CN112382354A - Cr12MoV material hardness prediction method and device - Google Patents
Cr12MoV material hardness prediction method and device Download PDFInfo
- Publication number
- CN112382354A CN112382354A CN202011357828.4A CN202011357828A CN112382354A CN 112382354 A CN112382354 A CN 112382354A CN 202011357828 A CN202011357828 A CN 202011357828A CN 112382354 A CN112382354 A CN 112382354A
- Authority
- CN
- China
- Prior art keywords
- regression
- barkhausen noise
- hardness
- noise data
- prediction
- 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.)
- Granted
Links
- 239000000463 material Substances 0.000 title claims abstract description 87
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000012549 training Methods 0.000 claims abstract description 96
- 230000005330 Barkhausen effect Effects 0.000 claims abstract description 94
- 239000011159 matrix material Substances 0.000 claims abstract description 37
- 238000013507 mapping Methods 0.000 claims abstract description 30
- 239000013598 vector Substances 0.000 claims description 27
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 230000002068 genetic effect Effects 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 6
- 238000009659 non-destructive testing Methods 0.000 abstract description 5
- 238000007542 hardness measurement Methods 0.000 abstract 1
- 238000004364 calculation method Methods 0.000 description 9
- 238000011156 evaluation Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000012935 Averaging Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000002349 favourable effect Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Evolutionary Biology (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Molecular Biology (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Investigating Or Analyzing Materials By The Use Of Magnetic Means (AREA)
Abstract
The invention relates to a Cr12MoV material hardness prediction method and a device, belongs to the technical field of electromagnetic nondestructive testing, and solves the problems of poor precision and high complexity of the existing mechanical material hardness testing method. The method comprises the following steps: acquiring Barkhausen noise data of a Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set; training a training data set based on a cascade regression Laguerre polynomial fitting method to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters; obtaining a trained regression model based on the regression coefficient, the principal component mapping matrix and the boundary parameter; the method comprises the steps of obtaining Barkhausen noise data of a Cr12MoV material to be predicted, and conducting hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted based on a trained regression model to obtain a prediction result. The method realizes the prediction of the hardness of the Cr12MoV material, and improves the prediction efficiency and the prediction precision.
Description
Technical Field
The invention relates to the technical field of electromagnetic nondestructive testing, in particular to a Cr12MoV material hardness prediction method and a Cr12MoV material hardness prediction device.
Background
Under the influence of temperature, chemical corrosion and other factors, the mechanical properties of the material usually affect the operation state and life cycle of key components in the machine structure. The current common material mechanical property detection means comprise destructive and nondestructive detection methods. In order to save cost and highlight efficiency, an electromagnetic nondestructive testing method is a new mechanical property monitoring means for materials at present.
The existing method for predicting the mechanical property of the material by utilizing the electromagnetic nondestructive testing technology mainly extracts signal characteristics and utilizes a prediction model to detect the mechanical property of the material. The method has the following two defects in two links of feature extraction and prediction model selection correspondingly: firstly, the features extracted in the current detection model are mainly some more general and mutually isolated features, and although some of the extracted features may have certain physical significance, the extracted features are greatly disturbed by the pseudo-randomness of the Barkhausen noise signal, so that the accuracy of mechanical property prediction is influenced; secondly, most of prediction methods used in the current detection model are mature algorithms, mainly including approximate linear regression and a shallow neural network, and of the two algorithms, the approximate linear regression method is simple to implement and clear in theory, but the main problem is that the approximation precision of the method to the nonlinear fitting problem is poor; the shallow neural network is mainly limited by the problems of relatively simple structure, poor convergence effect, more parameters needing to be optimized and the like of the shallow neural network. In recent years, researchers have proposed that fitting of a multivariate nonlinear problem is achieved by using a combination of orthogonal polynomials, but such a method is relatively expensive in storage and calculation, and does not have conditions for practical use.
Disclosure of Invention
In view of the foregoing analysis, embodiments of the present invention provide a Cr12MoV material hardness prediction method and apparatus, so as to solve the problems of poor accuracy and high complexity of the existing mechanical material hardness detection method.
On one hand, the embodiment of the invention provides a Cr12MoV material hardness prediction method, which comprises the following steps:
acquiring Barkhausen noise data of a Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set;
training the training data set based on a cascade regression Laguerre polynomial fitting method to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters; obtaining a trained regression model based on the regression coefficient, the principal component mapping matrix and the boundary parameter;
and acquiring Barkhausen noise data of the Cr12MoV material to be predicted, and performing hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result.
The method comprises the following steps of obtaining Barkhausen noise data of a Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material, wherein the characteristic information comprises a peak value, a peak position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width;
splicing the characteristic information to obtain a characteristic vector of the Barkhausen noise data;
and establishing a training data set based on the eigenvectors and the hardness labels of the Barkhausen noise data.
Training the training data set based on cascade regression Laguerre polynomial fitting to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters, and the method comprises the following steps of:
obtaining optimal laguerre polynomial parameters based on the training data set and a genetic algorithm;
and obtaining the regression coefficient, the principal component mapping matrix and the boundary parameter of each sub-regression in the whole cascade regression based on the optimal Laguerre polynomial parameter.
Further, obtaining optimal laguerre polynomial parameters based on the training data set and genetic algorithm, comprising the steps of:
generating a population member set for a given laguerre polynomial parameter;
calculating the fitness of each population member in the population member set based on the training data set;
sorting and grouping the population member sets based on the fitness to obtain the optimal and worst population members of each subgroup;
and performing iterative updating of the worst population member in the group based on the optimal and worst population members of each subgroup until an iterative condition is met to obtain an optimal Laguerre polynomial parameter.
Further, hardness prediction is carried out on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result, and the method comprises the following steps:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material to be predicted, wherein the characteristic information comprises a peak value, a peak position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width;
splicing the characteristic information to obtain a characteristic vector;
obtaining a sample to be predicted based on the feature vector, and initializing a predicted value to be zero;
and inputting the sample to be predicted into a regression model for hardness prediction to obtain a prediction result.
Further, the method for predicting the hardness of the sample to be predicted is input into a regression model to perform hardness prediction to obtain a prediction result, and comprises the following steps of:
projecting the input sample of each iteration to a principal component space based on the principal component mapping matrix output in the training process;
calculating the order values of the unary Laguerre polynomial corresponding to the input sample;
generating regression output of the iteration process by using the weighted summation of the regression coefficients of the corresponding regressors generated aiming at the hardness labels in the training process, and accumulating the regression output generated by the iteration to a predicted value;
carrying out weighted summation on the projected samples according to regression coefficients of corresponding regressors generated aiming at the input samples, generating regression output of the iteration process, and acquiring regression residual error as input of the next iteration;
repeating iterative updating until the total iteration times are reached to obtain a predicted value;
and carrying out inverse normalization processing on the generated final prediction value to obtain a final prediction result.
On the other hand, an embodiment of the present invention provides a Cr12MoV material hardness prediction apparatus, including:
the data acquisition module is used for acquiring Barkhausen noise data of the Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set;
the training module is used for training the training data set according to cascade regression Laguerre polynomial fitting to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters, and a trained regression model is obtained based on the regression coefficients, the principal component mapping matrixes and the boundary parameters;
and the prediction module is used for acquiring the Barkhausen noise data of the Cr12MoV material to be predicted, and performing hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result.
Further, the data acquisition module is configured to:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material, wherein the characteristic information comprises a peak value, a peak position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width;
splicing the characteristic information to obtain a characteristic vector of the Barkhausen noise data;
and establishing a training data set based on the eigenvectors and the hardness labels of the Barkhausen noise data.
Further, the training module is to:
obtaining optimal laguerre polynomial parameters based on the training data set and a genetic algorithm;
and obtaining the regression coefficient, the principal component mapping matrix and the boundary parameter of each sub-regression in the whole cascade regression based on the optimal Laguerre polynomial parameter.
Further, obtaining optimal laguerre polynomial parameters based on the training data set and genetic algorithm, comprising the steps of:
generating a population member set for a given laguerre polynomial parameter;
calculating the fitness of each population member in the population member set based on the training data set;
sorting and grouping the population member sets based on the fitness to obtain the optimal and worst population members of each subgroup;
and performing iterative updating of the worst population member in the group based on the optimal and worst population members of each subgroup until an iterative condition is met to obtain an optimal Laguerre polynomial parameter.
Compared with the prior art, the invention can realize at least one of the following beneficial effects:
1. a method and a device for predicting hardness of a Cr12MoV material are disclosed, and the method comprises the steps of firstly obtaining Barkhausen noise data of the Cr12MoV material through a Barkhausen noise signal sensor, obtaining a hardness label value of the Cr12MoV material through external equipment of a professional evaluation mechanism, establishing a training data set through the Barkhausen noise data and the hardness label value corresponding to the Barkhausen noise data, then training the training data set through a cascade regression Laguerre polynomial fitting method to obtain a trained regression model, and finally conducting hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted through the trained regression model to obtain a prediction result. The method is simple and easy to implement, realizes the hardness prediction of the Cr12MoV material to be predicted, improves the prediction precision and has higher practical value.
2. The method comprises the steps of obtaining Barkhausen noise data of a Cr12MoV material by using a Barkhausen noise signal sensor, obtaining a hardness label value of the Cr12MoV material by using a special evaluation mechanism external device, carrying out feature extraction on the Barkhausen noise data to obtain a plurality of features, obtaining a feature vector of the Barkhausen noise data according to the features, and constructing a training data set by using the feature vector and the hardness label value of the Barkhausen noise data, so that technical support and basis are provided for later training of a Laguerre polynomial regression model based on the training data set, meanwhile, the extracted features are comprehensive, and the model training by using a feature matrix formed by the feature vectors is favorable for improving the precision of the model.
