Creep life assessment method

Abstract

The present invention relates to a creep life assessment method having high precision and accuracy, especially for using in an inspection and assessment of the service life of material at high temperature including easy to apply with the conventional creep assessment. Moreover, said method intends to assess the actual creep life considering the creep life prediction from the model built from machine learning from the factor and operating condition data that affects the creep such as temperature, stress, including Larson-Miller parameter and Larson-Miller constant, which have been adjusted to consider the factor and operating condition, resulting in more precise and accurate creep life prediction. Said method also considers the creep life fraction assessment by considering the physical property of the material together with the accumulated creep condition simulation, so that the creep life fraction assessment of the material from model built from machine learning is more precise. Therefore, the actual creep life can be assessed.

Description

[0001] CREEP LIFE ASSESSMENT METHOD

[0002] Technical Field

[0003] The present invention relates to the field of mechanical engineering or material engineering, in particular, to the creep life assessment.

[0004] Background of the Invention

[0005] The creep assessment is the assessment of material from damage mechanics from operation and deterioration at high temperature. Said creep assessment can be assessed in both creep stage or creep life. The popular method for creep assessment can be divided into 3 parts as the followings.

[0006] 1) The creep assessment by calculation from factors relating to the operation (operating condition) such as pressure, heat, etc. The creep assessment by calculation from factors relating to the operating condition has several methods such as the calculation with Larson-Miller parameter (LMP) according to The American Petroleum Institute (API) 530 standard which is one of the popular methods for creep assessment. This can identify the creep life due to factors directly used in the operation such as stress, temperature, etc. Also, this method takes into account the deterioration of the material by considering the thickness lost from use, which can be used in the calculation of the service life of the material under the environment causing the creep.

[0007] 2) The creep assessment from the material which is the creep assessment by considering the property of the material such as physical property, chemical property, etc. The examples of the creep assessment by considering the material property are as the followings.

[0008] For the creep assessment from microstructure of the material by replica, this creep assessment method needed specialist in the microstructure analysis. Moreover, Thai patent document no. 1801002825 discloses the material life assessment damaged from creep including the creep stage from the characteristic of carbide formation by on-site extraction replica (OSER) process which can identify the creep stage that occurs.

[0009] For the creep assessment from potential drop measurement such as the creep assessment from measuring the alternating current potential drop (ACPD), the creep assessment from measuring the direct current potential drop (DCPD), etc., this is the resistance measurement of the material which is used in the measurement of the fatigue crack growth. However, normally the direct current potential drop measurement is often used for the crack assessment.

[0010] In the present, there is a development and research relating to the creep assessment method from measuring the potential drop such as the thesis entitled “Creep monitoring using permanently installed potential drop sensors (2015, Imperial College London) by Joseph Corcoran that discloses the development of the measurement method of potential drop by installation of sensor on the test specimen at high temperature with an attempt to use the potential drop to identify the damage from creep by compensating the resistance changed when the material is exposed to high temperature.

[0011] 3) The creep assessment from the creep testing for use in the life assessment by the accelerated creep test and test according to creep life fraction range using accumulated creep condition simulation for damage assessment from creep in each creep stage.

[0012] Apart from the popular creep assessment methods described above, at present there is a development of the creep life assessment method employing the machine learning in the creep life prediction. The machine learning is the process of utilizing the data to automatically identify the pattern of the data to be used to build the model or template for use in analysis of data sciences. This can be divided from the objective of the learning into 3 types which are: 1) supervised learning; 2) unsupervised learning, and 3) reinforcement learning. At present, there are disclosures of the development of creep life assessment methods employing the machine learning such as research by Mengyu Chai et., al. (Applied Sciences, 2023, 13(8), 4972) that discloses the machine learning framework for predicting the creep rupture life of the modified 9chromium-l molybdenum steel (modified 9Cr-lMo steel), research by Xu Zhang et., al. (Materials, 2023, 16(20), 6804) that discloses the method for predicting the creep rupture life of small-sample. The research by Osman Mamun et., al. (npj Materials Degradation, 2021, 20) that discloses the method for predicting the creep rupture life of ferritic and austenitic steel. The research by Yue Liu et., al. (Acta Materialia, 2020, 195, 454-467) that discloses the method for predicting the creep rupture life of Ni-based single crystal superalloys using divide-and-conquer approach based machine learning.

[0013] However, the creep life assessment methods employing the machine learning according to the research documents mentioned above still have limitations such as precision in the actual creep life assessment and applicability to the conventional creep assessment techniques.

[0014] Therefore, the present invention aims to develop the creep life assessment method that provides high precision and accuracy and can be easily applied to conventional creep assessment. The creep life assessment method according to the present invention focuses on the actual creep life assessment by considering the creep life prediction from the model built from machine learning from factor and operating condition data that affects the creep such as temperature, stress, etc., including Larson-Miller parameter and Larson-Miller constant, which have been adjusted to consider said factor and operating condition, resulting in the Larson-Miller parameter and Larson- Miller constant employed in the machine learning and model building being more precise. This results in more precise and accurate creep life prediction. Moreover, the creep life assessment method according to the present invention also considers the creep life fraction assessment by considering the physical property of the material together with the accumulated creep condition simulation, so that the creep life fraction assessment of the material from the model built from machine learning is more precise. Therefore, the creep life assessment method according to the present invention can assess the actual creep life. of the Invention

[0015] The present invention relates to a creep life assessment method. The objective is to develop the creep life assessment method having high precision and accuracy, especially for using in an inspection and assessment of the service life of material at high temperature. Said method can also be easily applied to the conventional creep assessment.

[0016] The other objective of the present invention is to provide the creep life assessment method that focuses on the actual creep life assessment by considering the creep life prediction from the model built from machine learning from the factor and operating condition data that affects the creep such as temperature, stress, etc., including Larson-Miller parameter and Larson-Miller constant, which have been adjusted to consider in said factor and operating condition. This makes more precise Larson-Miller parameter and Larson-Miller constant employed in the machine learning and model building, resulting in more precise and accurate creep life prediction. Moreover, the creep life assessment method according to the present invention also considers the creep life fraction assessment by considering the physical property of the material together with the accumulated creep condition simulation, so that the creep life fraction assessment of the material from model built from machine learning is more precise. Therefore, the creep life assessment method according to the present invention can assess the actual creep life.

