CN112836433B - Construction method and size identification method of high-temperature alloy grain size identification model - Google Patents
Construction method and size identification method of high-temperature alloy grain size identification model Download PDFInfo
- Publication number
- CN112836433B CN112836433B CN202110186836.5A CN202110186836A CN112836433B CN 112836433 B CN112836433 B CN 112836433B CN 202110186836 A CN202110186836 A CN 202110186836A CN 112836433 B CN112836433 B CN 112836433B
- Authority
- CN
- China
- Prior art keywords
- sample set
- temperature alloy
- prediction model
- model
- original sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B17/00—Measuring arrangements characterised by the use of infrasonic, sonic or ultrasonic vibrations
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention discloses a method and a system for constructing a high-temperature alloy grain size identification model and a method and a system for identifying the size of a high-temperature alloy grain. The construction method comprises the following steps: obtaining an original sample set of the high-temperature alloy crystal grains, wherein the original sample is an ultrasonic characteristic parameter of the high-temperature alloy crystal grains, and the label is the size of the crystal grains; obtaining an extended virtual sample based on Gaussian distribution according to an original sample set; judging whether the prediction precision of the second prediction model is higher than that of the first prediction model, and if so, determining the expanded virtual sample as an effective virtual sample; the first prediction model is a machine learning model obtained by training an original sample set, and the second prediction model is a machine learning model obtained by training an original sample set and an extended virtual sample; and training a high-temperature alloy grain size recognition model by adopting a reconstruction sample set formed by an original sample set and effective virtual samples. The invention expands the sample size and improves the accuracy and effectiveness of grain size identification.
Description
Technical Field
The invention relates to the technical field of high-temperature alloy grain size measurement, in particular to a method and a system for constructing a high-temperature alloy grain size identification model and a method and a system for identifying the size of a high-temperature alloy grain.
Background
The high-temperature alloy can keep excellent fatigue resistance, oxidation resistance and corrosion resistance and good mechanical property under a complex working environment, is generally and widely applied to component parts of aeroengines and various industrial gas turbines, and the grain size has obvious influence on the mechanical property of the high-temperature alloy, so that the effective evaluation of the grain size is very important for in-service inspection and quality detection of the alloy material. In the research field of high-temperature alloy ultrasonic evaluation, the scale of a sample set and the information content of a sample influence the effect of a model, so that the improvement of a small sample space has important value for improving the performance of a prediction model.
However, the sample acquisition cost in the high-temperature alloy material is high or the spatial distribution of the sample information is unbalanced, so that the number of samples is small or the sample information is insufficient, and the model prediction error is large, so that the grain size cannot be accurately identified.
Disclosure of Invention
The invention aims to provide a method and a system for constructing a high-temperature alloy grain size identification model and a method and a system for identifying the grain size of a high-temperature alloy.
In order to achieve the purpose, the invention provides the following scheme:
a method for constructing a high-temperature alloy grain size identification model comprises the following steps:
obtaining an original sample set of the high-temperature alloy crystal grains, wherein the original sample set is composed of high-temperature alloy crystal grain ultrasonic characteristic parameter samples with the labels of the crystal grain sizes;
obtaining an expanded virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to an original sample set;
judging whether the prediction precision of the second prediction model is higher than that of the first prediction model or not, and determining the expanded virtual sample as an effective virtual sample when the prediction precision of the second prediction model is higher than that of the first prediction model; the first prediction model is a machine learning model obtained by training the original sample set, and the second prediction model is a machine learning model obtained by training the original sample set and the extended virtual sample;
and training a high-temperature alloy grain size recognition model by adopting a reconstructed sample set formed by the original sample set and the effective virtual samples.
Optionally, obtaining the extended virtual sample by using a multi-distribution overall trend diffusion method based on gaussian distribution according to the original sample set specifically includes:
based on the original sample set, generating a virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution;
calculating an acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and an acceptable boundary of the crystal grain sizes by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to the original sample set;
and screening the virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample.
Optionally, the method further includes:
constructing a first prediction model based on a regression method of a support vector machine;
training the first predictive model using the original sample set.
Optionally, the method further includes:
constructing a second prediction model based on a support vector machine regression method;
and training the second prediction model by adopting the original sample set and the extended virtual sample to form a sample set.