3. The training data set is trained through a genetic algorithm and a cascade regression Laguerre polynomial fitting method to obtain regression coefficients, boundary parameters and a principal component mapping matrix generated by each iteration, and then a regression model is obtained.
In the invention, the technical schemes can be combined with each other to realize more preferable combination schemes. Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and drawings.
Drawings
The drawings are only for purposes of illustrating particular embodiments and are not to be construed as limiting the invention, wherein like reference numerals are used to designate like parts throughout.
FIG. 1 is a flowchart of a Cr12MoV material hardness prediction method in one embodiment;
FIG. 2 is a flow diagram of fitness calculation in one embodiment;
FIG. 3 is a flow chart of hardness prediction in one embodiment;
FIG. 4 is a drawing showing a hardness predicting apparatus for Cr12MoV material according to another embodiment;
reference numerals:
100-data acquisition module, 200-training module, 300-prediction module.
Detailed Description
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate preferred embodiments of the invention and together with the description, serve to explain the principles of the invention and not to limit the scope of the invention.
The extracted features of the existing nondestructive testing method are general or isolated from each other; or, the fitting of the multi-element nonlinear problem is realized by utilizing the combination form of the orthogonal polynomials, the storage and calculation expenses are consumed comparatively, and the condition of practical application is not met. In summary, the existing method for detecting the mechanical hardness of the material has the problems of poor precision and high complexity. The method comprises the steps of firstly obtaining Barkhausen noise data of a Cr12MoV material through a Barkhausen noise signal sensor, obtaining a hardness label value of the Cr12MoV material through external equipment of a professional evaluation mechanism, establishing a training data set through the Barkhausen noise data and the hardness label value corresponding to the Barkhausen noise data, then training the training data set through a cascade regression Laguerre polynomial fitting method to obtain a trained regression model, and finally conducting hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted through the trained regression model to obtain a prediction result. The method is simple and easy to implement, realizes the hardness prediction of the Cr12MoV material to be predicted, improves the prediction precision and has higher practical value.
In an embodiment of the present invention, a Cr12MoV material hardness prediction method is disclosed, as shown in fig. 1, including the following steps S1-S3.
And step S1, acquiring Barkhausen noise data of the Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set. Specifically, the barkhausen noise data of the Cr12MoV material is acquired by a barkhausen noise signal sensor, and the hardness label value of the Cr12MoV material is acquired by a professional evaluation mechanism external device. After the Barkhausen noise data and the corresponding hardness label value are obtained through corresponding equipment, a training data set can be established through the following steps:
s101, extracting characteristic information of Barkhausen noise data of the Cr12MoV material, wherein the characteristic information comprises a peak value, a peak value position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width. The Barkhausen noise data are single-cycle signals and comprise two peak values, the mean value of the two peak values is the peak value corresponding to the Barkhausen noise data, and the mean value of the positions of the two peak values is the peak value position. Meanwhile, average value calculation of the adjacent 11 sampling points is carried out at every 5 sampling points, the average values are used as envelope signal points of the Barkhausen noise signal, the corresponding envelope signal is generated after the Barkhausen noise is traversed, and the envelope average value is obtained by carrying out integral averaging on the signal.
And taking the second highest value on the left of the first peak value and the second highest value on the left of the second peak value of the Barkhausen noise data, averaging the two values to obtain a left secondary peak, and taking the second highest value on the right of the second peak value and the second highest value on the right of the second peak value of the Barkhausen noise data and averaging the two values to obtain a right secondary peak in the same way. And taking the zero crossing point value of the Barkhausen noise data as the tangential magnetic field intercept. The Barkhausen noise data is a single-cycle signal, the difference value of two 75% peak points of each peak value corresponding to a time point is taken as a 75% peak width, the two 75% peak widths are averaged to obtain the final 75% peak width, and similarly, the difference value of two half-peak points of each peak value corresponding to a time point is taken as a half-peak width, and the two half-peak widths are averaged to obtain the final half-peak width.
And S102, sequentially splicing the characteristic information obtained in the step S101 to obtain a characteristic vector.
S103, establishing a training data set based on the feature vector and the hardness label of the Barkhausen noise data. Specifically, a feature vector can be obtained by one Barkhausen noise data, a plurality of feature vectors form a feature matrix, and the feature matrix and the hardness label form a training data set, namely an input sample. Meanwhile, after the training data set is obtained, the training data set can be normalized, and the normalized value is used as the input of a laguerre polynomial regression model to train the model, specifically, the normalized calculation formula is as follows:
where X ' is input sample data after normalization, X is input sample data, min (X) is the minimum value of the input sample data, max (X) is the maximum value of the input sample data, Y ' is a hardness label after normalization, Y ' is a hardness label, min (Y) is the minimum value of the hardness label, max (Y) is the maximum value of the hardness label, and ∈ is 0.001.