[0017] In order to accomplish the objective above, the present invention provides the creep life assessment method, wherein said creep life assessment method comprising: (a) prediction of the creep life from the model built from machine learning from the factor and operating condition data comprising:

[0018] (a.l) step of inputting data for learning, wherein said step of inputting data for learning comprising:

[0019] (a.1.1) receiving data comprising stress, temperature, and design life (IDL), and optionally further comprising building the stress function model, wherein said stress function model is developed from received stress and temperature data;

[0020] (a.1.2) receiving the Larson-Miller constant (C) data for the starting of the learning; wherein step (a.1.1) and (a.1.2) are interchangeable;

[0021] (a.1.3) receiving the Larson-Miller parameter (LMP) data for the starting of the learning comprising: building of the Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter; wherein said Larson-Miller parameter used in the development of Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant obtained from (a.1.2);

[0022] (a.2) step of machine learning and building of model for predicting the creep life, wherein the starting data for machine learning and building of model for predicting the creep life is selected from:

[0023] - stress data;

[0024] - Larson-Miller parameter data; and

[0025] - Larson-Miller constant data; (a.3) step of predicting the creep life, wherein the step of predicting the creep life is operated to predict the creep life using the model obtained from step (a.2) and the conditions related to the operation of the material desired to assess the creep life;

[0026] (b) assessing the creep life fraction using the physical property from the model built from machine learning, comprising:

[0027] (b.l) step of inputting data for learning, wherein said data comprises the potential drop and creep life fraction;

[0028] (b.2) step of machine learning and building of model for assessing the creep life fraction from the potential drop, comprising the step of:

[0029] (b.2.1) cleaning the data, wherein said cleaning of the data is operated by removing creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage; and

[0030] (b.2.2) machine learning and building of the model for assessing the creep life fraction from the potential drop using data obtained from step (b.2.1);

[0031] (b.3) step of assessing the creep life fraction of the material desired to assess the creep life, comprising:

[0032] (b.3.1) assessing the creep stage together with measuring the potential drop on the material desired to assess the creep life; and

[0033] (b.3.2) assessing the creep life fraction from the creep life fraction assessment model from the potential drop obtained from step (b.2.2); and

[0034] (c) assessing the remaining creep life of the material desired to assess the creep life, wherein the remaining creep life assessment of said material is performed using the creep life obtained from step (a) and the creep life fraction obtained from step (b).

[0035] Figure 1 shows a schematic diagram of the creep life prediction from the model built from the machine learning from the factor and operating condition data.

[0036] Figure 2 shows a graph showing relationship between stress and temperature at each design life from the standard data of API530 and the stress function model (as a function with temperature) in the form of 5thdegree polynomial obtained from the building of the stress function model (as a function with temperature).

[0037] Figure 3 shows a graph showing relationship between Larson-Miller parameter and stress and the Larson-Miller parameter model that is a function with the stress in the form of 3rddegree polynomial obtained from the building of the Larson-Miller parameter model that is a function with the stress using the minimum Larson-Miller constant and the average Larson-Miller constant from the standard data of API530.

[0038] Figure 4 shows a schematic diagram of the creep life fraction assessment using the physical property from the model built from machine learning.

[0039] Figure 5 shows a measurement of direct current potential drop that comprising the current supplying probe (1) connected to the power source that supplies current via probe tip into material, current receiving probe (2) connected to the current circuit as the current receiver, and resistance measuring probe (3) that measures the resistance and converts into the potential drop, wherein the area between the two probe tips is the resistance measurement area of the material for use in the creep life assessment. Figure 6 shows an example of potential drop and creep life fraction data obtained from the accumulated creep condition simulation on the specimen that is exposed to the temperature of 580 °C and the force of 165 MPa.

[0040] Figure 7 shows a graph showing relationship between normalized direct current potential drop (NDCPD) and creep life fraction from the data that excludes the creep life fraction data in the range of 0-5%, which is data of creep life range in the initial stage or not in the creep stage, and the creep life fraction assessment model from potential drop obtained from linear regression analysis and polynomial regression analysis.

[0041] Detailed Description

[0042] Any aspect being demonstrated herein also means to include the application to other aspects of this invention unless stated otherwise.

[0043] Definitions

[0044] Technical terms or scientific terms used herein have definitions as understood by an ordinary person skilled in the art unless stated otherwise.

[0045] Any tools, devices, methods, materials, equations, or calculation formulas named herein mean tools, devices, methods, materials, equations, or calculation formulas being operated or used commonly by those person skilled in the art unless stated otherwise that they are tools, devices, methods, materials, equations, or calculation formulas specific only in this invention.

[0046] Use of singular noun or singular pronoun with “consisting of’, “comprising”, “further comprising”, and “including” in claims or specification means “one” and also including “one or more”, “at least one”, and “one or more than one” .

[0047] The terms “consist(s) of’, “comprise(s)”, “contain(s)”, and “include(s)” are open-end verbs. For example, any method which “consists of’, “comprises”, “contains”, or “includes” one component or multiple components or one step or multiple steps is not limited to only one component or one step or multiple components or multiple steps as specified but shall encompass components or steps that are not specified.

[0048] All compositions and / or methods disclosed and claims in this application are intended to cover embodiments from any operation, performance, modification, or adjustment any factors without any experiment that significantly different from this invention, and obtain with object with utility and resulted as same as the present embodiment according to person ordinary skilled in the art although without specifically stated in claims. Therefore, substitutable or similar object to the present embodiment, including any minor modification or adjustment that can be apparent to person skilled in the art should be construed as remains in spirit, scope, and concept of invention as appeared in appended claims.

[0049] Throughout this application, term “about” means any number that appeared or expressed herein that could be varied or deviated from any error of equipment, method, or personal using said equipment or method, including variations or deviations occurred from changes in reaction conditions of uncontrollable factors such as humidity and temperature.

[0050] The creep life in this invention means the material life that damages from creep or the creep rupture life.

[0051] The creep in this invention means the deformation of the material occurred by load at high temperature, or may mean the rupture or creep that happens slowly and continuously. This occurs in metallic material or equipment subjected to the force while operating at high temperature, selected from but limited to boiler, boiler parts, pressure vessel, steam engine, turbine, propeller, fired pressure vessel. Hereafter, invention embodiments are shown without any purpose to limit any scope of the invention.