The invention also provides a method for identifying the grain size of the high-temperature alloy, which comprises the following steps:
carrying out ultrasonic detection on the alloy crystal grain to be identified, and extracting ultrasonic characteristic parameters of the alloy crystal grain to be identified;
and inputting the ultrasonic characteristic parameters into the high-temperature alloy grain size identification model to obtain the size of the alloy grain to be identified.
The invention also provides a construction system of the high-temperature alloy grain size identification model, which comprises the following steps:
the system comprises an original sample set acquisition module, a high-temperature alloy crystal grain ultrasonic characteristic parameter acquisition module and a high-temperature alloy crystal grain ultrasonic characteristic parameter acquisition module, wherein the original sample set acquisition module is used for acquiring an original sample set of high-temperature alloy crystal grains, and the original sample set is composed of high-temperature alloy crystal grain ultrasonic characteristic parameter samples with the labels of the sizes of the crystal grains;
the extended virtual sample obtaining module is used for obtaining extended virtual samples by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to the original sample set;
the effective virtual sample determining module is used for judging whether the prediction precision of the second prediction model is higher than that of the first prediction model or not and determining the expanded virtual sample as an effective virtual sample when the prediction precision of the second prediction model is higher than that of the first prediction model; the first prediction model is a machine learning model obtained by training the original sample set, and the second prediction model is a machine learning model obtained by training the original sample set and the extended virtual sample;
and the Jin Jingli size recognition model training module is used for training a high-temperature alloy grain size recognition model by adopting a reconstruction sample set formed by the original sample set and the effective virtual samples.
Optionally, the extended virtual sample obtaining module specifically includes:
the virtual sample generating unit is used for generating virtual samples by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution based on the original sample set;
the acceptable boundary calculation unit is used for calculating the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the acceptable boundary of the crystal grain sizes according to the original sample set by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution;
and the extended virtual sample screening unit is used for screening a virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample.
Optionally, the system further includes:
the first prediction model building module is used for building a first prediction model based on a support vector machine regression method;
and the first prediction model training module is used for training the first prediction model by adopting the original sample set.
Optionally, the system further includes:
the second prediction model building module is used for building a second prediction model based on the support vector machine regression method;
and the second prediction model training module is used for training the second prediction model by adopting the original sample set and the extended virtual sample to form a sample set.
The invention also provides a system for identifying the grain size of the high-temperature alloy, which comprises the following components:
the ultrasonic detection module is used for carrying out ultrasonic detection on the alloy crystal grains to be identified and extracting ultrasonic characteristic parameters;
and the size identification module is used for inputting the ultrasonic characteristic parameters into the high-temperature alloy crystal grain size identification model to obtain the size of the alloy crystal grain to be identified.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: according to the construction method and the system of the high-temperature alloy grain size identification model, the original sample space is enlarged by using the method of virtual sample generation, the effective virtual sample for model training is obtained through effective screening, the problems of small sample quantity and insufficient sample information in the prior art are solved, and the accuracy and the effectiveness of high-temperature alloy grain size identification are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a method for constructing a grain size identification model of a superalloy provided in embodiment 1 of the present invention;
FIG. 2 is a schematic diagram illustrating the principle of the multi-distribution global trend diffusion technique based on Gaussian distribution in embodiment 1 of the present invention;
FIG. 3 is a schematic diagram illustrating a process of generating a dummy sample according to embodiment 1 of the present invention;
FIG. 4 shows metallographic structures of TC4 superalloy of the present invention at different forging temperatures and forging deformations, FIG. 4 (a) at 920 deg.C, 23% deformation, FIG. 4 (b) at 920 deg.C, 38% deformation, FIG. 4 (c) at 930 deg.C, 23% deformation, FIG. 4 (d) at 930 deg.C, 40% deformation, FIG. 4 (e) at 940 deg.C, 26% deformation, FIG. 4 (f) at 940 deg.C, 40% deformation, FIG. 4 (g) at 950 deg.C, 25% deformation, FIG. 4 (h) at 950 deg.C, 38% deformation, FIG. 4 (i) at 960 deg.C, 25% deformation, FIG. 4 (j) at 960 deg.C, 41% deformation, FIG. 4 (k) at 970 deg.C, 26% deformation, FIG. 4 (l) at 970 deg.C, 42% deformation, FIG. 4 (m) at 980%, 4 (n) at 41% deformation, 4 (o) at 990% deformation, and P) at 990% deformation;
FIG. 5 is a graph showing the comparison between the real sample value and the predicted value under the first prediction model and the grain size identification model of the superalloy.
