CN109033586B

CN109033586B - Method and system for determining alloy grain size based on mapping monotonicity

Info

Publication number: CN109033586B
Application number: CN201810768923.XA
Authority: CN
Inventors: 陈昊; 董金龙; 黎明; 邬冠华; 陈曦; 张聪炫; 李军华
Original assignee: Nanchang Hangkong University
Current assignee: Nanchang Hangkong University
Priority date: 2018-07-13
Filing date: 2018-07-13
Publication date: 2022-08-12
Anticipated expiration: 2038-07-13
Also published as: CN109033586A

Abstract

The invention discloses a method and a system for determining alloy grain size based on mapping monotonicity. The invention adopts the correlation measurement criterion to gradually screen the effective ultrasonic detection parameters, then establishes a grain size soft measurement model according to the finally screened effective ultrasonic detection parameter set by taking monotonicity as the maximum optimization target, so that the effective ultrasonic detection parameters can not only keep a monotone increasing or monotone decreasing form under the condition of ordered arrangement along with the grain size, but also realize the effectiveness of the test outside the sample set. Therefore, the grain size soft measurement model established by the determination method and the determination system provided by the invention has the advantages of high measurement precision, good monotonicity, small error, visual and complete result diagram of the model and stable evaluation effect.

Description

Method and system for determining alloy grain size based on mapping monotonicity

Technical Field

The invention relates to the field of ultrasonic detection technology and alloy grain size representation thereof, in particular to a method and a system for determining alloy grain size based on mapping monotonicity.

Background

The titanium alloy has elasticity resistance, corrosion resistance, fatigue resistance and good heat resistance, is widely used in national defense equipment such as advanced airplanes, spacecrafts, high thrust-weight ratio aeroengines, ships and the like, serves as a part of key parts such as aeroengine fans, gas compressor wheel discs, blades and the like, and is known as space metal. The primary alpha-phase grain size of the titanium alloy has certain influence on the yield strength, the fatigue performance and the corrosion resistance, and each macroscopic performance of the titanium alloy forms corresponding characteristic response along with different changes of the grain size. Because the specificity and importance of the application of the titanium alloy are irreplaceable, it is crucial to design an effective method for characterizing the primary alpha-phase grain size of the titanium alloy.

The existing grain size detection methods can be divided into two categories, namely destructive and nondestructive detection. The destructive detection mainly comprises methods such as metallographic detection, Electron Back Scattering Diffraction (EBSD) detection and the like. Nondestructive evaluation includes ultrasonic detection, eddy current detection and the like. Although the destructive method has high detection precision, the detection process is complicated, the detection efficiency is low, and irreversible damage can be caused to the test piece. Compared with the prior art, the nondestructive testing method can ensure higher testing efficiency under the condition of not damaging the tested workpiece, so that the establishment of the nondestructive evaluation method for representing the grain size of the material is a key problem of the current research. The ultrasonic nondestructive evaluation method has the advantages of high penetration capacity, high flaw detection sensitivity, easiness in automatic inspection and the like, and is most commonly focused in nondestructive characterization of grain sizes of high-temperature alloys and titanium alloys at present.

The grain size has different degrees of influence on linear ultrasonic detection parameters such as sound velocity, attenuation coefficient and the like. With the increasing requirements on the specificity and sensitivity of the microstructure response of materials, more and more researchers focus on the characterization relationship between the ultrasonic nonlinear parameters and the microscopic grain sizes. However, the linear relationship established by only a single ultrasonic detection parameter and the grain size results in a lack of accuracy of the evaluation model due to differences in information about the grain size carried by characteristic parameters such as sound velocity, attenuation coefficient, and nonlinear coefficient. For the existing multi-parameter evaluation scheme, due to the increase of complex information of the microscopic grain size of the detected material, the grain size ultrasonic evaluation method constructed by taking the minimum error as the target cannot effectively represent the grain size of the complex alloy material, and the formed incomplete monotonicity curve causes the evaluation accuracy to be lower and even loses the evaluation effect.

Therefore, how to establish an effective nondestructive testing method for accurately determining the grain size of the complex alloy material becomes a technical problem to be solved by those skilled in the art.

Disclosure of Invention

The invention aims to provide a method and a system for determining the size of alloy crystal grains based on mapping monotonicity.

