High-temperature alloy characterization parameter and microstructure mapping model construction method and application thereof
Through neural network model, the lossless characterization parameters are mapped with the microstructure characteristics of nickel-based high-temperature alloys, which solves the problem of destructive sample preparation for microstructure characterization of nickel-based high-temperature alloys in the prior art, and realizes an efficient and lossless characterization method.
Patent Information
- Application Number
- CN202311610451.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-28
- Publication Date
- 2025-05-30
AI Technical Summary
In the prior art, the microstructure characterization of nickel-based high-temperature alloys requires destructive sample preparation, with long periods, high costs and limited application scenarios, making it difficult to achieve non-destructive testing.
By establishing a neural network model, the non-destructive characterization parameters are used to map the microstructure characteristics of the nickel-based high-temperature alloy standard samples to construct a high-temperature alloy characterization parameters and microstructure mapping model to achieve the correspondence between the non-destructive detection results and microstructure.
The non-destructive characterization of nickel-based high-temperature alloys is realized, which improves characterization efficiency, reduces test costs, and avoids the limitation of destructive sample preparation.
Smart Images
Figure CN120072121A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of nondestructive testing, and particularly relates to a method for constructing a mapping model between characterization parameters and microstructures of superalloys and its application. Background Art
[0002] Nickel-based superalloys have good high-temperature mechanical properties, durability, and creep resistance, and are thus widely used in the manufacture of aeroengine components. Nickel-based superalloys are usually heat-treatable alloys, and need to undergo heat treatment in different processes to form specific microstructures in the nickel-based superalloy structure to obtain the comprehensive properties required by the part design requirements. Nickel-based superalloys will form various different precipitation phase particles such as γ', γ", δ, etc. in the γ matrix under different heat treatment systems. The morphology and distribution of these phases have a significant impact on the macroscopic properties of nickel-based superalloy parts. Therefore, characterizing the microstructure of nickel-based superalloys is an important basis for optimizing the structural design and heat treatment system of nickel-based superalloy parts. At present, the characterization of the microstructure of nickel-based superalloys usually requires cutting samples from parts and then characterizing them by means such as scanning electron microscopy (SEM) or transmission electron microscopy (TEM). This process has a long cycle, high test costs, a complex sample preparation process, and is destructive to the parts themselves, with limited application scenarios. Therefore, providing a method for constructing a mapping model between the characterization parameters of nondestructive testing of nickel-based superalloys and the microstructure of nickel-based superalloys has high practical value for optimizing the characterization method of the microstructure of nickel-based superalloy parts. Summary of the Invention
[0003] The purpose of the present invention is to provide a method for constructing a mapping model between characterization parameters and microstructures of superalloys, which can establish the correspondence between the nondestructive testing results of nickel-based superalloys and the microstructure. The present invention also provides a nondestructive characterization system and method for superalloys.
[0004] According to an embodiment of one aspect of the present invention, there is provided a method for constructing a mapping model between characterization parameters and microstructures of superalloys, the method comprising the following steps: providing the microstructure characteristics of a plurality of nickel-based superalloy standard samples and a plurality of nondestructive characterization parameters corresponding to the nickel-based superalloy standard samples as training samples; establishing a neural network model, training the neural network model using the training samples, predicting the microstructure characteristics according to the nondestructive characterization parameters and performing iteration, and when the prediction error is less than a given error threshold, obtaining a mapping model between the characterization parameters and microstructures of superalloys; when the prediction error cannot be less than the error threshold after iteration, replacing the nondestructive characterization parameters in the training samples.
[0005] Through this method, a mapping relationship between the non-destructive characterization test results of superalloys and the microstructural characteristics of superalloys can be established using a neural network, and a mapping model capable of predicting the microstructure based on the non-destructive test results can be obtained.
[0006] Further, in some embodiments, before training the neural network model, there is also a step of verifying the training samples: establishing the microstructural characteristics in the training samples as a microstructural matrix A, and establishing the non-destructive characterization parameters in the training samples as a non-destructive characterization matrix B. When r(A) > r(B), replace the non-destructive characterization parameters in the training samples, where r is the rank of the matrix.