[0052] The present invention relates to the creep life assessment method that provides high precision and accuracy, especially for using in an inspection and assessment of the service life of the material at high temperature, and can be easily applied to the conventional creep assessment, wherein said creep life assessment method comprises:

[0053] (a) prediction of the creep life from the model built from machine learning from the factor and operating condition data comprising:

[0054] (a.l) step of inputting data for learning, wherein said step of inputting data for learning comprises:

[0055] (a.1.1) receiving data comprising stress, temperature, and design life (IDL), and optionally further comprising building the stress function model, wherein said stress function model is developed from received stress and temperature data;

[0056] (a.1.2) receiving the Larson-Miller constant (C) data for the starting of the learning; wherein step (a.1.1) and (a.1.2) are interchangeable;

[0057] (a.1.3) receiving the Larson-Miller parameter (LMP) data for the starting of the learning comprising: building of the Larson-Miller parameter model that is a functions with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant obtained from (a.1.2); (a.2) step of machine learning and building of model for predicting the creep life, wherein the starting data for machine learning and building of model for predicting the creep life is selected from:

[0058] - stress data;

[0059] - Larson-Miller parameter data; and

[0060] - Larson-Miller constant data;

[0061] (a.3) step of predicting the creep life, wherein step of predicting the creep life is operated to predict the creep life using the model obtained from step (a.2) and the conditions related to the operation of the material desired to assess the creep life;

[0062] (b) assessing the creep life fraction using the physical property from the model built from machine learning, comprising:

[0063] (b.l) step of inputting data for learning, wherein said data comprises the potential drop and creep life fraction;

[0064] (b.2) step of machine learning and building of model for assessing the creep life fraction from the potential drop, comprising the step of:

[0065] (b.2.1) cleaning the data, wherein said cleaning of the data is operated by removing creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage; and

[0066] (b.2.2) machine learning and building of the model for assessing the creep life fraction from the potential drop using data obtained from step (b.2.1);

[0067] (b.3) step of assessing the creep life fraction of the material desired to assess the creep life, comprising: (b.3.1) assessing the creep stage together with measuring the potential drop on the material desired to assess the creep life; and

[0068] (b.3.2) assessing the creep life fraction from the creep life fraction assessment model from the potential drop obtained from step (b.2.2);

[0069] (c) assessing the remaining creep life of the material desired to assess the creep life, wherein the remaining creep life assessment of said material is performed using the creep life obtained from step (a) and the creep life fraction obtained from step (b).

[0070] In one aspect of the invention, step (a) and (b) are interchangeable.

[0071] In one aspect of the invention, data in step (a.1.1 ) is selected from material standard data, experimental data, or combination thereof. Preferably, the data in step (a.1.1) is the material standard data selected from the standard data of The American Petroleum Institute (API) 530, the standard data of The American Petroleum Institute (API) 579, the standard data of The Welding Research Council (WRC) 541, and the standard data of R5 Assessment Procedure for the High Temperature Response of Structures.

[0072] In one aspect of the invention, step (a.1.1) is receiving data comprising stress, temperature, and design life, and further building the stress function model, wherein said stress function model is developed from received stress and temperature data. Preferably, said stress model is developed by defining in the form of 5thdegree polynomial.

[0073] In one aspect of the invention, the Larson-Miller constant data started in the step (a.1.2) is selected from the Larson-Miller constant from the material standard data and the Larson-Miller constant obtained from the relationship between stress with temperature and design life in step

[0074] (a.1.1). In one aspect of the invention, said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial.

[0075] For the creep life assessment method according to the present invention, step (a.l) may further comprise the machine learning and precision testing of the data, wherein said machine learning and precision testing of the data is operated using model selected from random forest, neural network, regression, clustering, and classification. Preferably, said machine learning and precision testing of the data is operated using random forest model, wherein the precision of the data is considered from the consistency of the distribution of the Larson-Miller constant.

[0076] In one aspect of the invention, the starting data for machine learning and building of model for predicting the creep life in step (a.2) is stress data, wherein said stress data is the data obtained from step (a.1.1). In said aspect, the Larson-Miller parameter model and the Larson-Miller constant obtained in step (a.2) may refer from the Larson-Miller parameter model and Larson-Miller constant obtained from step (a.l). Or, in said aspect, the Larson-Miller parameter model obtained in step (a.2) is operated by machine learning and building the Larson-Miller parameter model that is a function with the stress, and the Larson-Miller constant obtained in step (a.2) refers to the Larson-Miller constant obtained from step (a.l).

[0077] In one aspect of the invention, the starting data for machine learning and building of model for predicting the creep life in step (a.2) is selected from:

[0078] - the Larson-Miller parameter data, wherein said Larson-Miller parameter data is the Larson-Miller parameter data obtained from step (a.1.3); or

[0079] - the Larson-Miller constant data, wherein said Larson-Miller constant data is the Larson- Miller constant data obtained from step (a.1.2).

[0080] In one aspect of the invention, step (a.2) comprises the step of: (a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, preferably said Larson-Miller constant model is developed by the multiple regression analysis; and

[0081] (a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress, preferably said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial; wherein step (a.2.1) and (a.2.2) are interchangeable.

[0082] In one aspect of the invention, the starting data for machine learning and building of model for predicting creep life in step (a.2) is the Larson-Miller parameter data, wherein said Larson- Miller parameter data is the Larson-Miller parameter data obtained from step (a.1.3). In said aspect, preferably step (a.2) comprises the step of:

[0083] (a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1), temperature received in step (a.1.1), and Larson-Miller constant. Preferably, said Larson-Miller constant model is developed by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller parameter obtained from the starting data; and

[0084] (a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter. Preferably, said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant calculated from the Larson-Miller constant model obtained in step (a.2.1).

[0085] In one aspect of the invention, the starting data for machine learning and building of model for predicting creep life in step (a.2) is the Larson-Miller constant data, wherein said Larson-Miller constant data is the Larson-Miller constant data obtained from step (a.1.2). In said aspect, preferably step (a.2) comprises the step of:

[0086] (a.2.1) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter. Preferably, said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant obtained from the starting data; and

[0087] (a.2.2) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1), temperature received in step (a.1.1), and Larson-Miller constant. Preferably, said Larson-Miller constant model is developed by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller parameter calculated from the Larson-Miller parameter model obtained in step (a.2.1). For the creep life assessment method according to the present invention, step (a) may further comprise inputting additional data by data augmentation.

[0088] In one aspect of the invention, the starting data for machine learning and building of model for predicting the creep life in step (a.2) is the stress data, wherein said stress data is the data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation. In said aspect, the Larson-Miller parameter model and Larson-Miller constant obtained in step (a.2) may refer from the Larson-Miller parameter model and Larson-Miller constant obtained from step (a.l). Or, in said aspect, the Larson-Miller parameter model obtained in step (a.2) is operated by machine learning and building the Larson-Miller parameter model that is a function with stress and the Larson-Miller constant obtained in step (a.2) refers to the Larson- Miller constant obtained from step (a.l).

[0089] In one aspect of the invention, the starting data for machine learning and building of model for predicting the creep life in step (a.2) is selected from:

[0090] - the Larson-Miller parameter data, wherein said Larson-Miller parameter data is the Larson-Miller parameter data obtained from step (a.1.3) and further comprised the Larson-Miller parameter data calculated from Larson-Miller parameter model obtained in step (a.1.3) using data from inputting additional data by data augmentation; or

[0091] - the Larson-Miller constant data, wherein said Larson-Miller constant data is the Larson- Miller constant data obtained from step (a.1.2).