In order to achieve the purpose, the invention provides the following scheme:

a method of determining alloy grain size based on mapping monotonicity, the method comprising:

acquiring ultrasonic fixed-point scanning signals, average thickness values and grain size values of all experimental samples;

determining each ultrasonic detection parameter value of each experimental sample according to the average thickness value and the ultrasonic fixed-point scanning signal;

acquiring an interval step length, a lowest threshold and a current selection moment;

determining a selection interval corresponding to the selection moment according to the interval step length and the selection moment;

determining effective ultrasonic detection parameters at each selection moment by adopting a correlation measurement criterion according to the minimum threshold, the selection interval, each ultrasonic detection parameter value and the grain size value;

determining a final effective ultrasonic detection parameter set according to the effective ultrasonic detection parameters at each selected moment;

establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set by taking the monotonicity maximum as an optimization target;

and determining the grain size of the tested alloy by adopting the grain size soft measurement model.

Optionally, the determining, according to the interval step length and the selection time, a selection interval corresponding to the selection time specifically includes:

according toThe formula:

determining a selection interval corresponding to the selection time, wherein t represents the selection time,

denotes the interval step size, θ _t Indicates a selection interval corresponding to the selection time t.

Optionally, the determining, according to the minimum threshold, the selection interval, the ultrasonic detection parameter values, and the grain size value, the effective ultrasonic detection parameter at each selection time by using a correlation metric criterion specifically includes:

judging whether the maximum value of the selection interval is greater than or equal to the minimum threshold value or not, and obtaining a first judgment result;

when the first judgment result shows that the maximum value of the selection interval is larger than or equal to the minimum threshold value, respectively calculating the correlation coefficient between the grain size value of each experimental sample and each ultrasonic detection parameter value of each experimental sample by adopting a Pearson correlation coefficient analysis method to obtain each size-parameter correlation coefficient;

selecting the ultrasonic detection parameters corresponding to the ultrasonic detection parameter values of the size-parameter correlation coefficient in the selection interval as the primarily selected ultrasonic detection parameters;

respectively calculating the average correlation coefficient of each initially selected ultrasonic detection parameter of each experimental sample by adopting a Pearson correlation coefficient analysis method;

selecting the initial ultrasonic detection parameters with the average correlation coefficient smaller than the minimum value of the selection interval and the initial ultrasonic detection parameters with the average correlation coefficient larger than the minimum value of the selection interval and the maximum average correlation coefficient as effective ultrasonic detection parameters;

and updating the selection time, and returning to the step of determining the selection interval according to the interval step length and the selection time.

Optionally, establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set with monotonicity being the maximum optimization target specifically includes:

constructing a multi-dimensional effective parameter vector according to each effective ultrasonic detection parameter;

constructing a dimension reduction mapping function, and reducing the multi-dimensional effective parameter vector into a single-dimensional effective parameter by adopting the dimension reduction mapping function;

carrying out normalization processing on the single-dimensional effective parameters to obtain normalized single-dimensional effective parameters;

constructing a first fitting function, wherein a dependent variable of the first fitting function is the grain size, and an independent variable of the first fitting function is a normalized single-dimensional effective parameter;

carrying out inverse transformation on the first fitting function to obtain a second fitting function, wherein a dependent variable of the second fitting function is a normalized single-dimensional effective parameter, and an independent variable of the second fitting function is a grain size;

constructing an optimization function by taking the maximum number of the difference values of the dependent variables corresponding to the adjacent independent variables of the second fitting function as positive numbers or negative numbers;

solving the optimization function by adopting a self-adaptive differential evolution algorithm to obtain an optimal dimensionality reduction coefficient and an optimal fitting coefficient which enable the number of dependent variables corresponding to adjacent independent variables of the second fitting function to be the largest and enable the difference values to be both positive numbers or both negative numbers, wherein the optimal dimensionality reduction coefficient is the optimal coefficient of the dimensionality reduction mapping function, and the optimal fitting coefficient is the optimal coefficient of the first fitting function;

and substituting the optimal fitting coefficient into the first fitting function to obtain a grain size soft measurement model.

A system for determining alloy grain size based on mapping monotonicity, the system comprising:

the sample parameter acquisition module is used for acquiring ultrasonic fixed point scanning signals, average thickness values and grain size values of all experimental samples;

an ultrasonic detection parameter value determining module, configured to determine, according to the average thickness value and the ultrasonic fixed-point scanning signal, each ultrasonic detection parameter value of each experimental sample;

the selection parameter acquisition module is used for acquiring the interval step length, the lowest threshold and the current selection moment;

a selection interval determining module, configured to determine a selection interval corresponding to the selection time according to the interval step length and the selection time;

an effective ultrasonic detection parameter determining module, configured to determine, according to the minimum threshold, the selection interval, the ultrasonic detection parameter values, and the grain size value, an effective ultrasonic detection parameter at each selection time by using a correlation metric criterion;

a final effective ultrasonic detection parameter set determining module, configured to determine a final effective ultrasonic detection parameter set according to the effective ultrasonic detection parameters at each selected time;

the grain size soft measurement model establishing module is used for establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set by taking monotonicity as a maximum optimization target;

and the grain size measuring module is used for determining the grain size of the measured alloy by adopting the grain size soft measurement model.

Optionally, the selection interval determining module is configured to:

Optionally, the effective ultrasonic detection parameter determining module specifically includes:

the first judgment unit is used for judging whether the maximum value of the selection interval is greater than or equal to the lowest threshold value or not and obtaining a first judgment result;

a size-parameter correlation coefficient determining unit, configured to, when the first determination result indicates that the maximum value of the selection interval is greater than or equal to a minimum threshold value, respectively calculate, by using a pearson correlation coefficient analysis method, a correlation coefficient between a grain size value of each experimental sample and each ultrasonic detection parameter value of each experimental sample, and obtain each size-parameter correlation coefficient;

the primary selection ultrasonic detection parameter screening unit is used for selecting the ultrasonic detection parameters corresponding to the ultrasonic detection parameter values of the size-parameter correlation coefficient in the selection interval as the primary selection ultrasonic detection parameters;

the average correlation coefficient calculating unit is used for calculating the average correlation coefficient of each primarily selected ultrasonic detection parameter of each experimental sample by adopting a Pearson correlation coefficient analysis method;

an effective ultrasonic detection parameter selection unit, configured to select, as effective ultrasonic detection parameters, the initial ultrasonic detection parameters whose average correlation coefficients are smaller than the minimum value of the selection interval and the initial ultrasonic detection parameters whose average correlation coefficients are larger than the minimum value of the selection interval and whose average correlation coefficients are the largest;

and the selection time updating unit is used for updating the selection time.

Optionally, the grain size soft measurement model establishing module specifically includes:

the multi-dimensional effective parameter vector construction unit is used for constructing a multi-dimensional effective parameter vector according to each effective ultrasonic detection parameter;

the dimension reduction unit is used for constructing a dimension reduction mapping function and reducing the multi-dimensional effective parameter vector into a single-dimensional effective parameter by adopting the dimension reduction mapping function;

the normalization processing unit is used for performing normalization processing on the single-dimensional effective parameters to obtain normalized single-dimensional effective parameters;

the fitting function constructing unit is used for constructing a first fitting function, the dependent variable of the first fitting function is the grain size, and the independent variable of the first fitting function is a normalized single-dimensional effective parameter;

the inverse transformation unit is used for carrying out inverse transformation on the first fitting function to obtain a second fitting function, a dependent variable of the second fitting function is a normalized single-dimensional effective parameter, and an independent variable of the second fitting function is a grain size;

the optimization function constructing unit is used for constructing an optimization function by taking the maximum number of the dependent variables corresponding to the adjacent independent variables of the second fitting function as a target, wherein the difference values of the dependent variables are both positive numbers or both negative numbers;

the adaptive differential evolution algorithm solving unit is used for solving the optimization function by adopting an adaptive differential evolution algorithm to obtain an optimal dimension reduction coefficient and an optimal fitting coefficient, wherein the optimal dimension reduction coefficient is the optimal coefficient of the dimension reduction mapping function, and the optimal fitting coefficient is the optimal coefficient of the first fitting function, and the difference values of dependent variables corresponding to adjacent independent variables of the second fitting function are both positive numbers or negative numbers with the largest number;

and the soft measurement model determining unit is used for substituting the optimal fitting coefficient into the first fitting function to obtain a grain size soft measurement model.

According to the specific embodiment provided by the invention, the invention discloses the following technical effects:

according to the method and the system for determining the alloy grain size based on the mapping monotonicity, provided by the invention, effective ultrasonic detection parameters are gradually screened by adopting a correlation measurement criterion, then the monotonicity is the maximum optimization target, and a grain size soft measurement model is established according to the finally screened effective ultrasonic detection parameter set, so that under the condition of ordered arrangement of the grain sizes, the effective ultrasonic detection parameters can not only keep a monotone increasing or monotone decreasing form, but also realize the effectiveness of the out-of-set test of a sample. Therefore, the grain size soft measurement model established by the determination method and the determination system provided by the invention has the advantages of high measurement precision, good monotonicity, small error, visual and complete result diagram of the model and stable evaluation effect.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.

FIG. 1 is a flowchart of a method for determining alloy grain size based on mapping monotonicity according to example 1 of the present invention;

FIG. 2 is a block diagram of a system for determining alloy grain size based on mapping monotonicity, provided in example 2 of the present invention;

FIG. 3 is a typical microstructure morphology map of TC4 titanium alloy at different forging temperatures and different forging deformation according to example 3 of the present invention;

FIG. 4 is a graph illustrating the effectiveness of the average grain size provided in example 3 of the present invention;

FIG. 5 is a diagram illustrating the calculation monotonicity and non-monotonicity provided in embodiment 3 of the present invention;

FIG. 6 is a graph of 5 models and fitting relations for evaluating the average grain size according to example 3 of the present invention;

fig. 7 is a model and a fitting relationship curve of 5 standard deviations of the evaluated grain sizes provided in example 3 of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

Example 1:

fig. 1 is a flowchart of a method for determining alloy grain size based on mapping monotonicity according to embodiment 1 of the present invention. As shown in fig. 1, a method for determining the grain size of an alloy based on mapping monotonicity, the method comprising:

step 101: acquiring ultrasonic fixed-point scanning signals, average thickness values and grain size values of all experimental samples;

step 102: determining each ultrasonic detection parameter value of each experimental sample according to the average thickness value and the ultrasonic fixed-point scanning signal;

step 103: acquiring an interval step length, a lowest threshold and a current selection moment;

step 104: according to the interval step length and the selection time, adopting a formula:

denotes the interval step size, θ _t A selection interval corresponding to the selection time t is shown;

step 105: determining effective ultrasonic detection parameters at each selection moment by adopting a correlation measurement criterion according to the minimum threshold, the selection interval, each ultrasonic detection parameter value and the grain size value;

step 106: determining a final effective ultrasonic detection parameter set according to the effective ultrasonic detection parameters at each selected moment;

step 107: establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set by taking the monotonicity maximum as an optimization target;

step 108: and determining the grain size of the tested alloy by adopting the grain size soft measurement model.

Specifically, the step 105: determining effective ultrasonic detection parameters at each selection moment by adopting a correlation measurement criterion according to the minimum threshold, the selection interval, the ultrasonic detection parameter values and the grain size value, and specifically comprising the following steps:

updating the selection time, and returning to the step 104: determining a selection interval according to the interval step length and the selection moment;

when the first judgment result indicates that the maximum value of the selection interval is smaller than the lowest threshold, step 106 is executed: and determining a final effective ultrasonic detection parameter set according to the effective ultrasonic detection parameters at each selected moment.

Specifically, the step 107: establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set by taking the monotonicity maximum optimization target, and specifically comprising the following steps of:

Example 2:

FIG. 2 is a block diagram of a system for determining alloy grain size based on mapping monotonicity, provided in example 2 of the present invention. As shown in fig. 2, a system for determining the grain size of an alloy, the system comprising:

a sample parameter obtaining module 201, configured to obtain an ultrasonic fixed-point scanning signal, an average thickness value, and a grain size value of each experimental sample;

an ultrasonic detection parameter value determining module 202, configured to determine, according to the average thickness value and the ultrasonic fixed-point scanning signal, each ultrasonic detection parameter value of each experimental sample;

a selection parameter obtaining module 203, configured to obtain an interval step length, a lowest threshold, and a current selection time;

a selection interval determination module 204 configured to:

an effective ultrasonic detection parameter determining module 205, configured to determine, according to the minimum threshold, the selection interval, the ultrasonic detection parameter values, and the grain size value, an effective ultrasonic detection parameter at each selection time by using a correlation metric criterion;

a final effective ultrasound detection parameter set determining module 206, configured to determine a final effective ultrasound detection parameter set according to the effective ultrasound detection parameters at each selected time;

a grain size soft measurement model establishing module 207, configured to establish a grain size soft measurement model according to the final effective ultrasonic detection parameter set with monotonicity being the maximum optimization target;

and a grain size measurement module 208 for determining the grain size of the measured alloy using the grain size soft measurement model.

Specifically, the effective ultrasound detection parameter determining module 205 specifically includes:

and the selection time updating unit is used for updating the selection time.

Specifically, the grain size soft measurement model establishing module 207 specifically includes:

the adaptive differential evolution algorithm solving unit is used for solving the optimization function by adopting an adaptive differential evolution algorithm to obtain an optimal dimensionality reduction coefficient and an optimal fitting coefficient, wherein the optimal dimensionality reduction coefficient is the optimal coefficient of the dimensionality reduction mapping function, and the optimal fitting coefficient is the optimal coefficient of the first fitting function, and the difference values of dependent variables corresponding to adjacent independent variables of the second fitting function are both positive numbers or both negative numbers;

Example 3:

in this embodiment, the influence of monotonicity is taken as a strategy, a plurality of ultrasonic detection parameters are preferably reduced to be one-dimensional parameters and normalized, the maximum monotonicity (the sequential sequence difference of sample points reduced to be one-dimensional ultrasonic detection parameters is positive or negative) is fitted with the primary alpha-phase grain size by one time, an optimization target is determined by combining with a self-adaptive differential evolution (SADE) algorithm to determine an optimization problem, and the undetermined coefficients of a corresponding mapping function and the fitting function are obtained by solving, so that the TC4 primary alpha-phase grain size ultrasonic evaluation method for mapping monotonicity is established.

The method comprises the following steps: and carrying out experiments to obtain ultrasonic fixed point scanning signals, average thickness values and grain size values of all experimental samples, and determining all ultrasonic detection parameter values of all experimental samples according to the average thickness values and the ultrasonic fixed point scanning signals.

The samples to be detected are prepared by different forging temperatures (920-990 ℃) and different forging deformation amounts (23-42%). Carrying out an ultrasonic detection experiment, respectively adopting a pulse reflection method (5077PR pulse generator, frequency is set to be 10MHz) and a collinear harmonic method (RAM-5000-SNAP nonlinear ultrasonic testing system, emission frequency is 2.5MHz and receiving frequency is 5MHz) to detect a sample, extracting a linear ultrasonic original A scanning signal (ultrasonic fixed-point scanning signal), storing information by utilizing EVA data processing software, and calculating sound velocity and attenuation coefficient according to the following formula:

in the formula, C _L Representing the speed of sound, at is the transition time between the peak of the surface echo and the peak of the primary bottom echo,

is the average thickness of the specimen, P _SE Is the peak of the surface echo, P _F1 Alpha represents the attenuation coefficient for the primary bottom echo peak.

The calculation formula of the nonlinear coefficient beta extracted in the nonlinear ultrasonic detection mode is as follows:

wherein x is the propagation distance, k is the wave number, A ₀ Is the fundamental amplitude, A ₂ The second harmonic amplitude. X and k are generally approximated as constant values, with beta being replaced by a relatively non-linear coefficient, beta', i.e.

The extracted ultrasonic detection parameters adopt a statistical mode of averaging and standard deviation and correspondingly calculate each sample, and the extracted characteristic parameters are respectively as follows: speed of sound, including average of speed of sound

Standard deviation of harmony velocity

Attenuation coefficient, including average value of attenuation coefficient

And standard deviation of attenuation coefficient

Primary bottom echo frequency peak P _F1 And the frequency peak P of the secondary bottom wave _F2 Frequency offset of primary bottom wave

And secondary bottom wave frequency offset

Relative non-linear coefficients, including average values of relative non-linear coefficients

And standard deviation of relative nonlinear coefficient

The corresponding ultrasonic testing parameters of this example are shown in table 1.

Ultrasonic testing parameter table extracted from table 1

The metallographic preparation is carried out on the sample, and 4 steps of sample cutting, grinding, polishing and corroding are followed. The metallographic abrasive paper is KmTBCr15Mo water-milled abrasive paper, the prepared corrosive liquid is HF, HNO3, H2O, 3, 8 and 89, the surface of the metallographic abrasive paper is corroded, and the sample structure is observed under an optical microscope. Typical microstructure morphologies of the obtained TC4 titanium alloy at different forging temperatures and different forging deformation amounts are shown in FIG. 3, in which part (a) of FIG. 3 shows typical microstructure morphologies of the TC4 titanium alloy at a forging temperature of 920 ℃ and a forging deformation amount of 23%, part (b) of FIG. 3 shows typical microstructure morphologies of the TC4 titanium alloy at a forging temperature of 920 ℃ and a forging deformation amount of 38%, part (c) of FIG. 3 shows typical microstructure morphologies of the TC4 titanium alloy at a forging temperature of 940 ℃ and a forging deformation amount of 26%Typical microstructure morphology of gold, portion (d) of fig. 3 shows typical microstructure morphology of TC4 titanium alloy at a forging temperature of 940 ℃, a forging deformation of 40%, portion (e) of fig. 3 shows typical microstructure morphology of TC4 titanium alloy at a forging temperature of 990 ℃, a forging deformation of 26%, portion (f) of fig. 3 shows typical microstructure morphology of TC4 titanium alloy at a forging deformation of 42%. The gold phase diagram of the TC4 titanium alloy sample was nascent using ImageJ software _α Analyzing the proportion of the phase white area, measuring the area S, and combining the calculation formula

The grain size (average value) of the equivalent primary alpha phase is obtained through statistics

Standard deviation of

) In the present example, the primary alpha phase grain size of the TC4 titanium alloy is expressed as a grain size. The relevant process parameters obtained by metallographic examination are shown in Table 2.

TABLE 2 primary alpha-phase grain size of TC4 at different forging temperatures and forging deformation

Step two: obtaining an interval step length, a lowest threshold and a current selection moment, determining a selection interval corresponding to the selection moment according to the interval step length and the selection moment, and determining effective ultrasonic detection parameters of each selection moment by adopting a correlation measurement criterion according to the lowest threshold, the selection interval, the ultrasonic detection parameter values and the grain size value.

Expressing the extracted ultrasonic detection parameters as Y ═ { Y ═ in the form of variables ₁ ,Y ₂ ,…,Y _k The mean and standard deviation of primary alpha-phase grain sizes are expressed as

Wherein the content of the first and second substances,

represents the average of primary alpha phase grain sizes,

the standard deviation of the primary alpha phase grain size is indicated. Because the direct modeling of the multidimensional ultrasonic detection parameters and the grain sizes shows information redundancy and uncertainty influence, a strategy is needed to remove invalid and interference parameters from the multidimensional parameters. In this embodiment, correlation measurement is performed by using Pearson correlation coefficient analysis, which has the following formula

In the formula, ρ _XY The correlation coefficient is shown. X represents the grain size, Y represents the ultrasonic testing parameters,

represents the average of the grain size within the sample,

represents the average value of the ultrasonic testing parameter within the sample.

The correlation analysis form of the grain size X and the ultrasonic detection parameter set Y is recorded as

The correlation coefficient of the grain size X and the ith ultrasonic detection parameter is shown.

The ultrasonic detection parameter internal correlation analysis form is as follows:

in the form of matrixρ _YY Representing the correlation between the ultrasound examination parameters,

indicating the ith ultrasonic testing parameter Y _i And j ultrasonic detection parameter Y _j The correlation of (c).

In the formula (I), the compound is shown in the specification,

indicating the ith ultrasonic testing parameter Y _i And (4) summing the absolute values of the correlation of the rest parameters and calculating an average correlation coefficient, wherein k represents the total number of all the ultrasonic detection parameters.

In this embodiment, when analyzing the ultrasonic testing parameters according to the correlation, the selection interval is set

2 correlation measurement criteria are formulated to assist in screening out effective ultrasonic detection parameters, and the criteria are as follows

Criterion 1: correlation rho of ultrasonic detection parameter Y and grain size X _XY Selection interval theta at time t ^t And (4) inside. The formula is as follows:

wherein t represents the current set time, k is the total number of all ultrasonic detection parameters,

the ultrasonic detection parameter set selected according to the criterion 1 is Y for the interval range of the current moment _t ＝{Y _t1 ,Y _t2 ,…,Y _tp }，Y _t The correlation of the ultrasonic detection parameters and the grain sizes belongs to characteristic parameters in a selected interval.

