A method and system for ceramic material bond strength prediction

Abstract

The application relates to a ceramic material bonding strength prediction method and system, and belongs to the technical field of data processing.The ceramic material bonding strength prediction method provided by the application can accurately realize the prediction of the bonding strength of the ceramic material in the thermal barrier layer by adopting a bonding strength prediction model according to the acquired process parameters of the ceramic material in the to-be-predicted thermal barrier layer, and further solves the problems of low bonding strength judgment efficiency and high cost in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a ceramic material bonding strength prediction method and system. BACKGROUND

[0002] Thermal barrier coatings (TBCs) are widely used to provide thermal protection for hot metal components in aero-engines, diesel engines, to improve the thermal efficiency and performance of the components. TBCs are usually composed of a metallic bond coat and a ceramic layer, and yttria-stabilized zirconia (YSZ) ceramic is considered to be an ideal TBCs material. When Y2O3 is added to ZrO2 as a stabilizer, it can form a stable or partially stable structure at high temperatures. For TBCs of turbine blades, the adhesion performance of the coating (usually considered to be the bonding strength between the coating and the bond coat interface, sometimes considered to be the bonding strength of the coating) is an important indicator for evaluating the quality of the coating. Strong adhesion between the coating and the bond coat is very important for achieving effective thermal insulation, and will significantly affect the service life of the coating. Existing researches change the process parameters of the coating to obtain ceramic materials with higher bonding strength, but there are problems of low efficiency and high cost in judging the bonding strength. SUMMARY

[0003] To solve the above problems existing in the prior art, the present application provides a ceramic material bonding strength prediction method and system.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] A ceramic material bonding strength prediction method, comprising:

[0006] Obtaining process parameters of ceramic materials in a thermal barrier layer to be predicted and a bonding strength prediction model; the process parameters include: current, voltage, spraying power, Ar flow rate ratio, H2 flow rate ratio, Ar / H2 flow rate ratio and spraying thickness; the bonding strength prediction model is a support vector machine model that has been trained and tested;

[0007] Inputting the process parameters into the bonding strength prediction model to obtain a bonding strength prediction result.

[0008] Preferably, the construction process of the bonding strength prediction model comprises:

[0009] Obtaining experimental sample data;

[0010] Pretreating the experimental sample data to obtain a training set and a test set;

[0011] Training and testing a support vector machine model using the training set and the test set to obtain the bonding strength prediction model.

[0012] Preferably, the experimental sample data is preprocessed to obtain a training set and a test set, specifically including:

[0013] The experimental sample data is subjected to variable elimination processing to obtain an initial sample data set.

[0014] The sample data in the initial sample data set is subjected to data augmentation processing to obtain a sample data set.

[0015] The sample data set is divided into the training set and the test set according to a preset ratio.

[0016] Preferably, the sample data in the initial sample data set is subjected to data augmentation processing by using a Gaussian mixture model to obtain a sample data set.

[0017] According to the specific embodiments of the present application, the following technical effects are disclosed:

[0018] The ceramic material bonding strength prediction method provided by the present application can accurately predict the bonding strength of the ceramic material in the thermal barrier layer by using a bonding strength prediction model and according to the obtained process parameters of the ceramic material in the thermal barrier layer to be predicted, thereby solving the problems of low bonding strength judgment efficiency and high cost in the prior art.

[0019] Corresponding to the above-mentioned ceramic material bonding strength prediction method, the present application further provides a ceramic material bonding strength prediction system, which comprises:

[0020] A parameter model acquisition module is configured to acquire process parameters of a ceramic material in a thermal barrier layer to be predicted and a bonding strength prediction model; the process parameters include current, voltage, spraying power, Ar flow rate ratio, H2 flow rate ratio, Ar / H2 flow rate ratio, and spraying thickness; and the bonding strength prediction model is a support vector machine model that has been trained and tested.

