Method for predicting service conditions of directionally solidified nickel-based superalloy blade

Abstract

The invention discloses a method for predicting the service condition of a directionally solidified nickel-based superalloy blade, and relates to the field of predicting the creep service condition of the blade, and the method comprises the following steps: S1, carrying out a creep interruption experiment on a material for many times; s2, after each creep interruption experiment is finished, characteristic parameters of a gamma'strengthening phase are extracted, and corresponding creep temperature and creep stress are sorted; s3, taking the characteristic parameters as input and the creep temperature as output, performing training by using a recurrent neural network, and constructing a temperature prediction model; s4, taking the characteristic parameters and the creep temperature as input, taking the creep stress as output, performing training by utilizing a recurrent neural network, and constructing a stress prediction model; and S5, characteristic parameters of the blade made of the material after service are obtained, and a temperature prediction model and a stress prediction model are used for prediction. According to the method, through linkage of a plurality of machine learning prediction models, the creep temperature and the creep stress of the directionally solidified nickel-based superalloy blade are rapidly predicted.

Description

Technical Field

[0001] The invention relates to the field of predicting blade creep service conditions, in particular to a method for predicting the service conditions of a directionally solidified nickel-based high-temperature alloy blade. Background Art

[0002] Creep deformation and failure are typical forms of damage to gas turbine blades. Due to the high temperature, high stress service environment of the blades and the flexible start-stop process, it is impossible to accurately determine the service temperature field and stress field through real-time monitoring, and then determine the service life. Accurately predicting the service temperature field and stress field of oriented nickel-based high-temperature alloys is of great guiding significance for blade design, manufacturing, operation and maintenance services, especially for the prediction of remaining service life. Through prediction, improvement measures can be taken in advance for the subsequent high stress areas and weak points of the service blades, thereby improving the safety and reliability of product design, manufacturing, and operation and maintenance.

[0003] In recent years, with the development of machine learning technology, it has shown great potential in the field of materials science. By analyzing a large amount of experimental data and simulation data, machine learning algorithms can identify complex nonlinear relationships and provide more accurate and efficient prediction results. Chinese patent CN113420502A discloses "A method for predicting the life of nickel-based single crystal high-temperature alloys by linking two models using machine learning methods". This method introduces the prediction of service temperature mainly using statistical mathematical models, but it is necessary to establish a large number of databases with different service temperatures, and it is impossible to accurately predict the service temperature with a small amount of data, and it is impossible to predict the service stress. Chinese patent CN110411850 discloses "A method for evaluating the service conditions of high-temperature alloy turbine blades". By establishing a database of microstructure evolution of DZ125 alloy under near-service conditions, combined with the BP network machine learning method, the equivalent maximum service temperature of turbine blades and its corresponding service stress and damage time under the temperature and stress conditions are evaluated. The prediction characteristic parameters, prediction process and prediction model of this patent and this method are significantly different from this method, and the prediction accuracy is significantly lower than this method; Chinese patent CN115035963A discloses "A method for predicting the creep life of high-temperature alloys using BPNN neural network and Larson-Miller parameter equation", which is also not a prediction of blade service conditions. CN202311836085 discloses a method for judging the creep stage of aluminum-containing heat-resistant alloy furnace tube materials at a service temperature of 1100°C, which is also not a prediction of blade service conditions.

[0004] In summary, the current existing technology cannot make accurate and rapid predictions of the creep temperature and creep stress of directionally solidified nickel-based high-temperature alloy blades. Summary of the invention

[0005] The purpose of the present invention is to provide a method for predicting the service conditions of directionally solidified nickel-based high-temperature alloy blades in response to the above-mentioned problems. The method makes a rapid prediction of the creep temperature and creep stress of directionally solidified nickel-based high-temperature alloy blades through the linkage of multiple machine learning prediction models.

[0006] The technical solution adopted by the present invention is as follows: A method for predicting the service conditions of a directionally solidified nickel-based high-temperature alloy blade comprises the following steps: S1: creep interruption experiments were conducted on directionally solidified nickel-based superalloys for many times, and the conditions of each creep interruption experiment were different; S2: After each creep interruption experiment, metallographic sample preparation is performed, the microstructure morphology is photographed, the characteristic parameters of the γ' strengthening phase are extracted, and the corresponding creep temperature and creep stress are sorted out; S3: Taking the characteristic parameters as input and the creep temperature as output, a temperature prediction model is constructed by training using a recurrent neural network; S4: Taking characteristic parameters and creep temperature as input and creep stress as output, a stress prediction model is constructed by training using a recurrent neural network; S5: Obtain characteristic parameters of the directionally solidified nickel-based high-temperature alloy blade after service, and use the temperature prediction model and the stress prediction model to predict the predicted temperature and predicted stress of the directionally solidified nickel-based high-temperature alloy blade during creep service; the creep temperature input into the stress prediction model is the predicted temperature output by the temperature prediction model.