[0092] In one aspect of the invention, step (a.2) comprises the step of:

[0093] (a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature. Preferably, said Larson-Miller constant model is developed by multiple regression analysis; and (a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress. Preferably, said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial; wherein step (a.2.1) and (a.2.2) are interchangeable.

[0094] In one aspect of the invention, the starting data for machine learning and building of model for predicting the creep life in step (a.2) is the Larson-Miller parameter data, wherein said Larson- Miller parameter data is the Larson-Miller parameter data obtained from step (a.1.3) and further comprised the Larson-Miller parameter data calculated from the Larson-Miller parameter model obtained in step (a.1.3) by using data from inputting additional data by data augmentation. In said aspect, preferably step (a.2) comprises the step of:

[0095] (a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, temperature received in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant. Preferably, said Larson-Miller constant model is developed by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter obtained from the starting data;

[0096] (a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter. Preferably, said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant calculated from the Larson-Miller constant model obtained in step (a.2.1).

[0097] In one aspect of the invention, the starting data for machine learning and building of model for predicting the creep life in step (a.2) is the Larson-Miller constant data, wherein said Larson- Miller constant data is the Larson-Miller constant data obtained from step (a.1.2). In said aspect, preferably step (a.2) comprises the step of:

[0098] (a.2.1) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter. Preferably, said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant obtained from the starting data;

[0099] (a.2.2) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, temperature received in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant. Preferably, said Larson-Miller constant model is developed by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter calculated from the Larson-Miller parameter model obtained in step (a.2.1).

[0100] For the creep life assessment method according to the present invention, step (a.2) may further comprise:

[0101] - the testing algorithm, preferably the testing algorithm is operated by considering the model accuracy from the coefficient of determination (R2) and considering the precision in prediction by the method selected from mean absolute percentage error (MAPE), root mean square error (RMSE), or median performance error (MDPE). Most preferably, the testing algorithm is operated by considering the model accuracy from the coefficient of determination and considering the precision in prediction from mean absolute percentage error; and

[0102] - the iterative machine learning by using the results obtained from building of model for predicting the creep life as the input data for machine learning, preferably the iterative machine learning stops its operation when the obtained mean absolute percentage error changes less than 10 % of the mean absolute percentage error in the previous testing of the prediction precision.

[0103] In one aspect of the invention, step (a.3) is operated to predict the creep life using the model obtained from step (a.2) and the conditions related to the operation of the material desired to assess the creep life, comprising stress and temperature of the material according to the operation. In one aspect of the invention, said potential drop is the direct current potential drop (DCPD).

[0104] In one aspect of the invention, step (b.l) comprises the step of:

[0105] (b.1.1) receiving the data from accumulated creep condition simulation; and

[0106] (b.1.2) normalizing the data, wherein said normalizing comprises the step of compensation of losing area of the specimen in the accumulated creep condition simulation and the compensation of potential drop of the specimen that simulates the accumulated creep condition.

[0107] In one aspect of the invention, step (b.1.1) further comprises the receiving data from creep stage assessment from characteristic of carbide formation by on-site extraction replica (OSER) process.

[0108] In one aspect of the invention, step (b.2.1) the creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage is obtained from the creep stage assessment from characteristic of carbide formation by on-site extraction replica (OSER) process.

[0109] In one aspect of the invention, step (b.2.2) machine learning and building of the model for assessing the creep life fraction from the potential drop is operated by the analysis method selected from linear regression, polynomial regression, or operated by said both analysis methods.

[0110] In one aspect of the invention, step (b.3.1) assessing the creep stage is operated by the technique selected from microstructure assessment by replica process, analysis of carbide formation characteristic by on-site extraction replica (OSER) process, omega method, or combination of two or more said techniques. Preferably, step (b.3.1) assessing the creep stage is operated by microstructure assessment by replica process together with analysis of carbide formation characteristic by on-site extraction replica process.

[0111] In one aspect of the invention, step (b.3.2) comprises the step of: (b.3.2.1) selecting of creep life fraction assessment model from the potential drop by considering the potential drop measured on the material desired to assess the creep life; and

[0112] (b.3.2.2) calculating the creep life fraction from the model obtained in step (b.3.2.1) using the potential drop measured on the material desired to assess the creep life.

[0113] In one aspect of the invention, the material desired to assess the creep life is selected from 9chromium-l molybdenum- vandium steel (9Cr-lMo-V steel; P91 / T91), 2.25chromium- Imolybdenum steel (2.25Cr-lMo steel; P22 / T22), and 1.25chromium-0.5molybdenum steel (1.25Cr-0.5Mo steel; Pl l / Tl l).

[0114] The creep life assessment method according to the present invention is used for assessing the creep life of the material selected from 9chromium-lmolybdenum-vandium steel (9Cr-lMo-V steel; P91 / T91), 2.25chromium-l molybdenum steel (2.25Cr-lMo steel; P22 / T22), and 1.25chromium-0.5molybdenum steel (1.25Cr-0.5Mo steel; Pl l / Tl l).

[0115] The use of the creep life assessment method according to the present invention assesses the creep life of the equipment selected from boiler, boiler parts, pressure vessel, steam engine, turbine, propeller, fired pressure vessel.

[0116] The following examples are only for demonstrating one aspect of this invention, not for limiting the scope of this invention in any way.

[0117] The creep life assessment method of 9chromium-l molybdenum- vandium steel or 9Cr- IMo-V steel (P91 / T91)

[0118] The creep life assessment method according to the invention of 9chromium-lmolybdenum- vandium steel or 9Cr-lMo-V steel (P91 / T91) can be operated by the following process.

[0119] (a) The prediction of creep life from the model built from machine learning from the factor and operating condition data For the prediction of creep life from the model built from machine learning from the factor and operating condition data, said obtained model is used for creep life prediction. Said creep life prediction considers the factor and operating condition data such as temperature, stress, etc., including Larson-Miller parameter and Larson-Miller constant, which have been adjusted to consider said factor and operating condition. The schematic diagram for creep life prediction from the model built from machine learning from factor and operating condition data is shown in figure

[0120] 1. Said creep life prediction from the model built from machine learning from factor and operating condition data can be operated by the following steps.

[0121] (a.l) The step of inputting data for learning, comprising the following steps.