Fig. 6 is a schematic structural diagram of a system for constructing a grain size identification model of a superalloy provided in embodiment 3 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for constructing a high-temperature alloy grain size identification model and a method and a system for identifying the grain size of a high-temperature alloy.
In order to make the aforementioned objects, features and advantages of the present invention more comprehensible, the present invention is described in detail with reference to the accompanying drawings and the detailed description thereof.
Example 1
Referring to fig. 1, the present embodiment provides a method for constructing a grain size identification model of a superalloy, including the following steps:
step 101: and obtaining an original sample set of the high-temperature alloy crystal grains, wherein the original sample set is composed of the high-temperature alloy crystal grain ultrasonic characteristic parameter samples with the labels of the crystal grain sizes. Each original sample data in the original sample set is derived from ultrasonic characteristic parameters obtained by carrying out ultrasonic detection on the reference test block, and the label of the sample data is the grain size obtained by carrying out metallographic sample preparation on the reference test block.
Step 102: and obtaining the expanded virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to the original sample set.
Step 103: judging whether the prediction precision of the second prediction model is higher than that of the first prediction model or not, and determining the expanded virtual sample as an effective virtual sample when the prediction precision of the second prediction model is higher than that of the first prediction model; the first prediction model is a machine learning model obtained by training an original sample set, and the second prediction model is a machine learning model obtained by training the original sample set and an extended virtual sample. The extended virtual sample may be input into the first prediction model, and the obtained output result may be used as a tag of the extended virtual sample.
Step 104: and training a high-temperature alloy grain size recognition model by adopting a reconstruction sample set formed by an original sample set and effective virtual samples.
As a preferred implementation manner of this embodiment, step 102 is implemented by the following method:
based on the original sample set, generating a virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution; calculating an acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and an acceptable boundary of the crystal grain sizes by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to an original sample set; and screening the virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample to serve as an expanded virtual sample.
The specific operation can be as follows:
representing the ultrasound characteristic parameter in the form of a variable as X = (X) 1 ,X 2 ,...,X M ) The crystal grain size is expressed in the form of variables as Y = (Y) 1 ,Y 2 ,...,Y N ) Wherein M represents the dimension of the ultrasonic characteristic parameter, and N represents the number of original training samples.
1) Referring to fig. 2 and 3, the acceptable boundaries (L, U) of X can be estimated by the multi-distribution global trend diffusion technique based on the gaussian distribution, and the input of the virtual samples is generated within the acceptable boundaries, and the calculation formula of L and U is
Wherein the content of the first and second substances,
wherein L (U) is the lower bound (upper bound) of the acceptable range for X, CL is the data center, and min (max) is the minimum value of X ((max))Max), skaw L (Skew U ) For left (right) skewness of the asymmetric diffusion feature of the data,
is the variance of the original sample set, N L (N U ) Is a number less than (greater than) CL, N represents the number of original samples in the original sample set, X is a value in X,
is the average value of X, odd is odd, even is even, X [] For order statistics in X, m =1.
Within [ min, max ], a gaussian distribution is used to generate virtual samples, which fill the information interval of the original discrete observation points.
2) Training a first prediction model: constructing a first prediction model based on a regression method of a support vector machine; training the first predictive model using the original sample set. Specifically, the following may be used:
normalizing the X to obtain X * ＝(X 1 * ,X 2 * ,...,X M * ) Wherein, in the step (A),
m is the M dimension of X, M is more than or equal to 1 and less than or equal to M, X nm Is composed of
N is more than or equal to 1 and less than or equal to N.
Constructing a prediction model based on a regression method of a support vector machine by using original training samples (X, Y), wherein the function of the prediction model is
Wherein f (x) is the prediction result of the original training sample, and K (x) p ,x i ) Being radial basis functions, α i 、
In order to be a lagrange multiplier,
is the mean value of the offsets, x p For the input of the prediction samples corresponding to the original training samples, x i Is the input of the original training sample.