[0021] A strength prediction module is configured to input the process parameters into the bonding strength prediction model to obtain a bonding strength prediction result.

[0022] Preferably, the system further comprises:

[0023] A data acquisition module is configured to acquire experimental sample data.

[0024] A preprocessing module is configured to preprocess the experimental sample data to obtain a training set and a test set.

[0025] A training and testing module is configured to train and test a support vector machine model by using the training set and the test set to obtain the bonding strength prediction model.

[0026] Preferably, the preprocessing module comprises:

[0027] a variable elimination unit, configured to perform variable elimination processing on the experimental sample data to obtain an initial sample data set;

[0028] a data augmentation unit, configured to perform data augmentation processing on sample data in the initial sample data set to obtain a sample data set;

[0029] a data division unit, configured to divide the sample data set into the training set and the test set according to a preset ratio.

[0030] Preferably, the data augmentation unit is configured to perform data augmentation processing on sample data in the initial sample data set by using a Gaussian mixture model to obtain a sample data set.

[0031] The technical effects achieved by the ceramic material bonding strength prediction system provided in the application are the same as those achieved by the ceramic material bonding strength prediction method provided in the application, and thus will not be described here again. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0033] Figure 1 a flowchart of the ceramic material bonding strength prediction method provided in the application;

[0034] Figure 2 a data processing flowchart of the ceramic material bonding strength prediction method provided in the embodiment of the application;

[0035] Figure 3 a regression model modeling result graph of the YSZ ceramic material bonding strength provided in the embodiment of the application;

[0036] Figure 4 a leave-one-out cross-validation result graph of the regression model of the YSZ ceramic material bonding strength provided in the embodiment of the application;

[0037] Figure 5 a test set result graph of the regression model of the YSZ ceramic material bonding strength provided in the embodiment of the application;

[0038] Figure 6 a prediction result graph of the regression model of the YSZ ceramic material bonding strength provided in the embodiment of the application;

[0039] Figure 7A structural schematic diagram of a ceramic material bonding strength prediction system provided by the present application is shown. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0041] The present application aims to provide a ceramic material bonding strength prediction method and system, which can improve the prediction accuracy and solve the problems of low bonding strength judgment efficiency and high cost in the prior art.

[0042] In order to make the above objectives, characteristics and advantages of the present application more apparent, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0043] As shown in the drawings, Figure 1 The present application provides a ceramic material bonding strength prediction method, which comprises the following steps:

[0044] Step 100: Obtain the process parameters and bonding strength prediction model of the ceramic material in the thermal barrier layer to be predicted. The process parameters include current, voltage, spraying power, Ar flow rate ratio, H2 flow rate ratio, Ar / H2 flow rate ratio and spraying thickness. The bonding strength prediction model is a support vector machine model that has been trained and tested.

[0045] Step 101: Input the process parameters into the bonding strength prediction model to obtain the bonding strength prediction result.

[0046] The construction process of the bonding strength prediction model used above comprises the following steps:

[0047] Step 1: Obtain experimental sample data.

[0048] Step 2: Preprocess the experimental sample data to obtain a training set and a test set. This step specifically comprises the following steps:

[0049] Step 2-1: Perform variable elimination processing on the experimental sample data to obtain an initial sample data set. In the variable elimination process, the spraying power and the Ar / H2 flow rate ratio in the collected experimental sample data need to be calculated based on the existing characteristic variables, which are redundant information, so they need to be eliminated. Constant process variable H2 flow rate ratio can also be eliminated.

[0050] Step 2-2: Perform data augmentation on the sample data in the initial sample dataset to obtain the final sample dataset. Specifically, a Gaussian mixture model is used to augment the sample data in the initial sample dataset to obtain the final sample dataset.

[0051] Steps 2-3: Divide the sample dataset into training and test sets according to a preset ratio.

[0052] Step 3: Train and test the support vector machine model using the training and test sets to obtain the binding strength prediction model.