[0007] Furthermore, in step S1, the creep conditions of the creep interruption experiment are the experimental temperature and the experimental stress.

[0008] Furthermore, during the creep interruption experiment, the experimental temperature range is 800-1000°C, and the experimental stress range is 125-600MPa.

[0009] Furthermore, among the multiple creep interruption experiments, the creep interruption experiments with a creep time range of 20-918h were selected as effective experiments.

[0010] Furthermore, the characteristic parameters include rafting degree Ω, channel width , area fraction ;in:

[0011] in: is the spacing between the connected γ' phases on the image bisector; n is the number of spacings;

[0012] in: is the area of ​​the γ' enhanced phase in the image; is the total area of ​​the image.

[0013] Furthermore, in step S2, when photographing the microscopic tissue morphology, multiple microscopic tissue morphologies at different positions should be photographed.

[0014] Furthermore, in step S3, the structure, learning rate and number of training times of the recurrent neural network are adjusted during training so that the error between the predicted temperature output by the temperature prediction model and the creep temperature in the creep interruption experiment is within an allowable range.

[0015] Furthermore, in step S4, the structure, learning rate and number of training times of the recurrent neural network are adjusted during training so that the error between the predicted stress output by the stress prediction model and the creep stress in the creep interruption experiment is within an allowable range.

[0016] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: The present invention establishes a temperature prediction model and a stress prediction model through machine learning, and under the linkage of the temperature prediction model and the stress prediction model, it can make a rapid prediction of the creep temperature and creep stress of the directionally solidified nickel-based high-temperature alloy blade, thereby providing reliable data for the study of the service process of the directionally solidified nickel-based high-temperature alloy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The present invention will now be described by way of example with reference to the accompanying drawings, in which: Figure 1 The flow chart of the method for predicting creep service conditions of directionally solidified nickel-based superalloys is shown; Figure 2 It is a schematic diagram of the temperature prediction model structure; Figure 3 It is a schematic diagram of the stress prediction model structure; Figure 4 The creep temperature prediction results outside the data set obtained according to the creep condition prediction method of the oriented nickel-based superalloy are: Figure 5 The creep stress prediction results outside the data set obtained according to this directional nickel-based superalloy creep condition prediction method. DETAILED DESCRIPTION

[0018] In the description of this specification, it should be noted that if the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc. appear, the orientation or position relationship indicated is based on the orientation or position relationship shown in the drawings, or is the orientation or position relationship in which the product of this specification is usually placed when used. It is only for the convenience of describing this specification and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this specification.

[0019] In addition, if the terms "horizontal" or "vertical" appear in the description of this specification, it does not mean that the components are required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0020] In the description of this specification, it should also be noted that, unless otherwise clearly stipulated and limited, the terms "set", "installed", "connected" and "connected" should be understood in a broad sense. For example, the connection can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be a connection between the internal parts of two elements.

[0021] like Figure 1-Figure 5 As shown, a method for predicting the service conditions of a directionally solidified nickel-based high-temperature alloy blade is studied for a directionally solidified nickel-based high-temperature alloy blade with a grade of Mar-M247LC, comprising the following steps: S1: Using experimental temperature and experimental stress as creep conditions, creep interruption experiments were carried out on directionally solidified nickel-based high-temperature alloys for many times. In each creep interruption experiment, the experimental temperature was selected within 800-1000°C, and the experimental stress was selected within 125-600MPa; the creep conditions of each creep interruption experiment were different.

[0022] It should be noted that, among the multiple creep interruption experiments mentioned above, creep interruption experiments with a creep time range of 20-918h were selected as effective experiments to ensure the effectiveness of the creep interruption experiments and avoid data errors caused by damage to the directionally solidified nickel-based high-temperature alloy blade material due to other reasons.

[0023] S2: After each creep interruption experiment, for effective experiments, metallographic sample preparation is performed, microstructure morphology is photographed, and characteristic parameters of the γ' strengthening phase are extracted. The characteristic parameters are rafting degree Ω, channel width , area fraction , and sort out the corresponding creep temperature and creep stress; among which:

[0024] in: is the spacing between the connected γ' phases on the bisector; n is the number of spacings;

[0025] in: is the area of ​​the γ' enhanced phase in the image; is the total area of ​​the image.