The output of the virtual sample, that is, the output of the virtual sample, can be predicted by inputting the input value of the virtual sample obtained by the multi-distribution overall trend diffusion technique based on the gaussian distribution into the prediction model function.
Obtaining an acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and an acceptable boundary of the crystal grain sizes according to the calculation method provided in 1), and screening out the virtual samples of which the input and output of the virtual samples are respectively positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the acceptable boundary of the crystal grain sizes to serve as the extended virtual samples.
As a preferred implementation manner of this embodiment, the process of comparing the accuracy of the prediction model in step 103 may be as follows:
setting two indexes for testing the effectiveness of the virtual sample, and carrying out effectiveness analysis on the virtual sample according to the effectiveness indexes, wherein the specific indexes are
In the formula, y i In order to be the true value of the value,
for the prediction of real samples, MAPE is the mean absolute percentage error value, which is used to measure the error of the model. And the OTR is an optimization rate and is used for measuring the improvement amplitude of the model precision. MAPE 1 Is the first stepMeasuring the mean absolute percentage error, MAPE, between the true sample value and the predicted value in the model 2 The average absolute percentage error of the true sample value and the predicted value in the second prediction model.
Inputting the real sample into a second prediction model to obtain the prediction output of the real sample, calculating an OTR value based on the real output and the prediction output of the real sample, judging whether the OTR value is more than or equal to 0, and if so, considering the expanded virtual sample as an effective virtual sample.
Repeating the steps 102 and 103 for a plurality of times, obtaining a plurality of effective virtual samples, forming a reconstructed sample set together with the original samples in the original sample set, and training the grain size identification model of the high-temperature alloy based on the reconstructed sample set.
As a preferred implementation manner of this embodiment, the training process of the second prediction model may be as follows:
constructing a second prediction model based on a support vector machine regression method; and training the second prediction model by adopting the original sample set and the extended virtual sample to form a sample set. Wherein, the second prediction model is specifically:
wherein Q (x) is the prediction result of the reconstructed sample set, Q is the number of the reconstructed sample sets, and K (x) 1p ,x 1i ) Is a radial basis function, beta i 、
In order to be a lagrange multiplier,
is the mean value of the offsets, x 1p For input of prediction samples corresponding to reconstructed samples, x 1i Is the input to reconstruct the sample.
Example 2
The embodiment provides a method for identifying the grain size of a high-temperature alloy, which comprises the following steps: carrying out ultrasonic detection on the alloy crystal grain to be identified, and extracting ultrasonic characteristic parameters of the alloy crystal grain to be identified; and inputting the ultrasonic characteristic parameters into the high-temperature alloy grain size identification model constructed in the embodiment 1 to obtain the size of the alloy grain to be identified. The invention realizes accurate, efficient and nondestructive measurement of the grain size of the high-temperature alloy.
The effect of examples 1 and 2 will be described by taking the titanium alloy TC4 as an example. 16 reference blocks, 4 of which are tested, T 1 、T 2 、T 3 And T 4 . Actual grain size average T of test block 1 ＝13.86μm、T 2 ＝14.58μm、T 3 =15.17 μm and T 4 ＝14.76μm。
By adopting the method provided by the invention, the ultrasonic characteristic parameters of 16 reference test blocks are firstly obtained, the metallographic sample preparation is carried out next step, the microstructure morphology shown in figure 4 is obtained by using an optical microscope, and the average grain size is calculated. And calculating the upper limit and the lower limit of the acceptable range of the ultrasonic characteristic parameters, and generating the input of the virtual sample by using Gaussian distribution. And constructing a first prediction model and predicting the output of the virtual sample. And screening the virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample. Judging whether the prediction precision of the second prediction model is higher than that of the first prediction model, determining the expanded virtual sample as an effective virtual sample when the prediction precision of the second prediction model is higher than that of the first prediction model, and training a high-temperature alloy grain size recognition model by adopting a reconstruction sample set formed by the original sample set and the effective virtual sample after obtaining a plurality of effective virtual samples. The prediction model parameters, the high temperature alloy grain size ultrasonic evaluation function model parameters generated based on the virtual sample, and the traditional single-parameter polynomial fitting evaluation model parameters are shown in table 1.