[0053] The following is as follows Figure 2 The data processing flow shown is used as an example to illustrate the specific implementation process of the above-mentioned method for predicting the bonding strength of ceramic materials, taking the prediction of the bonding strength of yttrium-stabilized zirconia (YSZ) ceramic materials as an example.

[0054] Example 1

[0055] 1) Experimental data collection:

[0056] The TBC system used in this experiment includes a metal substrate, a metal adhesive layer, and a ceramic layer. A nickel-based superalloy was selected as the substrate to prepare a cuboid with dimensions of 20mm × 10mm × 2mm and a ceramic layer with dimensions of… Cylindrical samples were prepared. NiCrCoAlY powder was vacuum plasma-sprayed to create an adhesive layer on the substrate. The chemical composition of the adhesive layer powder and the spraying parameters were recorded. The ceramic layer was then formed by atmospheric plasma spraying of ZrO2-4 mol% Y2O3 onto the adhesive layer. Process parameters for coating performance, such as spraying power, spraying thickness, and Ar / H2 flow rate ratio, were adjusted. The YSZ coating and the adhesive layer were measured using a universal testing machine (Instron-5592). The bonding strength between the substrates was measured. Each sample was measured three times to reduce error, resulting in a total of nine samples. The experimental samples are shown in Table 1.

[0057] Table 1 Sample Data Table

[0058]

[0059]

[0060] 2) Establish a sample set:

[0061] Based on the nine YSZ ceramic material experimental samples obtained in step 1), a dataset was constructed with bonding strength as the target variable and seven process conditions (current, voltage, spraying power, Ar, H2, Ar / H2 flow rate ratio, and spraying thickness) as feature variables. Since spraying power and Ar / H2 were calculated based on existing feature variables and were considered redundant, they were removed. The constant process variable H2 was also removed.

[0062] 3) Gaussian mixture sample enhancement:

[0063] Based on the nine experimental sample data in step 2), a Gaussian Mixture Model (GMM) was used to expand this small sample dataset. When using GMM to fit the data distribution, the number of individual Gaussian models constituting the GMM has a significant impact on the obtained sample distribution. Too few models will result in inaccurate model fitting, while too many models will lead to overfitting. This invention adaptively selects the number of individual Gaussian models based on the Akaike information criterion (AIC) to balance the model's accuracy and generalization performance. The smaller the AIC value, the better the obtained sample distribution is guaranteed to be optimal. After obtaining the Gaussian mixture distribution of the original data, virtual samples were generated using random sampling based on the probability density function of the obtained original data, resulting in 450 virtual sample data.

[0064] 4) Divide the dataset into training and testing sets:

[0065] The 450 virtual samples obtained in step 3) are randomly divided into a training set and a test set. For example, the ratio of the training set to the test set is 4:1.

[0066] 5) Construct a binding strength prediction model:

[0067] The support vector machine model is trained using the training set from step 4), and hyperparameter optimization and leave-one-out cross-validation are performed. Based on the trained support vector machine model, the accuracy is tested using sample data from the test set.

[0068] The Gaussian mixture model is a probabilistic model that assumes all data points are generated from a mixture of a finite number of Gaussian distributions. If there are n observations X = {X1, ..., Xn}... n} is generated by a mixture distribution P, and each vector X is generated by a mixture distribution P. i Since all are p-dimensional, and the distribution P is composed of G components, the maximum mixture likelihood function of the distribution is shown in equation (1).

[0069]

[0070] In the formula, f k (x i |θ k ) represents X i It is the density function of the k-th class, θ k These are the corresponding parameters, π k It is a weight parameter, which represents the probability that a certain observation belongs to the k-th class.