[0122] (a.1.1) Receiving data comprising stress, temperature, and design life (tor), wherein said data is received from API530 standard data which is the creep life design data of material according to API530 standard, providing design life data from 20,000 - 100,000 hours. Then the stress model (as a function with temperature) is built as a stress function model, wherein said stress model is developed from the stress and temperature data obtained from the above, by plotting relationship graph between stress and temperature to fit curve and acquire the mathematical functional equation of said curve in the form of 5thdegree polynomial according to WRC 541 stress equation (equation 1) as the following: (equation 1); wherein said obtained mathematical functional equation in the form of 5thdegree polynomial is the stress function model that is a function with the temperature as shown in figure

[0123] 2.

[0124] (a.1.2) Receiving the Larson-Miller constant (C) data for the starting of the learning, wherein said Larson-Miller constant for the starting of the learning is the Larson-Miller constant from API530 standard data comprising minimum Larson-Miller constant (Cmin) and average Larson-Miller constant (Cavg).

[0125] (a.1.3) Receiving the Larson-Miller parameter (LMP) data for the starting of learning, comprising the building of Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter by plotting relationship graph between Larson-Miller parameter and stress to fit curve and acquire the mathematical functional equation of said curve in the form of 3rddegree polynomial; wherein the stress obtained from step (a.1.1) used in building of said Larson-Miller parameter model is the stress calculated from stress function model obtained from step (a.1.1 ) ; and wherein the Larson-Miller parameter used in building of said Larson-Miller parameter model is the Larson-Miller parameter calculated from Larson-Miller equation (equation 2) using temperature and design life data obtained in step (a.1.1) above, and the Larson-Miller constant obtained from step (a.1.2):

[0126] LMP = T[C + log(t)] (equation 2);

[0127] Said obtained mathematical functional equation in the form of 3rddegree polynomial is the

[0128] Larson-Miller parameter model that is a function with the stress as shown in figure 3.

[0129] (a.1.4) Machine learning and precision testing of data, wherein the machine learning is operated using random forest model to test the accuracy of the data in previous step and test the precision of the model by training the model using training data set and testing the model using testing data set (training / testing) and the precision is considered from the consistency of the distribution of the Larson-Miller constant. (a.2) The step of machine learning and building the creep life prediction model, wherein said step of machine learning and building the creep life prediction model is the supervised learning by finding target from training and testing of the learning and building said creep life prediction model. Said machine learning and building the creep life prediction model can be operated by selecting different starting data as the followings.

[0130] Type 1 the machine learning and building the creep life prediction model using stress data as the starting data, wherein said stress data is the data obtained from step (a.1.1). Therefore, the obtained Larson-Miller parameter model and Larson-Miller constant can refer from the Larson- Miller parameter model and Larson-Miller constant obtained from step (a.l).

[0131] Type 2 the machine learning and building the creep life prediction model using the Larson- Miller parameter data as the starting data, wherein said Larson-Miller parameter data is the Larson- Miller parameter data obtained from step (a.1.3). Therefore, step (a.2) comprises the step of:

[0132] (a.2.1) machine learning and building of Larson-Miller constant model that is a function with the stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1), temperature obtained in step (a.1.1), and Larson-Miller constant by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation (equation 3) using temperature and design life data obtained in step (a.1.1) and Larson-Miller parameter obtained from the starting data; (equation 3); wherein the mathematical functional equation obtained from said multiple regression analysis is the Larson-Miller constant model that is a function with the stress and temperature; (a.2.2) machine learning and building of Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter by defining said model in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation (equation 2) using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant calculated from the Larson- Miller constant model obtained in step (a.2.1).

[0133] Type 3 the machine learning and building the creep life prediction model using the Larson- Miller constant data as the starting data, wherein said Larson-Miller constant data is the Larson- Miller constant data obtained from step (a.1.2). Therefore, step (a.2) comprises the step of:

[0134] (a.2.1) machine learning and building of Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter by defining said model in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation (equation 2) using temperature and design life data obtained in step (a.1.1 ) and Larson-Miller constant obtained from the starting data;

[0135] (a.2.2) machine learning and building of Larson-Miller constant model that is a function with the stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1), temperature obtained in step (a.1.1), and Larson-Miller constant by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation (equation 3) using temperature and design life data obtained in step (a.1.1) and Larson-Miller parameter calculated from the Larson- Miller parameter model obtained in step (a.2.1).

[0136] Moreover, it further comprises inputting additional data by data augmentation, wherein the additional data inputted is the creep testing data from outside and inside the laboratory such as the test data of NIMS (National Institute for Material Science, JAPAN) and testing data of WELLab (WELLab, King Mongkut’s University of Technology Thonburi) to be operated according to the step of machine learning and building the creep life prediction model. The starting data after inputting additional data by data augmentation can be selected as the followings.

[0137] Type 4 the machine learning and building the creep life prediction model using stress data as the starting data, wherein said stress data is the data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation. Therefore, the obtained Larson-Miller parameter model and Larson-Miller constant can refer from the Larson-Miller parameter model and Larson-Miller constant obtained from step (a.l).

[0138] Type 5 the machine learning and building the creep life prediction model using Larson- Miller parameter data as the starting data, wherein said Larson-Miller parameter data is the Larson- Miller parameter data obtained from step (a.1.3) and further comprised the Larson-Miller parameter data calculated from the Larson-Miller parameter model obtained in step (a.1.3) using data from inputting additional data by data augmentation. Therefore, step (a.2) comprises the step of:

[0139] (a.2.1) machine learning and building of Larson-Miller constant model that is a function with the stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, temperature obtained in step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from the Larson-Miller constant equation (equation 3) using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation and Larson-Miller parameter obtained from the starting data;

[0140] (a.2.2) machine learning and building of Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter by defining said model in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation (equation 2) using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant calculated from Larson-Miller constant model obtained in step (a.2.1).

[0141] Type 6 the machine learning and building the creep life prediction model using Larson- Miller constant data as the starting data, wherein said Larson-Miller constant data is the Larson- Miller constant data obtained from step (a.1.2). Therefore, step (a.2) comprises the step of:

[0142] (a.2.1) machine learning and building of Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter by defining said model in the form of 3rddegree polynomial; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation (equation 2) using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant obtained from the starting data;

[0143] (a.2.2) machine learning and building of Larson-Miller constant model that is a function with the stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, temperature obtained in step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant by multiple regression analysis; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from the Larson-Miller constant equation (equation 3) using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter calculated from the Larson-Miller parameter model obtained in step (a.2.1).