TABLE 1 relevant parameters of the respective models
Finally, generating a high-temperature alloy grain size ultrasonic evaluation function model pair test block T based on the virtual sample 1 、T 2 、T 3 And T 4 The average grain size was identified without loss and compared with the identification results of the other models, and table 2 shows the identification results of each model and the present invention.
TABLE 2 verification of the respective models in the test block
As can be seen from FIG. 5, the predicted value of the real sample in the superalloy grain size identification model provided by the present invention is closer to the actual measured value of the grain size of the reference block than the predicted value of the real sample in the first prediction model. As shown in table 1, the method effectively improves the accuracy of the evaluation model because the virtual sample information is added. As shown in Table 2, it can be seen from the recognition results of the test block under each model that the MAPE value of the superalloy grain size ultrasonic evaluation function model generated based on the virtual sample is smaller than that of other models. Obviously, compared with the traditional ultrasonic method using a single ultrasonic characteristic parameter and the evaluation method using a plurality of ultrasonic characteristic parameters, the method provided by the invention has the advantages that the range of the high-temperature alloy which can be evaluated is wider, and the accuracy is higher.
Example 3
Referring to fig. 6, the present embodiment provides a system for constructing a grain size identification model of a superalloy, the system including:
an original sample set obtaining module 601, configured to obtain an original sample set of the superalloy crystal grains, where the original sample set is composed of samples of ultrasonic characteristic parameters of the superalloy crystal grains labeled with a crystal grain size;
an extended virtual sample obtaining module 602, configured to obtain an extended virtual sample by using a multi-distribution overall trend diffusion method based on gaussian distribution according to an original sample set;
an effective virtual sample determining module 603, configured to determine whether the prediction accuracy of the second prediction model is higher than that of the first prediction model, and determine that the extended virtual sample is an effective virtual sample when the prediction accuracy of the second prediction model is higher than that of the first prediction model; the first prediction model is a machine learning model obtained by training an original sample set, and the second prediction model is a machine learning model obtained by training an original sample set and an extended virtual sample;
jin Jingli size recognition model training module 604, configured to train a grain size recognition model of a superalloy using an extended sample set composed of an original sample set and valid virtual samples.
The extended virtual sample obtaining module 602 specifically includes:
the virtual sample generating unit is used for generating virtual samples by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution based on an original sample set;
the acceptable boundary calculation unit is used for calculating the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the acceptable boundary of the crystal grain sizes by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to the original sample set;
and the extended virtual sample screening unit is used for screening the virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample.
As a preferred implementation manner of this embodiment, the system for constructing a grain size identification model of a superalloy provided in this embodiment further includes:
the first prediction model building module is used for building a first prediction model based on a support vector machine regression method;
and the first prediction model training module is used for training the first prediction model by adopting the original sample set.
As a preferred implementation manner of this embodiment, the system for constructing a grain size identification model of a superalloy provided in this embodiment further includes:
the second prediction model building module is used for building a second prediction model based on the support vector machine regression method;
and the second prediction model training module is used for training the second prediction model by adopting the original sample set and the extended virtual sample to form a sample set.
Example 4
The embodiment provides a system for identifying grain size of a high-temperature alloy, which comprises: an ultrasonic detection module and a size identification module. The ultrasonic detection module is used for carrying out ultrasonic detection on the alloy crystal grains to be identified and extracting ultrasonic characteristic parameters; and the size identification module is used for inputting the ultrasonic characteristic parameters into the high-temperature alloy grain size identification model related to the embodiment 1, the embodiment 2 or the embodiment 3 to obtain the size of the alloy grain to be identified.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (9)
1. A method for constructing a high-temperature alloy grain size identification model is characterized by comprising the following steps:
obtaining an original sample set of high-temperature alloy crystal grains, wherein the original sample set is composed of high-temperature alloy crystal grain ultrasonic characteristic parameter samples with the labels of the crystal grain sizes;
obtaining an expanded virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to an original sample set;
judging whether the prediction precision of the second prediction model is higher than that of the first prediction model or not, and determining the expanded virtual sample as an effective virtual sample when the prediction precision of the second prediction model is higher than that of the first prediction model; the first prediction model is a machine learning model obtained by training the original sample set, and the second prediction model is a machine learning model obtained by training the original sample set and the extended virtual sample;
training a high-temperature alloy grain size recognition model by adopting a reconstruction sample set formed by the original sample set and the effective virtual sample;
the method includes the following steps that according to an original sample set, an extended virtual sample is obtained by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution, and specifically includes the following steps:
based on the original sample set, generating a virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution;
calculating an acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and an acceptable boundary of the crystal grain sizes by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to the original sample set;
and screening the virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample.