[0007] Further, in some embodiments, when training the neural network model, normalize the microstructural characteristics and the non-destructive characterization parameters in the training samples.
[0008] Further, in some embodiments, the neural network model uses a BP prediction model or a Kriging model.
[0009] Further, in some embodiments, the neural network model uses a BP prediction model, including an input layer, an output layer, and a hidden layer. The input layer includes 2 nodes, the output layer includes 6 nodes, and the number of nodes in the hidden layer is q. where N is the number of neurons in the input layer, M is the number of neurons in the output layer, and a is a constant between 0 and 10.
[0010] Further, in some embodiments, the nodes in the input layer include two of conductivity, microhardness, and Hall coefficient.
[0011] Further, in some embodiments, the nodes in the output layer include the average size of γ' phase, the area ratio of γ' phase, the average size of γ'' phase, the area ratio of γ'' phase, the average size of δ phase, and the area ratio of δ phase.
[0012] Further, in some embodiments, the given error threshold ≤ 5%.
[0013] According to an embodiment of another aspect of the present invention, a non-destructive characterization system for superalloys is provided. The system includes a memory and a processor, wherein the memory stores a calculation program, and when the calculation program is executed by the processor, it can implement the method for constructing the mapping between the superalloy characterization parameters and the microstructure provided in any of the foregoing embodiments.
[0014] According to an embodiment of another aspect of the present invention, a method for non-destructive characterization of superalloys is provided, including the following steps: providing a mapping model of superalloy characterization parameters and microstructure, inputting the characterization parameters of the superalloy sample to be characterized into the mapping model of superalloy characterization parameters and microstructure, and obtaining the microstructure result predicted by the mapping model of superalloy characterization parameters and microstructure, wherein the mapping model of superalloy characterization parameters and microstructure is constructed by using the method for constructing the mapping model of superalloy characterization parameters and microstructure provided in any of the foregoing embodiments.
[0015] Further, in some embodiments, the method further includes the following steps: providing a database of superalloy heat treatment parameters - microstructure, and determining the heat treatment parameters of the superalloy to be characterized according to the microstructure result predicted by the mapping model of superalloy characterization parameters and microstructure. Description of the Drawings
[0016] Figure 1 It is a flowchart of the method for non-destructive characterization of superalloys in an embodiment;
[0017] Figure 2 It is a schematic diagram of a neural network structure in an embodiment.
[0018] The purpose of the above-mentioned drawings is to make a detailed description of the present invention so that those skilled in the art can understand the technical concept of the present invention, rather than aiming to limit the present invention. Detailed Embodiments
[0019] The present invention will be further described in detail below through specific embodiments in conjunction with the drawings.
[0020] The mention of "embodiment" in this article means that the specific features, structures or characteristics described in connection with the embodiment may be included in at least one embodiment of this article. The phrase appearing at various positions in the specification does not necessarily refer to the same embodiment, nor is it limited to mutually exclusive independent or alternative embodiments. Those skilled in the art should be able to understand that the embodiments in this article can be combined with other embodiments without structural conflicts. In the description of this article, the meaning of "a plurality" is at least two.
[0021] Nickel-based superalloys are widely used in the manufacture of aeroengine components. Through heat treatment, γ', γ'' strengthening phases or δ precipitate phases are formed in the γ matrix phase, thereby obtaining good high-temperature mechanical properties. The microstructure of the precipitate phase will continuously evolve under various heat treatment processes during the manufacturing process, leading to the evolution of mechanical properties. By characterizing the microstructure, on the one hand, it can guide the optimization of heat treatment processes during the preparation of materials and components; on the other hand, during the service process of aeroengine components, they experience complex thermal cycles, and the precipitate phases of nickel-based alloys will change accordingly. The composition of the precipitate phase will affect the mechanical properties of nickel-based superalloys. By regularly detecting the microstructure of key components, the degradation of component performance can be accurately judged, thereby avoiding potential failure risks.
[0022] Currently, the characterization of the microstructure of nickel-based superalloys usually adopts a destructive sampling method to cut samples from parts, and then uses technical means such as SEM or TEM for microstructure analysis. For the actual production process, the sampling and characterization processes of SEM or TEM are time-consuming and costly, and the destructive sampling method is unacceptable for some application scenarios. For example, when conducting long-term tests on nickel-based superalloy parts, it is necessary to perform multiple characterizations on the parts at different service durations to determine the evolution process of their microstructure under long-term service conditions. In such scenarios, the destructive sampling and characterization method is unacceptable, and if multiple specimens are provided in parallel for destructive sampling, the cost is extremely high.