[0071] If f k (x i |θ k If θ is a multivariate normal distribution, then P is a Gaussian mixture distribution. k From the mean μ k The covariance matrix Σ k Composition. Density function f k (x i |θ k As shown in equation (2):

[0072]

[0073] The Gaussian mixture distribution can be described by a probability density function represented by the weighted average of G Gaussian density functions, as shown in equation (3) below:

[0074]

[0075] Fundamentally, GMM is a density estimation algorithm. As can be seen from formula (3), by adjusting the weights π k This greatly influences the probability density function curve of the mixture model, which is then used to fit a nonlinear function of arbitrary shape. Based on this, a generative probability model describing small sample data is constructed, and the parameters are solved using the EM algorithm. Finally, the obtained generative model is used to generate virtual samples that meet the expectations.

[0076] Support Vector Machine (SVM) algorithms are based on insensitivity functions and kernel functions. If the fitted mathematical model is expressed as a curve in a multidimensional space, the result obtained according to the insensitivity function is the "pipe" that encloses the curve and the training points. Of all the sample points, only the points distributed on the "pipe wall" determine the position of the pipe; these training samples are called "support vectors." SVM regression aims to find a regression plane that minimizes the distance to the hyperplane for all data points in a set. For SVM, the kernel function allows it to determine the separating hyperplane with fewer support vectors, giving it good performance even with small sample sizes.

[0077] The present application is based on nine experimental samples prepared under different process conditions, and a virtual sample generation algorithm is proposed using a Gaussian mixture model to sample and generate virtual samples by fitting the distribution of the original samples. Based on the generated 450 virtual samples, a support vector machine algorithm is used to establish a bonding strength prediction model, and the prediction results as shown in Figure 3 are obtained, effectively improving the accuracy of the model. This work is expected to help improve the adhesion strength, service life and thermal insulation performance of the coating.

[0078] Example Two:

[0079] This example has basically the same process framework as Example One, and the special features are as follows:

[0080] In this example, the 360 data samples in the training set divided in step 4) of Example One are sequentially numbered as A1, A2, …, A360. In the first step, A1, A2, …, A359 are used to form a new training set, and a model M1 is established to predict the bandgap width value of sample A360. Next, this is followed by A2, …, A359, and a model is established to predict the bandgap width value of the remaining single sample. After iteration, the target variables of A1, A2, …, A360 are predicted, and 359 models M1, M2, …, M360 are established based on the remaining 359 samples. The error between the predicted value and the experimental value is used to further judge the reliability of the modeling method and the stability of the model.

[0081] The leave-one-out cross-validation results of the YSZ ceramic material bonding strength model established based on 360 samples and the support vector machine algorithm are shown in Figure 4 . The correlation coefficient between the predicted value and the actual value of the sample in the leave-one-out internal cross-validation is 0.9881, and the RMSE value is 1.3019. This example further evaluates the stability and reliability of the YSZ ceramic material bonding strength prediction model, and can more accurately and efficiently predict the bonding strength of YSZ ceramic materials.

[0082] Example Three:

[0083] The process of this example is basically the same as the above examples, and the special features are as follows:

[0084] In this example, the bonding strength values of the test set samples are quickly predicted based on the established YSZ ceramic material bonding strength prediction model. The independent test set prediction results of the YSZ ceramic material bonding strength prediction model established based on 360 sample data and the support vector machine are shown in Figure 5 .

[0085] The method of the embodiment uses the bonding strength prediction model of the YSZ ceramic material established based on 360 training set samples to predict 90 samples in the independent test set. The predicted bonding strength values correspond to the actual values of the samples, and the correlation coefficient R obtained is 0.9899, indicating that the model has good extrapolation generalization ability and can efficiently and quickly predict the bonding strength of the YSZ ceramic material.