2. The method of constructing a superalloy grain size identification model according to claim 1, further comprising:
constructing a first prediction model based on a regression method of a support vector machine;
training the first predictive model using the original sample set.
3. The method of constructing a superalloy grain size identification model as claimed in claim 1, further comprising:
constructing a second prediction model based on a regression method of a support vector machine;
and training the second prediction model by adopting the original sample set and the extended virtual sample to form a sample set.
4. A method for identifying the grain size of a high-temperature alloy is characterized by comprising the following steps:
carrying out ultrasonic detection on the alloy crystal grain to be identified, and extracting ultrasonic characteristic parameters of the alloy crystal grain to be identified;
inputting the ultrasonic characteristic parameters into the high-temperature alloy grain size identification model in any one of claims 1-3 to obtain the size of the alloy grain to be identified.
5. A system for constructing a grain size identification model of a high-temperature alloy is characterized by comprising the following components:
the system comprises an original sample set acquisition module, a high-temperature alloy crystal grain detection module and a high-temperature alloy crystal grain detection module, wherein the original sample set acquisition module is used for acquiring an original sample set of high-temperature alloy crystal grains, and the original sample set is formed by high-temperature alloy crystal grain ultrasonic characteristic parameter samples with the labels of the crystal grain sizes;
the extended virtual sample obtaining module is used for obtaining extended virtual samples by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to an original sample set, and specifically comprises the following steps:
based on the original sample set, generating a virtual sample by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution;
calculating an acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and an acceptable boundary of the crystal grain sizes by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution according to the original sample set;
screening the virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample;
the effective virtual sample determining module is used for judging whether the prediction precision of the second prediction model is higher than that of the first prediction model or not and determining the expanded virtual sample as an effective virtual sample when the prediction precision of the second prediction model is higher than that of the first prediction model; the first prediction model is a machine learning model obtained by training the original sample set, and the second prediction model is a machine learning model obtained by training the original sample set and the extended virtual sample;
and the Jin Jingli size recognition model training module is used for training a high-temperature alloy grain size recognition model by adopting a reconstruction sample set formed by the original sample set and the effective virtual samples.
6. The system for constructing the grain size identification model of the superalloy according to claim 5, wherein the extended virtual sample acquisition module specifically comprises:
the virtual sample generating unit is used for generating virtual samples by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution based on the original sample set;
the acceptable boundary calculation unit is used for calculating the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the acceptable boundary of the crystal grain sizes according to the original sample set by adopting a multi-distribution overall trend diffusion method based on Gaussian distribution;
and the extended virtual sample screening unit is used for screening a virtual sample in which the ultrasonic characteristic parameters of the high-temperature alloy crystal grains are positioned in the acceptable boundary of the ultrasonic characteristic parameters of the high-temperature alloy crystal grains and the crystal grain sizes are positioned in the acceptable boundary of the crystal grain sizes in the virtual sample as the extended virtual sample.
7. The system for constructing a superalloy grain size identification model according to claim 5, further comprising:
the first prediction model building module is used for building a first prediction model based on a support vector machine regression method;
and the first prediction model training module is used for training the first prediction model by adopting the original sample set.
8. The system for constructing a superalloy grain size identification model according to claim 5, further comprising:
the second prediction model building module is used for building a second prediction model based on the support vector machine regression method;
and the second prediction model training module is used for training the second prediction model by adopting the original sample set and the extended virtual sample to form a sample set.