[0144] For step (a.2), the step of machine learning and building the creep life prediction model using all 6 types of the starting data further comprises:

[0145] - testing algorithm by operating the accuracy test of the model from the coefficient of determination (R2) which must be more than 0.9, and considering the precision in prediction from mean absolute percentage error (MAPE) as shown in equation 4; (equation 4);

[0146] - iterative machine learning by using the result obtained from building of creep life prediction model above as the input data for machine learning by building of the above model again for improving the accuracy and precision in the building of creep life prediction model, wherein the iterative machine learning stops its operation when the obtained mean absolute percentage error changes less than 10 % of the mean absolute percentage error in the previous testing of the prediction precision.

[0147] (a.3) the step of creep life prediction by predicting the creep life using model obtained from machine learning, comprising:

[0148] - selecting the creep life prediction model obtained from step (a.2), wherein said model provides highest precision in prediction;

[0149] - predicting the creep life using selected model by calculating the creep life by creep life prediction equation (equation 5), wherein the Larson-Miller parameter and Larson-Miller constant are calculated using selected model and condition related to the operation of the material desired to assess the creep life; comprising stress and temperature of material due to operation;

[0150] (equation 5).

[0151] (b) The creep life fraction assessment using the physical property from the model built from machine learning

[0152] The creep life fraction assessment using the physical property from the model built from machine learning is performed by using said model for the creep life fraction assessment, wherein said creep life fraction assessment considers the physical property of the material in the creep life assessment together with the accumulated creep condition simulation. The schematic diagram of the creep life fraction assessment using the physical property from the model built from machine learning as shown in figure 4. Said creep life fraction assessment using the physical property from the model built from machine learning can be operated by the following steps.

[0153] (b.1 ) The step of inputting data for learning, wherein said step of inputting data for learning comprises: (b.1.1) Receiving the data from accumulated creep condition simulation, wherein said data comprises the potential drop and creep life fraction, wherein said data is obtained from the accumulated creep condition simulation according to creep life fraction range and the measurement of direct current potential drop (DCPD) in each creep life fraction range to collect the potential drop and creep life fraction data. The specimen used in the accumulated creep condition simulation comprises the specimen exposed to temperature and force on the specimen and the comparative specimen exposed to temperature but no force on the specimen. The measurement of direct current potential drop can be operated by the measurement method of four- point direct current potential drop using probe to supply-receive the current and measure the resistance. The obtained potential drop is the result from the total resistance measurement of probe and material. The example of the measurement method of four-point direct current potential drop is shown in figure 5 and the example of potential drop and creep life fraction data obtained from the accumulated creep condition simulation are shown in figure 6.

[0154] Moreover, it further comprises the receiving data from creep stage assessment from characteristic of carbide formation by on-site extraction replica (OSER) process for assessing the creep stage occurred, wherein the specimen subjected to the accumulated creep condition simulation is used to analyze of characteristic of carbide formation by on-site extraction replica process.

[0155] (b.1.2) Normalizing the data by data normalization to obtain the potential drop that has been compensated or adjusted to the normalized form, wherein the data normalization comprises the step of:

[0156] - compensating the losing area of the specimen in the accumulated creep condition simulation by adjusting the area of the specimen that simulates the accumulated creep condition (As) compared to the area of the specimen before the accumulated creep condition simulation (Ao) as shown in equation 6;

[0157] (equation 6);

[0158] - compensating the potential drop of the specimen that simulates the accumulated creep condition by adjusting the potential drop of the specimen that simulates the accumulated creep condition (DCPDs) compared to the potential drop of the comparative specimen exposed to temperature but no force on the specimen (DCPDTB) as shown in equation 7; (tequation 7); wherein the potential drop of the specimen that simulates the accumulated creep condition (DCPDs) is the potential drop measured on the specimen that simulates the accumulated creep condition (DCPD), subjected to the area compensation by equation 8;

[0159] DCPDs = Narea x DCPD (equation 8).

[0160] (b.2) The step of machine learning and building of model for assessing the creep life fraction from the potential drop, comprising:

[0161] (b.2.1) Cleaning the data by removing the creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage for obtaining the cleaned data, wherein said creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage are obtained from the creep stage assessment from characteristic of carbide formation by onsite extraction replica process in step (b.1.1).

[0162] (b.2.2) Machine learning and building of the model for assessing the creep life fraction from the potential drop by plotting the relationship graph between potential drop obtained from data normalization (normalized DCPD; NDCPD) and creep life fraction from the cleaned data above, and then fitting curve and finding the mathematical functional equation suitable to the data set by linear regression and polynomial regression as shown in figure 7.

[0163] The mathematical functional equation obtained from said linear regression and polynomial regression is the creep life fraction assessment model from the potential drop.

[0164] Moreover, it further comprises the precision testing of the creep life fraction assessment model from the potential drop, wherein said precision testing is operated by training the model using training data set and testing the model using testing data set (training / testing) and the precision is considered from the statistical reliability.

[0165] (b.3) The step of assessing the creep life fraction of the material desired to assess the creep life, wherein said step can be operated by the method comprising:

[0166] (b.3.1) assessing the creep stage together with measuring the potential drop on the material desired to assess the creep life;

[0167] - assessing the creep stage on the material desired to assess the creep life by microstructure assessment technique by replica process together with analysis of carbide formation characteristic by on-site extraction replica process, which is performed by preparing specimen at the resolution of 1 pm, etching, replicating, and analyzing by scanning electron microscope with energy dispersive X-ray spectroscopy (SEM-EDX) in order to analyze the proportion of compound or element and determine the element proportion that forms carbide such as Cr to Fe ratio, and identify the creep stage range from carbide formation characteristic of the material. The creep stage range can be divided into 2 ranges which are initial stage and creep stage;

[0168] - measuring the potential drop on the material desired to assess the creep life, which is performed by the measurement of direct current potential drop on the surface of the specimen material desired to assess using high resolution nano voltmeter with setting the machine to supply the current at 1 ampere at the potential drop of 5 volts.

[0169] (b.3.2) assessing the creep life fraction from the creep life fraction assessment model from the potential drop, wherein said creep life fraction assessment is operated by using potential drop data measured from step (b.3.1) and the creep life fraction assessment model from potential drop obtained from step (b.2) by selecting the creep life fraction assessment model that provides precision in the assessment in range of said measured potential drop .

[0170] (c) The remaining creep life assessment of the material desired to assess the creep life

[0171] The remaining creep life assessment of the material desired to assess the creep life can be performed using the creep life obtained from step (a) and the creep life fraction obtained from step (b) and calculating the remaining creep life of the material desired as follows:

[0172] Remaining creep life of the material = (100 % - creep life fraction obtained from step (b) x creep life obtained from step (a).