[0023] To solve the above problems, an embodiment of one aspect of the present invention provides a method for constructing a mapping model between characterization parameters and microstructure of superalloys. By establishing a mapping relationship, the microstructure characteristics of materials can be predicted based on non-destructive characterization parameters (i.e., physical parameter data that can be obtained from non-destructive testing of material samples), thereby effectively improving the characterization efficiency of nickel-based superalloys and reducing the test cost.
[0024] In one embodiment, the construction method of the mapping model is as follows.
[0025] First, prepare multiple standard samples of superalloys. These standard samples have the same initial state after full annealing treatment and are heat-treated with different heat treatment parameters to have different microstructures. These standard samples are respectively characterized using non-destructive testing techniques to obtain a non-destructive characterization parameter data set; the microstructure characteristics of the standard samples are determined by SEM testing or querying against standard metallographic maps respectively to obtain a microstructure characteristic data set. The non-destructive characterization parameter data set - microstructure characteristic data set is used as a training sample to establish a training sample database.
[0026] Next, a neural network model is established based on the BP model, the Kriging model, or other approximation models. The neural network model is trained iteratively using the training sample database. The neural network model is used to predict the microstructure characteristics based on the non-destructive characterization parameters. When the prediction error is less than the given error threshold, the training is completed, and a mapping model between the superalloy characterization parameters and the microstructure is obtained. If, after a certain number of iterations, the prediction result of the neural network model cannot converge to less than the given error threshold, the non-destructive characterization parameters need to be replaced, and the types of non-destructive characterization parameters are increased or replaced. In a preferred embodiment, for the sake of simplicity in calculation, before training, the microstructure characteristics and the non-destructive characterization parameters are respectively normalized according to where X is the data to be normalized, X min and X max are respectively the maximum and minimum values in the data, and X' is the data after normalization. In a further preferred embodiment, before the neural network training, the samples in the training sample database can be first verified and calculated. The microstructure characteristics are established as a microstructure matrix A, and the non-destructive characterization parameters are established as a non-destructive characterization matrix B. The ranks r of the matrices are calculated respectively. When r(A) ≤ r(B), it indicates that each microstructure characteristic in the training sample database has a high probability of being expressed as at least a linear combination of a set of non-destructive characterization parameters, and the trained neural network model has a high probability of convergence.
[0027] The mapping model between the superalloy characterization parameters and the microstructure obtained through training can be used to predict the microstructure of the superalloy based on the non-destructive testing data. The process of constructing this mapping model can be written as a calculation program and stored in the memory of a general-purpose computer. When the processor of the general-purpose computer runs the calculation program, the construction of the mapping model can be completed according to the above process. In a preferred embodiment, the general-purpose computer can also store a superalloy heat treatment parameter - microstructure database. After the microstructure of the alloy sample is predicted using the mapping model, the prediction result can be associated with the database, and then the heat history of the alloy sample can be directly given, thereby guiding the optimization of the heat treatment process of the alloy.
[0028] In a preferred embodiment, taking the GH4169 alloy as the characterization object, the process of constructing the mapping model and predicting the microstructure is as Figure 1As shown, it includes two steps: mapping model construction 1 and non-destructive characterization of samples 2. During the aging process of nickel-based superalloys, with the precipitation of precipitate phases, their hardness will change, but this process is not monotonic: before peak aging, the hardness of the alloy matrix increases with aging, while after peak aging, the hardness decreases with aging. There is a similar non-monotonic relationship between the electrical conductivity or Hall coefficient of nickel-based superalloys, and there is an extreme value under specific microstructural states. Since this change relationship is not monotonic, no single parameter can be used to determine the heat treatment state of the alloy. However, the change trends of hardness, electrical conductivity, and Hall coefficient with the evolution of heat treatment microstructure are inconsistent. Therefore, by combining these characterization parameters that can be determined by non-destructive testing, the microstructural state of the alloy can be uniquely determined. Therefore, a mapping model of the characterization parameters and microstructure of GH4169 alloy can be constructed, and the microstructure can be predicted based on the results of non-destructive characterization.