The method comprises the steps of obtaining Barkhausen noise data of a Cr12MoV material by using a Barkhausen noise signal sensor, obtaining a hardness label value of the Cr12MoV material by using a special evaluation mechanism external device, carrying out feature extraction on the Barkhausen noise data to obtain a plurality of features, obtaining a feature vector of the Barkhausen noise data according to the features, and constructing a training data set by using the feature vector and the hardness label value of the Barkhausen noise data, so that technical support and basis are provided for later training of a Laguerre polynomial regression model based on the training data set, meanwhile, the extracted features are comprehensive, and the model training by using a feature matrix formed by the feature vectors is favorable for improving the precision of the model.
Step S2, training a training data set based on a cascade regression Laguerre polynomial fitting method to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters; and obtaining a trained regression model based on the regression coefficient, the principal component mapping matrix and the boundary parameter, comprising the following steps:
obtaining optimal Laguerre polynomial parameters based on a training data set and a genetic algorithm, comprising the following steps:
step S21, randomly generating a population member set for a given laguerre polynomial parameter based on a laguerre polynomial regression model.
And step S22, calculating the fitness of each population member in the population member set based on the training data set. Specifically, as shown in the fitness calculation flowchart shown in fig. 2, for each population member in the population member set, the normalized training data set is used to calculate the fitness value of the population member with the fitting error under the K-fold cross test as the objective function, where the calculation formula of the fitting error (fitness) RMSE is as follows:
wherein, SampleNum is the number of samples, y is the hardness value obtained by the regression model,is the hardness label value. The calculation of the fitness value of each population member comprises two sub-links, a training sub-link and a forecasting sub-link. The training data can be divided into training samples and testing samples by a K-time cross validation method, wherein the training samples are used for training the training sub-link, and the testing samples are used for testing the prediction sub-link.
Step S221, the training sub-link comprises the following steps:
calculating input training sample E in t-th circulation processtAnd a corresponding label FtSecond order correlation matrix ofAnd obtaining a principal component mapping vector p of the second order correlation matrixt(i.e., the eigenvector corresponding to the principal eigenvalue of the second order correlation matrix). The vector p is then mapped using the principal components of the second order correlation matrixtProjecting input samples into principal component space to form vectorsEtpt∈RSampleNum×1。
Aiming at the Laguerre polynomial parameter alpha of the current population member, utilizing a Laguerre polynomial three-term recurrence formula to obtain each order coefficient of a unitary Laguerre polynomial, wherein the formula is as follows:
wherein,(i takes 1, 2.. and n.) is the ith unary laguerre polynomial coefficient corresponding to the input sample data x.
Respectively aiming at each order coefficient of a unitary Laguerre polynomial of the input sample and the label, obtaining a regression coefficient, wherein the formula is as follows:
in the formula, EtRepresenting the input sample matrix during the t-th iteration, FtRepresenting an input sample label matrix in the t iteration process; et+1Residual matrix representing input samples, Ft+1Residual matrix representing hardness labels, ctRepresenting a residual matrix E for input samplest+1Of the regression coefficient matrix rtResidual matrix F representing the label for output stiffnesst+1The regression coefficient matrix of (2).
And (4) repeating the step of the training sub-link until the condition that the iteration time T is more than T (T is the total iteration time) is met, and obtaining a principal component mapping matrix.
Step S222, the predictor includes the following steps:
projecting the input test sample of each iteration to a principal component space E by using a principal component mapping matrix obtained in the process of training the subnodestpt(ii) a Calculating each order value of the unitary Laguerre polynomial corresponding to the test sample by using the Laguerre polynomial parameter alpha of the current population member; and using the regression coefficient r of the corresponding sub-regressor generated by the training sub-link aiming at the labeltWeighted summation is carried out on each order value of the unary Laguerre polynomial to generate regression output of the current cycle process, and the regression output generated by the current iteration is accumulated on a predicted value, namely
Wherein Y' represents a predicted value, and Y represents an initial predicted value; using regression coefficient c of corresponding sub-regressors generated for input samples in training processtWeighted summation is carried out on each order value of the unary Laguerre polynomial to generate regression output of the circulation process, and a regression residual error matrix E of input samples in the iteration process is obtainedt+1As input for the next iteration;
and (5) repeating the steps in the predictor link until the iteration condition T > T is met by letting T be T +1, and ending the predictor link.
And step S23, sorting and grouping the population member sets based on the fitness to obtain the optimal and worst population members of each subgroup. Specifically, the population member sets are sorted and grouped by fitness to divide the population member sets into M subgroups.
And step S24, performing iterative updating of the worst population members in the group based on the optimal and worst population members of each subgroup until an iterative condition is met, and obtaining optimal Laguerre polynomial parameters. The iterative updating of the worst population member specifically comprises the following steps:
according to the optimal population member alpha in the groupmbAnd worst population member alphamwWhen d is between the optimal population member alphambAnd worst population member alphamwIn between, i.e. alphamw≤d≤αmbThen, a random number D is obtainedm=rand(d:αmw≤d≤αmb) And updating the worst population member alphamnew=αmw+c*DmWherein c is a scaling factor, given by man. Calculating a fitness function RMSE of the newly generated worst population member based on the calculation formula of the fitness function in the step S22, and replacing the worst member in the current group if the fitness function value is less than the worst member; otherwise, generating a new member completely at random;
executing the iterative updating step of the worst population member in the group until all subgroups are updated, counting the fitness values of all population members in the current population, updating the global optimal solution, namely counting the fitness values of all samples in the current population after T iterations to obtain the optimal Laguerre polynomial parameter alphaGB。
And step S25, obtaining the regression coefficient and the principal component mapping matrix of each sub-regression in the whole cascade regression based on the optimal Laguerre polynomial parameters. In particular, the optimal Laguerre polynomial parameter alpha is obtainedGBThen, based on the parameters, calculating the coefficients of each order of the unary Laguerre polynomial of the training data set sample, and further obtaining the regression coefficient c of each sub-regression in the whole cascade regressiont,rtAnd a principal component mapping matrix p generated during each iterationtAnd simultaneously, outputting boundary parameters, wherein the boundary parameters comprise a maximum value max (X) of input sample data, a minimum value min (X) of the input sample data, a maximum value max (Y) of a hardness label and a minimum value min (Y) of the hardness label. Wherein, the regression coefficient ct,rtThe formula is as follows:
training the training data set by a genetic algorithm and a cascade regression Laguerre polynomial fitting method to obtain regression coefficients, boundary parameters and a principal component mapping matrix generated by each iteration, and further obtaining a regression model.
And step S3, acquiring Barkhausen noise data of the Cr12MoV material to be predicted, and performing hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result. The hardness prediction flowchart shown in fig. 3 includes the following steps:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material to be predicted, wherein the characteristic information comprises a peak value, a peak value position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width. And after the characteristic information of the Barkhausen noise data is extracted, sequentially splicing all the characteristic information to obtain a characteristic vector.
And carrying out normalization processing on the feature vectors to obtain the feature vectors of the Barkhausen noise data, wherein the normalization processing mode of the feature vectors is the same as the normalization mode of the feature vectors of the Barkhausen noise data in the training stage. After the normalized feature vectors are obtained, all the feature vectors form a feature matrix, namely a sample to be predicted, and meanwhile, the initialized predicted value is set to be zero.
And then, projecting the input sample of each iteration to a principal component space according to the principal component mapping matrix obtained in the training process, and calculating each order value of the unary Laguerre polynomial corresponding to the input sample according to the Laguerre polynomial parameters obtained by training. Weighting and summing the order values of the unitary Laguerre polynomial by using the regression coefficient of the corresponding regressor generated aiming at the hardness label in the training process to generate the regression output of the iteration process, and accumulating the regression output generated by the iteration to a predicted value; meanwhile, according to regression coefficients of corresponding regressors generated aiming at input samples, weighted summation is carried out on the projected samples, regression output of the iteration process is generated, and regression residual errors are obtained to be used as input of the next iteration;
repeating the prediction step to perform iterative update until the total iteration times is reached to obtain a predicted value, and performing inverse normalization processing on the generated predicted value to obtain a final prediction result, wherein the formula of the inverse normalization processing is as follows:
Y=Y′×[max(Y)-min(Y)]-min(Y)
wherein, Y is the final prediction result, Y' is the prediction value, max (Y) is the maximum value of the hardness label, and min (Y) is the minimum value of the hardness label.
And (4) predicting the hardness of the obtained Cr12MoV material to be predicted by using the regression model trained in the step S2, wherein the speed is high, the precision is high, and the result is more reliable.
Compared with the prior art, the method and the device for predicting the hardness of the Cr12MoV material provided by the embodiment are characterized in that the barkhausen noise data of the Cr12MoV material is obtained through a barkhausen noise signal sensor, the hardness label value of the Cr12MoV material is obtained through external equipment of a professional evaluation mechanism, a training data set is established through the barkhausen noise data and the hardness label value corresponding to the barkhausen noise data, then the training data set is trained through a cascade regression laguerre polynomial fitting method to obtain a trained regression model, and finally the trained regression model is used for predicting the hardness of the barkhausen noise data of the Cr12MoV material to be predicted, so that a prediction result is obtained. The method is simple and easy to implement, realizes the hardness prediction of the Cr12MoV material to be predicted, improves the prediction precision and has higher practical value.
In another embodiment of the present invention, a Cr12MoV material hardness predicting apparatus is disclosed, as shown in fig. 4, including: the data acquisition module 100 is used for acquiring Barkhausen noise data of a Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set; the training module 200 is used for training the training data set according to the cascade regression Laguerre polynomial fitting to obtain an optimal Laguerre polynomial parameter, a regression coefficient, a principal component mapping matrix and a boundary parameter, and obtaining a trained regression model based on the regression coefficient, the principal component mapping matrix and the boundary parameter; the prediction module 300 is configured to obtain the barkhausen noise data of the Cr12MoV material to be predicted, and perform hardness prediction on the barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result.
Since the implementation principle of the Cr12MoV material hardness prediction apparatus in the present application is similar to that of the Cr12MoV material hardness prediction method, further description is omitted here.
Those skilled in the art will appreciate that all or part of the flow of the method implementing the above embodiments may be implemented by a computer program, which is stored in a computer readable storage medium, to instruct related hardware. The computer readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention.
Claims (10)
1. The method for predicting the hardness of the Cr12MoV material is characterized by comprising the following steps of:
acquiring Barkhausen noise data of a Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set;
training the training data set based on a cascade regression Laguerre polynomial fitting method to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters; obtaining a trained regression model based on the regression coefficient, the principal component mapping matrix and the boundary parameter;
and acquiring Barkhausen noise data of the Cr12MoV material to be predicted, and performing hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result.
2. The Cr12MoV material hardness prediction method according to claim 1, wherein the Barkhausen noise data of the Cr12MoV material and the hardness label corresponding to the Barkhausen noise data are obtained to obtain a training data set, and the method comprises the following steps:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material, wherein the characteristic information comprises a peak value, a peak position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width;
splicing the characteristic information to obtain a characteristic vector of the Barkhausen noise data;
and establishing a training data set based on the eigenvectors and the hardness labels of the Barkhausen noise data.
3. The Cr12MoV material hardness prediction method according to claim 1, wherein the training dataset is trained based on a cascade regression Laguerre polynomial fit to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrices, and boundary parameters, comprising the steps of:
obtaining optimal laguerre polynomial parameters based on the training data set and a genetic algorithm;
and obtaining the regression coefficient, the principal component mapping matrix and the boundary parameter of each sub-regression in the whole cascade regression based on the optimal Laguerre polynomial parameter.
4. The Cr12MoV material hardness prediction method according to claim 3, wherein obtaining optimal Laguerre polynomial parameters based on the training data set and genetic algorithm comprises the following steps:
generating a population member set for a given laguerre polynomial parameter;
calculating the fitness of each population member in the population member set based on the training data set;
sorting and grouping the population member sets based on the fitness to obtain the optimal and worst population members of each subgroup;
and performing iterative updating of the worst population member in the group based on the optimal and worst population members of each subgroup until an iterative condition is met to obtain an optimal Laguerre polynomial parameter.
5. The Cr12MoV material hardness prediction method according to claim 3 or 4, wherein the hardness prediction is carried out on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result, and the method comprises the following steps:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material to be predicted, wherein the characteristic information comprises a peak value, a peak position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width;
splicing the characteristic information to obtain a characteristic vector of the Barkhausen noise data;
obtaining a sample to be predicted based on the feature vector of the Barkhausen noise data, and initializing a predicted value to be zero;
and inputting the sample to be predicted into a regression model for hardness prediction to obtain a prediction result.
6. The Cr12MoV material hardness prediction method according to claim 5, wherein the sample to be predicted is input into a regression model for hardness prediction to obtain a prediction result, and the method comprises the following steps:
projecting the input sample of each iteration to a principal component space based on the principal component mapping matrix output in the training process;
calculating the order values of the unary Laguerre polynomial corresponding to the input sample;
generating regression output of the iteration process by using the weighted summation of the regression coefficients of the corresponding regressors generated aiming at the hardness labels in the training process, and accumulating the regression output generated by the iteration to a predicted value;
carrying out weighted summation on the projected samples according to regression coefficients of corresponding regressors generated aiming at the input samples, generating regression output of the iteration process, and acquiring regression residual error as input of the next iteration;
repeating iterative updating until the total iteration times are reached to obtain a predicted value;
and carrying out inverse normalization processing on the generated final prediction value to obtain a final prediction result.
7. A Cr12MoV material hardness prediction device is characterized by comprising:
the data acquisition module is used for acquiring Barkhausen noise data of the Cr12MoV material and a hardness label corresponding to the Barkhausen noise data to obtain a training data set;
the training module is used for training the training data set according to cascade regression Laguerre polynomial fitting to obtain optimal Laguerre polynomial parameters, regression coefficients, principal component mapping matrixes and boundary parameters, and a trained regression model is obtained based on the regression coefficients, the principal component mapping matrixes and the boundary parameters;
and the prediction module is used for acquiring the Barkhausen noise data of the Cr12MoV material to be predicted, and performing hardness prediction on the Barkhausen noise data of the Cr12MoV material to be predicted based on the trained regression model to obtain a prediction result.
8. The Cr12MoV material hardness prediction device of claim 7, wherein the data acquisition module is configured to:
extracting characteristic information of Barkhausen noise data of the Cr12MoV material, wherein the characteristic information comprises a peak value, a peak position, an envelope mean value, a left secondary peak, a right secondary peak, a tangential magnetic field intercept, frequency spectrum information, 75% peak width and half peak width;
splicing the characteristic information to obtain a characteristic vector of the Barkhausen noise data;
and establishing a training data set based on the eigenvectors and the hardness labels of the Barkhausen noise data.
9. The Cr12MoV material hardness prediction device according to claim 8, wherein the training module is configured to:
obtaining optimal laguerre polynomial parameters based on the training data set and a genetic algorithm;
and obtaining the regression coefficient, the principal component mapping matrix and the boundary parameter of each sub-regression in the whole cascade regression based on the optimal Laguerre polynomial parameter.
10. The Cr12MoV material hardness prediction device according to claim 9, wherein obtaining optimal laguerre polynomial parameters based on the training data set and genetic algorithm comprises the steps of:
generating a population member set for a given laguerre polynomial parameter;
calculating the fitness of each population member in the population member set based on the training data set;
sorting and grouping the population member sets based on the fitness to obtain the optimal and worst population members of each subgroup;
and performing iterative updating of the worst population member in the group based on the optimal and worst population members of each subgroup until an iterative condition is met to obtain an optimal Laguerre polynomial parameter.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011357828.4A CN112382354B (en) | 2020-11-27 | 2020-11-27 | Cr12MoV material hardness prediction method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011357828.4A CN112382354B (en) | 2020-11-27 | 2020-11-27 | Cr12MoV material hardness prediction method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112382354A true CN112382354A (en) | 2021-02-19 |
CN112382354B CN112382354B (en) | 2021-08-10 |
Family
ID=74587642
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011357828.4A Active CN112382354B (en) | 2020-11-27 | 2020-11-27 | Cr12MoV material hardness prediction method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112382354B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112906991A (en) * | 2021-04-07 | 2021-06-04 | 中国科学院计算机网络信息中心 | Method and device for predicting RAFM steel radiation hardening behavior |
CN112990009A (en) * | 2021-03-12 | 2021-06-18 | 平安科技(深圳)有限公司 | End-to-end-based lane line detection method, device, equipment and storage medium |
WO2023233047A1 (en) * | 2022-06-02 | 2023-12-07 | Asociacion Centro Tecnologico Ceit | Method and system for measuring longitudinal distance to the hardened area of the surface of a part |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103617354A (en) * | 2013-11-22 | 2014-03-05 | 冶金自动化研究设计院 | Universal fitting calculation method of steel hardenability |
CN105033764A (en) * | 2015-06-26 | 2015-11-11 | 哈尔滨理工大学 | Method for detecting quenched steel die milling stability |
CN105718655A (en) * | 2016-01-21 | 2016-06-29 | 天津大学 | Closed loop neural stimulation simulation system for Parkinson state |
US20160210561A1 (en) * | 2007-01-04 | 2016-07-21 | Dan Steinberg | Methods and systems for automatic selection of preferred size classification and regression trees |
CN107340162A (en) * | 2017-07-10 | 2017-11-10 | 北京工业大学 | The preparation method of calibration sample during a kind of Barkhausen Noise detection grinding burn |
CN107435124A (en) * | 2016-05-27 | 2017-12-05 | 鲍凡凡 | High tungsten ledeburite pinion steel of hard high-wearing feature and preparation method thereof |
CN108595784A (en) * | 2018-03-30 | 2018-09-28 | 南京航空航天大学 | Based on MBN signals full width at half maximum than the yield strength method of estimation with envelope size |
CN108846173A (en) * | 2018-05-29 | 2018-11-20 | 南京航空航天大学 | A kind of barkhausen signal stress estimation method based on slow signature analysis |
CN108872359A (en) * | 2018-06-21 | 2018-11-23 | 北京工业大学 | A kind of magnetic mixing non-linear detection method for ferrimagnet hardness characterization |
CN108959175A (en) * | 2018-05-30 | 2018-12-07 | 南京航空航天大学 | A kind of ferrimagnet yield strength successive Regression estimation method based on MBN |
CN109142057A (en) * | 2018-09-11 | 2019-01-04 | 北京工业大学 | The Barkhausen noise non-destructive measuring method and system of ferrimagnet local stress and strain field |
CN109409271A (en) * | 2018-10-16 | 2019-03-01 | 北京工业大学 | Testing of Ferromagnetic Material Hardness prediction algorithm based on BP neural network innovatory algorithm |
WO2019236940A2 (en) * | 2018-06-08 | 2019-12-12 | Covestro Llc | System, method, and computer program product for predicting properties of a polymer |
CN111049718A (en) * | 2019-12-30 | 2020-04-21 | 北京京航计算通讯研究所 | CAN bus simulation monitoring method |
US20200253505A1 (en) * | 2014-11-13 | 2020-08-13 | Ori Diagnostic Instruments, LLC | Systems and methods for high frequency impedance spectroscopy detection of daily changes of dielectric properties of the human body to measure body composition and hydration status |
-
2020
- 2020-11-27 CN CN202011357828.4A patent/CN112382354B/en active Active
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160210561A1 (en) * | 2007-01-04 | 2016-07-21 | Dan Steinberg | Methods and systems for automatic selection of preferred size classification and regression trees |
CN103617354A (en) * | 2013-11-22 | 2014-03-05 | 冶金自动化研究设计院 | Universal fitting calculation method of steel hardenability |
US20200253505A1 (en) * | 2014-11-13 | 2020-08-13 | Ori Diagnostic Instruments, LLC | Systems and methods for high frequency impedance spectroscopy detection of daily changes of dielectric properties of the human body to measure body composition and hydration status |