[0173] From the creep life prediction method according to the invention in above example, it is found that the precision testing of the data for learning in building the model that is built from machine learning from the factor and operating condition data provides the distribution of Larson- Miller constant with consistency, indicating that the learned data is precise and accurate. From the testing algorithm, it is found that the creep life prediction model built shows high precision in prediction. Moreover, from the testing algorithm of creep life fraction assessment model from the potential drop, it is found that it is statistically reliable. Therefore, the creep life prediction method according to this invention can be used to assess the creep life with higher precision and accuracy.

[0174] The implementation of the method according to the present invention previously described which comprises various steps may be performed in any other order different from the one described. Any modifications and changes evident to a person of ordinary skilled in the art should be construed to be within the spirit, scope, and concept of the present invention

[0175] Best Mode of the Invention

[0176] Best mode or preferred embodiment of the invention is as provided in the description of the invention.

Claims

Claims1. A creep life assessment method, said method comprising:(a) prediction of the creep life from the model built from machine learning from the factor and operating condition data comprising:(a.l) step of inputting data for learning, wherein said step of inputting data for learning comprises:(a.1.1) receiving data comprising stress, temperature, and design life (IDL), and optionally further comprising building the stress function model, wherein said stress function model is developed from received stress and temperature data;(a.1.2) receiving the Larson-Miller constant (C) data for the starting of the learning; wherein step (a.1.1) and (a.1.2) are interchangeable;(a.1.3) receiving the Larson-Miller parameter (LMP) data for the starting of the learning comprising: building of the Larson-Miller parameter model that is a function with the stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1 ) and Larson-Miller parameter; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant obtained from (a.1.2);(a.2) step of machine learning and building of model for predicting the creep life, wherein the starting data for machine learning and building of model for predicting the creep life is selected from:- stress data;- Larson-Miller parameter data; and- Larson-Miller constant data;(a.3) step of predicting the creep life, wherein step of predicting the creep life is operated to predict the creep life using the model obtained from step (a.2) and the conditions related to the operation of the material desired to assess the creep life;(b) assessing the creep life fraction using the physical property from the model built from machine learning, comprising:(b.l) step of inputting data for learning, wherein said data comprises the potential drop and creep life fraction;(b.2) step of machine learning and building of model for assessing the creep life fraction from the potential drop, comprising the step of:(b.2.1) cleaning the data, wherein said cleaning of the data is operated by removing creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage; and(b.2.2) machine learning and building of the model for assessing the creep life fraction from the potential drop using data obtained from step (b.2.1);(b.3) step of assessing the creep life fraction of the material desired to assess the creep life, comprising:(b.3.1) assessing the creep stage together with measuring the potential drop on the material desired to assess the creep life; and(b.3.2) assessing the creep life fraction from the creep life fraction assessment model from the potential drop obtained from step (b.2.2);(c) assessing the remaining creep life of the material desired to assess the creep life, wherein the remaining creep life assessment of said material is performed using the creep life obtained from step (a) and the creep life fraction obtained from step (b).

2. The creep life assessment method according to claim 1, wherein step (a) and (b) are interchangeable.

3. The creep life assessment method according to claim 1, wherein the data in step (a.1.1) is selected from material standard data, experimental data, or combination thereof.

4. The creep life assessment method according to claim 3, wherein the data in step (a.1.1) is the material standard data selected from the standard data of The American Petroleum Institute (API) 530, the standard data of The American Petroleum Institute (API) 579, the standard data of The Welding Research Council (WRC) 541, and the standard data of R5 Assessment Procedure for the High Temperature Response of Structures.

5. The creep life assessment method according to claim 1 or 4, wherein step (a.1.1) is receiving data comprising stress, temperature, and design life, and further building the stress function model, wherein said stress function model is developed from received stress and temperature data.

6. The creep life assessment method according to claim 1 or 5, wherein said stress model is developed by defining in the form of 5thdegree polynomial.

7. The creep life assessment method according to claim 1, wherein the Larson-Miller constant data for the starting in step (a.1.2) is selected from the Larson-Miller constant from the material standard data and the Larson-Miller constant obtained from the relationship between stress with temperature and design life in step (a.1.1).

8. The creep life assessment method according to claim 1, wherein said Larson-Miller parameter model is developed by defining in the form of 3rddegree polynomial.

9. The creep life assessment method according to claim 1, wherein step (a.l) further comprises the machine learning and precision testing of the data.

10. The creep life assessment method according to claim 9, wherein the machine learning and precision testing of the data is operated using model selected from random forest, neural network, regression, clustering, and classification.

11. The creep life assessment method according to claim 10, wherein the machine learning and precision testing of the data is operated using random forest model.

12. The creep life assessment method according to claim 11, wherein the precision of the data is considered from the consistency of the distribution of the Larson-Miller constant.

13. The creep life assessment method according to claim 1, wherein the starting data for machine learning and building of model for predicting the creep life in step (a.2) is stress data, wherein said stress data is the data obtained from step (a.1.1).

14. The creep life assessment method according to claim 13, wherein the Larson-Miller parameter model and the Larson-Miller constant obtained in step (a.2) refer from the Larson-Miller parameter model and Larson-Miller constant obtained from step (a.l).

15. The creep life assessment method according to claim 13, wherein the Larson-Miller parameter model obtained in step (a.2) is operated by machine learning and building the Larson- Miller parameter model that is a function with the stress, and the Larson-Miller constant obtained in step (a.2) refers the Larson-Miller constant obtained from step (a.l).

16. The creep life assessment method according to claim 1, wherein the starting data for machine learning and building of model for predicting the creep life in step (a.2) is the Larson-Miller parameter data, wherein said Larson-Miller parameter data is the Larson-Miller parameter data obtained from step (a.1.3).

17. The creep life assessment method according to claim 1, wherein the starting data for machine learning and building of model for predicting the creep life in step (a.2) is the Larson- Miller constant data, wherein said Larson-Miller constant data is the Larson-Miller constant data obtained from step (a.1.2).

18. The creep life assessment method according to claim 16 or 17, wherein step (a.2) comprises the step of:(a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature; and(a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress; wherein step (a.2.1) and (a.2.2) are interchangeable.

19. The creep life assessment method according to claim 16, wherein step (a.2) comprises the step of:(a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1), temperature received in step (a.1.1), and Larson-Miller constant; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller parameter obtained from the starting data;(a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant calculated from the Larson-Miller constant model obtained in step (a.2.1).

20. The creep life assessment method according to claim 17, wherein step (a.2) comprises the step of:(a.2.1) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and Larson-Miller parameter; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller constant obtained from the starting data;(a.2.2) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1), temperature received in step (a.1.1), and Larson-Miller constant; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and Larson-Miller parameter calculated from the Larson-Miller parameter model obtained in step (a.2.1).

21. The creep life assessment method according to claim 1, wherein step (a) further comprises inputting additional data by data augmentation.