[0029] The specific process is as follows: First, perform the mapping model construction 1 step. Prepare standard specimens of GH4169 alloy. Each standard specimen uses fully annealed GH4169 alloy and is prepared through different heat treatment processes. Specifically, as shown in Table 1, peak aging temperature (720°C), a typical under-aging temperature (600°C), and a typical over-aging temperature (840°C) are respectively used, and 2 intermediate temperatures are evenly inserted among them to complete the temperature gradient setting. A geometric progression of heat treatment duration is established within 1h - 100h to complete the time gradient setting with 7 test times. In other embodiments, more test parameters can also be set according to the types of alloys and test conditions.
[0030]
[0031] Table 1 Heat treatment parameter table of GH4169 alloy standard specimens
[0032] Prepare 49 groups of GH4169 alloy standard specimens according to Table 1, with each group including at least 3 samples. Use non-destructive testing methods to detect the electrical conductivity and hardness of each sample respectively. Among them, the electrical conductivity can be detected by methods such as eddy current and alternating current potential difference, and the detection position should be more than 3mm away from the edge; the hardness can be measured by methods such as Vickers hardness tester or ultrasonic hardness tester; at least 10 sampling points are taken on each sample. Establish an electrical conductivity - microhardness database based on the results of non-destructive testing, and thus obtain the non-destructive characterization parameters of each sample.
[0033] Perform SEM microstructural characterization on each group of samples respectively. Randomly select 20 electron microscope photos at different positions, and use image analysis software to identify and count the average size and area ratio of γ', γ", and δ phases. Thus, the microstructural characteristic parameters of each sample are obtained.
[0034] The non-destructive characterization parameters and microstructural characteristic parameters of each alloy standard specimen together constitute the training samples. To simplify the calculation process, the non-destructive characterization parameters and microstructural characteristic parameters are respectively normalized.
[0035] Next, a neural network model based on the BP model as shown in Figure 2 is established. This neural network includes an input layer 3, a hidden layer 4, and an output layer 5. Among them, the input layer 2 has two nodes, with the conductivity σ and microhardness H as the input quantities respectively. The output layer 5 has six nodes, which are the average size S(γ') and area fraction F(γ') of the γ' phase, the average size S(γ") and area fraction F(γ") of the γ" phase, and the average size S(δ) and area fraction F(δ) of the δ phase. According to the research and experiments of the present invention, in the preferred embodiment, the empirical formula for the number of nodes q in the hidden layer 4 is where N is the number of neurons in the input layer, M is the number of neurons in the output layer, and a is a constant between 0 and 10.
[0036] The normalized training sample data is input into the neural network model for training. The RMSPRrop algorithm is used to iterate 500 times. Three samples are randomly selected from the samples and input into the neural network model for prediction. The error between the predicted value and the measured value of the microstructural parameters does not exceed 5%. This BP neural network is saved as the final mapping model. In the preferred embodiment, before training, a verification calculation is performed on the training samples: a microstructural characteristic parameter matrix A and a non-destructive characterization parameter matrix B are established, and the ranks r(A) and r(B) of matrices A and B are calculated respectively, and r(A) = r(B) is obtained, that is, a one-to-one correspondence can be achieved between the conductivity-hardness and the microstructural parameters in the training samples. It is easy to understand that although for each sample, the non-destructive characterization parameter is a 2D vector and the microstructural characteristic is a 6D vector, the average size and area fraction of each phase in the alloy microstructure are not independent variables.
[0037] Next, the non-destructive characterization step 2 of the sample is carried out. First, the GH4169 alloy sample to be detected is cleaned and polished. Ten positions are randomly selected in the surface area for non-destructive detection of conductivity and hardness. The measured data is normalized, and its average value is input into the mapping model for prediction to obtain the quantitative characteristics of the γ', γ", and δ phases in the microstructure of the sample, so as to realize the non-destructive detection of the alloy sample. The mapping model is associated with the GH4169 alloy heat treatment parameter - microstructure database, and the thermal history data experienced by the sample is provided according to the output result of the mapping model, so as to guide the optimization of the heat treatment process parameters of this part.
[0038] In different embodiments, the methods provided in the above embodiments can also be applied to other types of nickel-based superalloys. The mapping model of the characterization parameters and microstructure of the superalloy established in the above embodiments can be stored on a general-purpose computer for non-destructive characterization of nickel-based superalloy samples; relevant calculation programs can also be pre-stored on the computer, and after inputting different sample data, the program for constructing the mapping model of the characterization parameters and microstructure of the superalloy is automatically executed to complete the construction of the mapping model.
[0039] The purpose of the above embodiments is to further elaborate on the present invention in conjunction with the accompanying drawings so that those skilled in the art can understand the technical concept of the present invention. Within the scope disclosed by the present invention, optimizing or equivalently replacing the involved method steps, and combining the implementation manners in different embodiments without conflict in structure and principle all fall within the protection scope of the present invention.
Claims
1. A method for constructing a mapping model between characterization parameters and microstructures of superalloys, characterized in that, it includes the following steps: Provide the microstructural characteristics of multiple nickel-based superalloy standard samples, as well as multiple non-destructive characterization parameters corresponding to the nickel-based superalloy standard samples, as training samples; Establish a neural network model, use the training samples to train the neural network model, predict the microstructural characteristics according to the non-destructive characterization parameters and perform iteration. When the prediction error is less than a given error threshold, obtain a mapping model between the characterization parameters and microstructures of superalloys; when the prediction error cannot be less than the error threshold after iteration, replace the non-destructive characterization parameters in the training samples.
2. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 1, characterized in that, When training the neural network model, normalize the microstructural characteristics and the non-destructive characterization parameters in the training samples.
3. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 1 or 2, characterized in that, Before training the neural network model, it further includes a step of verifying the training samples: establish the microstructural characteristics in the training samples as a microstructure matrix A, and establish the non-destructive characterization parameters in the training samples as a non-destructive characterization matrix B. When r(A) > r(B), replace the non-destructive characterization parameters in the training samples, where r is the rank of the matrix.
4. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 1 or 2, characterized in that, The neural network model adopts a BP prediction model or a Kriging model.
5. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 4, characterized in that, The neural network model adopts a BP prediction model, including an input layer, an output layer and a hidden layer. The input layer includes 2 nodes, the output layer includes 6 nodes, and the number of nodes in the hidden layer is q. Where N is the number of neurons in the input layer, M is the number of neurons in the output layer, and a is a constant between 0 and 10.
6. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 5, characterized in that, The nodes of the input layer include two of conductivity, microhardness, and Hall coefficient.
7. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 5, characterized in that, The nodes of the output layer include the average size of γ' phase, the area ratio of γ' phase, the average size of γ'' phase, the area ratio of γ'' phase, the average size of δ phase, and the area ratio of δ phase.
8. The method for constructing a mapping model between characterization parameters and microstructures of superalloys according to claim 1 or 2, characterized in that, The given error threshold ≤ 5%.
9. A non-destructive characterization system for superalloys, including a memory and a processor, characterized in that, The memory stores a calculation program, and when the calculation program is executed by the processor, it can implement the method for constructing a mapping model between the characterization parameters and microstructures of superalloys as described in any one of claims 1 to 8.
10. A non-destructive characterization method for superalloys, characterized in that, it includes the following steps: Provide a mapping model of superalloy characterization parameters and microstructure. Input the characterization parameters of the superalloy sample to be characterized into the mapping model of superalloy characterization parameters and microstructure, and obtain the microstructure result predicted by the mapping model of superalloy characterization parameters and microstructure, wherein the mapping model of superalloy characterization parameters and microstructure adopts the mapping model of superalloy characterization parameters and microstructure provided by the construction method of the mapping model of superalloy characterization parameters and microstructure described in any one of claims 1 to 8.
11. According to the non-destructive characterization method of superalloy described in claim 10, characterized in that, it further comprises the following steps: Provide a database of superalloy heat treatment parameters - microstructure. Determine the heat treatment parameters of the superalloy to be characterized according to the microstructure result predicted by the mapping model of superalloy characterization parameters and microstructure.