[0086] Embodiment Four

[0087] The process of the present example is basically the same as the above-mentioned embodiments, and the special features are as follows:

[0088] In the present embodiment, the initial 9 experimental samples are predicted for bonding strength according to the established bonding strength prediction model of the YSZ ceramic material, and the prediction results are as shown in Table 2. Figure 6

[0089] The method of the present embodiment is based on the bonding strength prediction model of the YSZ ceramic material established based on 360 training set samples to predict the bonding strength of the 9 samples obtained by the initial experiment. The predicted bonding strength values correspond to the bonding strength values measured by the experiment, and the correlation coefficient R obtained is 0.7897. The model itself has good generalization ability and stability, and the prediction accuracy is greatly improved compared to directly using 9 samples for modeling, which can more efficiently predict the bonding strength of the YSZ ceramic material and save the time and resource costs of the researchers.

[0090] The above describes the embodiments of the present application in conjunction with the drawings, but the present application is not limited to the above-mentioned embodiments, and can be based on the present application as a starting point, and various changes can be made according to different research objects and target properties. Any change, modification, substitution, combination or simplification made in accordance with the spirit and principles of the present application shall be an equivalent replacement, as long as it conforms to the technical principles and concepts of the present application for quickly predicting the bonding strength of the YSZ ceramic material based on GMM-SVM, and belongs to the protection scope of the present application.

[0091] Based on the above description, the present application has the following advantages compared to the prior art:

[0092] 1. Compared with the commonly used experimental method, the present application uses a machine learning algorithm to predict the bonding strength of the YSZ ceramic material, which is more efficient and convenient, and saves a lot of time and resource costs. It can quickly predict the bonding strength based on known process conditions, greatly improving the calculation efficiency.

[0093] ​2. Regarding the model itself, this invention employs a Gaussian Mixture Model (GMM) to predict the bonding strength of ceramic materials. The virtual samples generated by the GMM extrapolate the original dataset to a certain extent, expanding the original training set. This results in a more efficient, faster, and lower-cost method that can quickly build a model to predict bonding strength based on a small number of experimental samples, significantly improving the accuracy of the prediction.

[0094] 3. This invention builds models based on a smaller number of experimental samples, reducing the consumption of experimental chemicals during performance exploration. The model is simple, requires fewer computational resources, and causes no environmental pollution, making it energy-efficient and environmentally friendly.

[0095] 4. This invention can provide guidance for the preparation of ceramic materials with high bonding strength, explore better process conditions, reduce the blindness of experiments, further save experimental resources, and the model is also applicable to other similar systems with the same small sample size.

[0096] Corresponding to the above-mentioned method for predicting the bonding strength of ceramic materials, the present invention also provides a system for predicting the bonding strength of ceramic materials, such as... Figure 7 As shown, the system includes:

[0097] The parameter model acquisition module 700 is used to acquire the process parameters and bonding strength prediction model of the ceramic material in the thermal barrier layer to be predicted. The process parameters include: current, voltage, spraying power, Ar flow rate ratio, H2 flow rate ratio, Ar / H2 flow rate ratio, and spraying thickness. The bonding strength prediction model is a well-trained and tested support vector machine model.

[0098] The strength prediction module 701 is used to input process parameters into the bonding strength prediction model to obtain the bonding strength prediction result.

[0099] To improve the accuracy of strength prediction, as an embodiment of the present invention, the ceramic material bonding strength prediction system provided above further includes:

[0100] The data acquisition module is used to acquire experimental sample data.

[0101] The preprocessing module is used to preprocess the experimental sample data to obtain the training set and the test set.

[0102] The training and testing module is used to train and test the support vector machine model using the training and testing sets to obtain the combination strength prediction model.

[0103] The preprocessing module includes:

[0104] The variable removal unit is used to remove variables from the experimental sample data to obtain the initial sample dataset.

[0105] A data augmentation unit is configured to perform data augmentation on sample data in an initial sample data set to obtain a sample data set.

[0106] A data division unit is configured to divide the sample data set into the training set and the test set according to a preset ratio.