9. A system for identifying grain size in a superalloy, comprising:
the ultrasonic detection module is used for carrying out ultrasonic detection on the alloy crystal grains to be identified and extracting ultrasonic characteristic parameters;
a size identification module, which is used for inputting the ultrasonic characteristic parameters into the high-temperature alloy grain size identification model in any one of claims 1-3 and 5-8 to obtain the size of the alloy grain to be identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110186836.5A CN112836433B (en) | 2021-02-18 | 2021-02-18 | Construction method and size identification method of high-temperature alloy grain size identification model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110186836.5A CN112836433B (en) | 2021-02-18 | 2021-02-18 | Construction method and size identification method of high-temperature alloy grain size identification model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112836433A CN112836433A (en) | 2021-05-25 |
| CN112836433B true CN112836433B (en) | 2023-03-14 |
Family
ID=75933690
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110186836.5A Active CN112836433B (en) | 2021-02-18 | 2021-02-18 | Construction method and size identification method of high-temperature alloy grain size identification model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112836433B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113804591B (en) * | 2021-09-03 | 2023-05-12 | 南昌航空大学 | High-dimensional ultrasonic evaluation method for nickel-based alloy grain size |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2002243403A1 (en) * | 2000-10-26 | 2002-06-18 | Vextec Corporation | Method and apparatus for predicting the failure of a component |
| CN108898547B (en) * | 2018-06-27 | 2022-06-07 | 太原理工大学 | Single-sample-based face image virtual sample expansion method and system |
| CN109325532A (en) * | 2018-09-18 | 2019-02-12 | 成都网阔信息技术股份有限公司 | The image processing method of EDS extended data set under a kind of small sample |
| CN109497951A (en) * | 2018-12-21 | 2019-03-22 | 中国科学院自动化研究所 | Brain structure and function feature display systems based on brain image |
| CN112132200A (en) * | 2020-09-17 | 2020-12-25 | 山东大学 | Lithology identification method and system based on multi-dimensional rock image deep learning |
-
2021
- 2021-02-18 CN CN202110186836.5A patent/CN112836433B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112836433A (en) | 2021-05-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Müller et al. | Accuracy of fatigue limits estimated by the staircase method using different evaluation techniques | |
| CN108399248A (en) | A kind of time series data prediction technique, device and equipment | |
| CN110321518B (en) | Method for judging trend type of hydrological time series | |
| CN106779217A (en) | Detection of Air Quality method and air quality detection system | |
| CN109313110A (en) | Use the application assessment of digital image correlation technique | |
| CN112836433B (en) | Construction method and size identification method of high-temperature alloy grain size identification model | |
| CN111104736A (en) | Abnormal data detection method, device, medium and equipment based on time sequence | |
| CN110990978A (en) | Bolt state monitoring method and device | |
| JP2023053249A (en) | Piping diagnostic device, piping diagnostic method, and program | |
| CN111044614A (en) | High-temperature alloy grain size circle-like mapping ultrasonic evaluation method | |
| CN110188399B (en) | Dam safety monitoring single-measuring-point evaluation method based on multiple correlation sequences | |
| Li et al. | Magnetic barkhausen noise technique for early-stage fatigue prediction in martensitic stainless-steel samples | |
| Meng et al. | Prediction of fault evolution and remaining useful life for rolling bearings with spalling fatigue using digital twin technology | |
| CN111241725B (en) | Structure response reconstruction method for generating countermeasure network based on conditions | |
| CN113804591B (en) | High-dimensional ultrasonic evaluation method for nickel-based alloy grain size | |
| Zhang et al. | Multivariate discrete grey model base on dummy drivers | |
| JP7281708B2 (en) | Manufacturing condition calculation device, manufacturing condition calculation method, and manufacturing condition calculation program for identifying equipment that contributes to the generation of defective products | |
| EP4116697A1 (en) | Method and system for generating a test coupon specification for predicting fatigue life of a component | |
| CN112949000B (en) | Component residual stress inversion method based on convolutional neural network model | |
| CN111210409B (en) | Condition-based generation confrontation network structure damage identification method | |
| CN111241614B (en) | Engineering structure load inversion method based on condition generation confrontation network model | |
| JP2010079465A (en) | System and method for estimating process | |
| Lindgren | Opportunities for nondestructive evaluation: quantitative characterization | |
| CN116050252B (en) | Structure fatigue damage prediction method and device and storage medium | |
| Katam et al. | SVM-assisted damage identification in cantilever steel beam using vibration-based method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |