estimation corresponding to the fatigue life of the machine learning calculation unit that performs supervised machine learning using the calculation result of the load stress estimation corresponding to the fatigue life according to the type of fatigue limit estimation used for the calibration target, according to the load stress estimation method corresponding to the fatigue life according to claim 11.

20. Furthermore, a step is provided to recommend the type of mechanical characteristics used in the decision tree of machine learning and the type of fatigue data of the metal structural material read from the fatigue data sheet to improve the accuracy of the load stress estimation corresponding to the fatigue life of the machine learning calculation unit, based on the evaluation result of the machine learning evaluation step, according to the load stress estimation method corresponding to the fatigue life according to claim 19.

21. Furthermore, considering the type of mechanical characteristics used in the recommended decision tree of machine learning and the metal structural material read from the fatigue data sheet, a step is provided to specify the type of fatigue limit estimation used in the step of specifying the type of fatigue limit estimation and the mechanical characteristics used in the decision tree of machine learning, and a step is provided to correct the mechanical characteristics used in the decision tree of machine learning, according to the load stress estimation method corresponding to the fatigue life according to claim 20.

Images