[0107] Embodiment five:

[0108] This embodiment is basically the same as the above-mentioned embodiments, and the particularity is that:

[0109] In this embodiment, a system for rapidly predicting the bonding strength of YSZ ceramic material is implemented to perform the method for predicting the bonding strength of YSZ ceramic material based on GMM-SVR with a small sample according to the above-mentioned example, and the system comprises:

[0110] An input module: data augmentation is performed on 9 samples obtained through experiments by using a Gaussian mixture model, and the samples are used as input data.

[0111] A data analysis module: the data obtained by the input module is used to perform the method for rapidly predicting the bonding strength of YSZ ceramic material based on GMM-SVR, and the bonding strength of YSZ ceramic material is rapidly predicted.

[0112] An output module: the data of rapidly predicting the bonding strength of YSZ ceramic material is output.

[0113] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method part.

[0114] The principles and implementation manners of the present application are described by using specific examples in this paper, and the above-mentioned embodiment is only used to help understand the method of the present application and its core idea; at the same time, for the general technical personnel in the art, according to the idea of the present application, the specific implementation manner and application range will be changed. In conclusion, the content of the specification should not be understood as the limitation of the present application.

Claims

1. A method of predicting the bonding strength of a ceramic material, characterized by, The method comprises the following steps: obtaining process parameters of ceramic materials in a thermal barrier coating layer to be predicted and a bonding strength prediction model; the process parameters include current, voltage, spraying power, Ar flow rate ratio, H2 flow rate ratio, Ar / H2 flow rate ratio and spraying thickness; and the bonding strength prediction model is a support vector machine model that has been trained and tested; inputting the process parameters into the bonding strength prediction model to obtain a bonding strength prediction result; the process of constructing the bonding strength prediction model comprises the following steps: obtaining experimental sample data; performing variable elimination processing on the experimental sample data to obtain a small sample data set; wherein the spraying power and Ar / H2 flow rate ratio in the experimental sample data are eliminated as variables; performing data expansion processing on sample data in the small sample data set by using a Gaussian mixture model to obtain a sample data set; wherein, when fitting the data distribution by using the Gaussian mixture model, the number of single Gaussian models is adaptively selected according to the Akaike information criterion; after obtaining the Gaussian mixture distribution of the original data, virtual sample data is generated by using the probability density function of the obtained original data in a random sampling manner, and 450 pieces of virtual sample data are obtained; dividing the sample data set into the training set and the test set according to a preset ratio; training and testing a support vector machine model by using the training set and the test set to obtain the bonding strength prediction model.

2. A ceramic material bond strength prediction system characterized by, The method comprises the following steps: a parameter model acquisition module is configured to obtain process parameters of ceramic materials in a thermal barrier coating layer to be predicted and a bonding strength prediction model; the process parameters include current, voltage, spraying power, Ar flow rate ratio, H2 flow rate ratio, Ar / H2 flow rate ratio and spraying thickness; and the bonding strength prediction model is a support vector machine model that has been trained and tested; a strength prediction module is configured to input the process parameters into the bonding strength prediction model to obtain a bonding strength prediction result; a data acquisition module is configured to obtain experimental sample data; a preprocessing module is configured to preprocess the experimental sample data to obtain a training set and a test set; a training and testing module is configured to train and test a support vector machine model by using the training set and the test set to obtain the bonding strength prediction model. The preprocessing module comprises the following steps: a variable elimination unit is configured to perform variable elimination processing on the experimental sample data to obtain a small sample data set; wherein the spraying power and Ar / H2 flow rate ratio in the experimental sample data are eliminated as variables; a data expansion unit is configured to perform data expansion processing on sample data in the small sample data set by using a Gaussian mixture model to obtain a sample data set; wherein, when fitting the data distribution by using the Gaussian mixture model, the number of single Gaussian models is adaptively selected according to the Akaike information criterion; after obtaining the Gaussian mixture distribution of the original data, virtual sample data is generated by using the probability density function of the obtained original data in a random sampling manner, and 450 pieces of virtual sample data are obtained; a data division unit is configured to divide the sample data set into the training set and the test set according to a preset ratio.

Images