22. The creep life assessment method according to claim 21, wherein the starting data for machine learning and building of model for predicting the creep life in step (a.2) is stress data, wherein said stress data is the data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation.

23. The creep life assessment method according to claim 22, wherein Larson-Miller parameter model and Larson-Miller constant obtained in step (a.2) refer from the Larson-Miller parameter model and Larson-Miller constant obtained from step (a.l).

24. The creep life assessment method according to claim 22, wherein the Larson-Miller parameter model obtained in step (a.2) is operated by machine learning and building the Larson- Miller parameter model that is a function with the stress and the Larson-Miller constant obtained in step (a.2) refers the Larson-Miller constant obtained from step (a.l).

25. The creep life assessment method according to claim 21, wherein the starting data for machine learning and building of model for predicting the creep life in step (a.2) is Larson-Miller parameter data, wherein said Larson-Miller parameter data is the Larson-Miller parameter data obtained from step (a.1.3) and further comprised the Larson-Miller parameter data calculated from Larson-Miller parameter model obtained in step (a.1.3) using data from inputting additional data by data augmentation.

26. The creep life assessment method according to claim 21, wherein the starting data for machine learning and building of model for predicting the creep life in step (a.2) is Larson-Miller constant data, wherein said Larson-Miller constant data is the Larson-Miller constant data obtained from step (a.1.2).

27. The creep life assessment method according to claim 25 or 26, wherein step (a.2) comprises the step of:(a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature; and(a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress; wherein step (a.2.1) and (a.2.2) are interchangeable.

28. The creep life assessment method according to claim 25, wherein step (a.2) comprises the step of:(a.2.1) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, temperature received in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter obtained from the starting data;(a.2.2) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter;wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant calculated from the Larson-Miller constant model obtained in step (a.2.1).

29. The creep life assessment method according to claim 26, wherein step (a.2) comprises the step of:(a.2.1) machine learning and building of Larson-Miller parameter model that is a function with stress, wherein said Larson-Miller parameter model is developed from stress data obtained from step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller parameter; wherein the Larson-Miller parameter used in the development of said Larson-Miller parameter model is calculated from Larson-Miller equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant obtained from the starting data;(a.2.2) machine learning and building of Larson-Miller constant model that is a function with stress and temperature, wherein said Larson-Miller constant model is developed from stress data obtained from step (a.1.1 ) and further comprised data from inputting additional data by data augmentation, temperature received in step (a.1.1) and further comprised data from inputting additional data by data augmentation, and Larson-Miller constant; wherein the Larson-Miller constant used in the development of said Larson-Miller constant model is calculated from Larson-Miller constant equation using temperature and design life data obtained in step (a.1.1) and further comprised data from inputting additional data by dataaugmentation, and Larson-Miller parameter calculated from the Larson-Miller parameter model obtained in step (a.2.1).

30. The creep life assessment method according to any one of claim 18, 19, 20, 27, 28, or 29, wherein said Larson-Miller constant model is developed by the relationship analysis using multiple regression.

31. The creep life assessment method according to claim 1, wherein step (a.2) further comprises testing algorithm and iterative machine learning by using the results obtained from building of model for predicting the creep life as the input data for machine learning.

32. The creep life assessment method according to claim 31 , wherein the testing algorithm is operated by considering the model accuracy from the coefficient of determination (R2) and considering the precision in prediction by the method selected from mean absolute percentage error (MAPE), root mean square error (RMSE), or median performance error (MDPE).

33. The creep life assessment method according to claim 32, wherein the testing algorithm is operated by considering the model accuracy from the coefficient of determination and considering the precision in prediction from mean absolute percentage error.

34. The creep life assessment method according to any one of claim 31 to 33, wherein the iterative machine learning stops its operation when the obtained mean absolute percentage error changes less than 10 % of the mean absolute percentage error in the previous testing of the prediction precision.

35. The creep life assessment method according to claim 1, wherein step (a.3) is operated to predict the creep life using the model obtained from step (a.2) and the conditions related to the operation of the material desired to assess the creep life comprising stress and temperature of the material according to the operation.

36. The creep life assessment method according to claim 1, wherein said potential drop is the direct current potential drop (DCPD).

37. The creep life assessment method according to claim 1, wherein step (b.l) comprises the step of:(b.1.1) receiving the data from accumulated creep condition simulation; and(b.1.2) normalizing the data, wherein said normalizing comprises the step of compensation of losing area of the specimen in the accumulated creep condition simulation and the compensation of potential drop of the specimen that simulates the accumulated creep condition.

38. The creep life assessment method according to claim 37, wherein step (b.1.1) further comprises the receiving data from creep stage assessment from characteristic of carbide formation by on-site extraction replica (OSER) process.

39. The creep life assessment method according to claim 1, wherein step (b.2.1) the creep life fraction data in the initial stage and the creep life fraction data that is not in the creep stage is obtained from creep stage assessment from characteristic of carbide formation by on-site extraction replica process.

40. The creep life assessment method according to claim 1, wherein step (b.2.2) machine learning and building of the model for assessing the creep life fraction from the potential drop is operated by the analysis method selected from linear regression, polynomial regression, or operated by said both analysis methods.

41. The creep life assessment method according to claim 1, wherein step (b.3.1) assessing the creep stage is operated by the technique selected from microstructure assessment by replica process, analysis of carbide formation characteristic by on-site extraction replica (OSER) process, omega method, or combination of two or more said techniques.

42. The creep life assessment method according to claim 1 or 41, wherein step (b.3.1) assessing the creep stage is operated by microstructure assessment by replica process together with analysis of carbide formation characteristic by on-site extraction replica process.

43. The creep life assessment method according to claim 1, wherein step (b.3.2) comprises the step of:(b.3.2.1) selecting of creep life fraction assessment model from the potential drop by considering the potential drop measured on the material desired to assess the creep life; and(b.3.2.2) calculating the creep life fraction from the model obtained in step (b.3.2.1) using the potential drop measured on the material desired to assess the creep life.

44. The creep life assessment method according to any one of the preceding claims, wherein the material desired to assess the creep life is selected from 9chromium-lmolybdenum- vandium steel (9Cr-lMo-V steel; P91 / T91), 2.25chromium-lmolybdenum steel (2.25Cr-lMo steel; P22 / T22), and 1.25chromium-0.5molybdenum steel (1.25Cr-0.5Mo steel; Pl l / Tl l).

45. A use of the creep life assessment method according to any one of the preceding claims for the creep life assessment of the equipment selected from boiler, boiler parts, pressure vessel, steam engine, turbine, propeller, fired pressure vessel.

Images