[0026] It should be noted that when performing step S2, since the metallographic structure may be segregated and present an uneven state, multiple microstructure morphologies at different positions should be taken when photographing the microstructure morphology to fully obtain the characteristic parameters of the γ' strengthening phase, thereby improving the accuracy of the subsequent prediction model establishment.

[0027] At the same time as step S2, after obtaining the characteristic parameters and the corresponding creep temperature and creep stress, these data are randomly divided to form a training set and a test set.

[0028] S3: Figure 2 As shown, a component temperature prediction model is constructed using data from a training set; specifically, a temperature prediction model is constructed by training a recurrent neural network using characteristic parameters as input and creep temperature as output; It should be noted that if Figure 4 As shown, the temperature prediction model is tested with the test set during training, and the parameters such as the structure of the recurrent neural network, the learning rate and the number of training times are continuously adjusted so that the error between the predicted temperature output by the temperature prediction model and the creep temperature (experimental temperature) in the test set with the characteristic parameters in the test set as input is within the allowable range, and the allowable error is within ±15°C, thereby ensuring the accuracy of the temperature prediction model.

[0029] S4: Figure 3 As shown, a component temperature prediction model is constructed using data from a training set; specifically, a stress prediction model is constructed by training a recurrent neural network using characteristic parameters and creep temperature as input and creep stress as output; It should be noted that if Figure 5 As shown, the stress prediction model is tested with the test set during training, and the parameters such as the structure of the recurrent neural network, the learning rate and the number of training times are continuously adjusted so that the characteristic parameters in the test set and the predicted temperature output by the temperature prediction model are used as input, and the error between the predicted stress output by the stress prediction model and the creep stress (experimental stress) in the test set is within the allowable range, and the allowable error is within ±10MPa, thereby ensuring the accuracy of the temperature prediction model.

[0030] S5: Obtain characteristic parameters of the directionally solidified nickel-based high-temperature alloy blade after service, and use the temperature prediction model and the stress prediction model to predict the predicted temperature and predicted stress of the directionally solidified nickel-based high-temperature alloy blade during creep service; the creep temperature input into the stress prediction model is the predicted temperature output by the temperature prediction model.

[0031] The present invention is not limited to the above-mentioned specific embodiments, but extends to any new features or any new combination disclosed in this specification, as well as any new method or process steps or any new combination disclosed.

Claims

1. A method for predicting the service conditions of a directionally solidified nickel-based superalloy blade, characterized in that: The following steps are involved: S1: creep interruption experiments were conducted on directionally solidified nickel-based superalloys multiple times, with different creep conditions in each experiment; S2: After each creep interruption experiment, metallographic sample preparation is performed, the microstructure morphology is photographed, the characteristic parameters of the γ' strengthening phase are extracted, and the corresponding creep temperature and creep stress are sorted out; S3: Taking the characteristic parameters as input and the creep temperature as output, a temperature prediction model is constructed by training using a recurrent neural network; S4: Taking characteristic parameters and creep temperature as input and creep stress as output, a stress prediction model is constructed by training using a recurrent neural network; S5: Obtain characteristic parameters of the directionally solidified nickel-based high-temperature alloy blade after service, and use the temperature prediction model and the stress prediction model to predict the predicted temperature and predicted stress of the directionally solidified nickel-based high-temperature alloy blade during creep service; the creep temperature input into the stress prediction model is the predicted temperature output by the temperature prediction model.

2. The method according to claim 1, characterized in that: In step S1, the creep conditions of the creep interruption experiment are the experimental temperature and the experimental stress.

3. The method according to claim 2, characterized in that: During the creep interruption experiment, the experimental temperature range is 800-1000°C and the experimental stress range is 125-600MPa.

4. The method according to claim 3, characterized in that: Among the multiple creep interruption experiments, the creep experiments with creep time ranging from 20 to 918 h were selected as effective experiments.

5. The method according to claim 1, characterized in that: The characteristic parameters include rafting degree Ω, channel width , area fraction ;in: in: is the spacing between the connected γ' phases on the bisector; n is the number of spacings; in: is the area of ​​the γ' enhanced phase in the image; is the total area of ​​the image.

6. The method according to claim 1, characterized in that: In step S2, when photographing the microscopic structure morphology, multiple microscopic structure morphologies at different positions should be photographed.

7. The method according to claim 1, characterized in that: In step S3, the structure, learning rate and number of training times of the recurrent neural network are adjusted during training so that the error between the predicted temperature output by the temperature prediction model and the creep temperature in the creep interruption experiment is within an allowable range.

8. The method according to claim 1, characterized in that: In step S4, the structure, learning rate and number of training times of the recurrent neural network are adjusted during training so that the error between the predicted stress output by the stress prediction model and the creep stress in the creep interruption experiment is within an allowable range.

Images