CN105033764A (en) * | 2015-06-26 | 2015-11-11 | 哈尔滨理工大学 | Method for detecting quenched steel die milling stability |
CN105718655A (en) * | 2016-01-21 | 2016-06-29 | 天津大学 | Closed loop neural stimulation simulation system for Parkinson state |
CN107435124A (en) * | 2016-05-27 | 2017-12-05 | 鲍凡凡 | High tungsten ledeburite pinion steel of hard high-wearing feature and preparation method thereof |
CN107340162A (en) * | 2017-07-10 | 2017-11-10 | 北京工业大学 | The preparation method of calibration sample during a kind of Barkhausen Noise detection grinding burn |
CN108595784A (en) * | 2018-03-30 | 2018-09-28 | 南京航空航天大学 | Based on MBN signals full width at half maximum than the yield strength method of estimation with envelope size |
CN108846173A (en) * | 2018-05-29 | 2018-11-20 | 南京航空航天大学 | A kind of barkhausen signal stress estimation method based on slow signature analysis |
CN108959175A (en) * | 2018-05-30 | 2018-12-07 | 南京航空航天大学 | A kind of ferrimagnet yield strength successive Regression estimation method based on MBN |
WO2019236940A2 (en) * | 2018-06-08 | 2019-12-12 | Covestro Llc | System, method, and computer program product for predicting properties of a polymer |
CN108872359A (en) * | 2018-06-21 | 2018-11-23 | 北京工业大学 | A kind of magnetic mixing non-linear detection method for ferrimagnet hardness characterization |
CN109142057A (en) * | 2018-09-11 | 2019-01-04 | 北京工业大学 | The Barkhausen noise non-destructive measuring method and system of ferrimagnet local stress and strain field |
CN109409271A (en) * | 2018-10-16 | 2019-03-01 | 北京工业大学 | Testing of Ferromagnetic Material Hardness prediction algorithm based on BP neural network innovatory algorithm |
CN111049718A (en) * | 2019-12-30 | 2020-04-21 | 北京京航计算通讯研究所 | CAN bus simulation monitoring method |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112990009A (en) * | 2021-03-12 | 2021-06-18 | 平安科技(深圳)有限公司 | End-to-end-based lane line detection method, device, equipment and storage medium |
CN112906991A (en) * | 2021-04-07 | 2021-06-04 | 中国科学院计算机网络信息中心 | Method and device for predicting RAFM steel radiation hardening behavior |
WO2023233047A1 (en) * | 2022-06-02 | 2023-12-07 | Asociacion Centro Tecnologico Ceit | Method and system for measuring longitudinal distance to the hardened area of the surface of a part |
Also Published As
Publication number | Publication date |
---|---|
CN112382354B (en) | 2021-08-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112382354B (en) | Cr12MoV material hardness prediction method and device | |
CN112784881B (en) | Network abnormal flow detection method, model and system | |
CN105354595A (en) | Robust visual image classification method and system | |
CN114239725B (en) | Electric larceny detection method for data poisoning attack | |
CN111046961B (en) | Fault classification method based on bidirectional long-time and short-time memory unit and capsule network | |
Tang et al. | Modeling collinear data using double-layer GA-based selective ensemble kernel partial least squares algorithm | |
CN111582315B (en) | Sample data processing method and device and electronic equipment | |
CN117237733A (en) | Breast cancer full-slice image classification method combining self-supervision and weak supervision learning | |
CN116451139B (en) | Live broadcast data rapid analysis method based on artificial intelligence | |
CN116306226B (en) | Fuel cell performance degradation prediction method | |
CN118151020B (en) | Method and system for detecting safety performance of battery | |
CN115659254A (en) | Power quality disturbance analysis method for power distribution network with bimodal feature fusion | |
CN114897085A (en) | Clustering method based on closed subgraph link prediction and computer equipment | |
CN117578438A (en) | Generating countermeasure network method and system for predicting new energy power generation | |
CN117594132A (en) | Single-cell RNA sequence data clustering method based on robust residual error map convolutional network | |
CN116245259B (en) | Photovoltaic power generation prediction method and device based on depth feature selection and electronic equipment | |
JP4495691B2 (en) | Influence factor estimation apparatus and influence factor estimation program | |
CN115510964B (en) | Computer calculation method for liquid chromatograph scientific instrument | |
CN116430234A (en) | Fuel cell fault diagnosis model construction method | |
CN116561569A (en) | Industrial power load identification method based on EO feature selection and AdaBoost algorithm | |
CN116185817A (en) | Screening method and system for software defect prediction rules | |
CN115062685A (en) | Fault diagnosis method and device, electronic equipment and storage medium | |
Moslemi et al. | Scaling up bayesian neural networks with neural networks | |
CN115599698B (en) | Software defect prediction method and system based on class association rule | |
CN117934963B (en) | Gas sensor drift compensation 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 |