Criterion 2: selecting average correlation coefficient between parameters

Is lower than

The calculation formula of the ultrasonic detection parameters is as follows:

in the formula (I), the compound is shown in the specification,

for the initially selected ultrasonic detection parameters whose average correlation coefficient is less than the minimum value of the selection interval, i.e. from Y for criterion 2 _t The 1 st part of the selected effective ultrasonic detection parameters p is the set Y _t The number of parameters contained in (1).

And for higher than

The formula of the preferred one of the ultrasonic detection parameters with the strongest correlation and all valid parameter sets is as follows:

in the formula, Q _t Represents Y _t Wherein the calculated values of the average correlation coefficients are all larger than

The parameter set of (2).

In this example, use

Representing a slave parameter set Q _t The ultrasonic parameter with the strongest average correlation is selected as part 2And dividing effective ultrasonic detection parameters.

In the formula (I), the compound is shown in the specification,

i.e. from Y in rule 2 _t The selected 2 nd part of the effective ultrasonic testing parameters.

The set of valid ultrasound detection parameters ultimately determined for this embodiment.

The specific process of selecting effective characteristic parameters from the multi-dimensional ultrasonic detection parameters according to the correlation rules 1 and 2 is as follows:

step 1: initializing parameters, giving ultrasonic detection parameters Y, initial time t equal to 1, and interval step length

Lowest threshold value ρ _υ 。

Step 2: according to the formula:

and determining a selection interval corresponding to the selection time.

Step 3: determining a current interval

Whether greater than ρ _υ ；

If yes, operation Step3 continues.

Step 4: at time t, the interval is

Selecting corresponding parameter Y according to the correlation criterion 1 (formula 7) _t 。

Step 5: according to the correlation criterion 2 (equations 8-10), Y can be selected from the set _t Select necessary characteristic parameters

Step 6: update interval of

Returning to Step2, the necessary characteristic parameters are continuously selected from Y and are classified into a set

Tables 3 and 4 are the correlation coefficients between the ultrasonic testing parameters and the grain size (mean, standard deviation) and ultrasonic testing parameters. With the selection method provided in this embodiment, k is 10,

taking the value 0.2, ρ _υ The value is 0.4. When the average grain size is used as the target, it can be selected according to Step4

And is available according to Step5

The necessary parameters are obtained according to Stpe 3-Step 6

A _F2 、

Also meets the requirement, then selects

As input parameters for ultrasound and grain size (mean) modeling.

Similarly, when the standard deviation of the grain size is taken as an object, the standard deviation can be selected

As ultrasound and grain size (standard deviation) modeling inputsAnd entering parameters.

TABLE 3 correlation of ultrasonic characteristic parameters with primary alpha-phase grain size (mean, standard deviation)

TABLE 4 correlation between the inside of ultrasound characteristic parameters

Step three: and establishing a multi-parameter ultrasonic evaluation model based on monotonicity according to the final effective ultrasonic detection parameter set as a grain size soft measurement model by taking the maximum monotonicity as an optimization target.

The existing research focuses on determining the optimal evaluation model by taking the minimum error between the ultrasonic detection parameters and the fitting values as the target, but the monotonicity of the model cannot be ensured as various interference information such as irregular distribution of material grain sizes is increased. Fig. 4 is a schematic diagram illustrating the effectiveness of the average grain size provided in example 3 of the present invention. Fig. 4 shows an evaluation model (a special failure situation) with a minimum error as a target, specifically, an evaluation model with a minimum error as a target, which is established by (sound velocity standard deviation, attenuation coefficient average value, secondary bottom wave frequency peak value, relative nonlinear coefficient standard deviation) and grain size average value, and the error between the found value and the fitting value of the obtained ultrasonic detection parameter in the figure is really small, but the following disadvantages exist:

1) from the global perspective, the fit line formed by the model is on a horizontal line, so that the reflected grain size is greatly deviated when corresponding ultrasonic detection parameters (parameters with unobvious sample value distinction) are given, and the evaluation effect is basically lost.

2) From a local perspective, the optimizing lines also do not exhibit completely monotonicity, and such local evaluation also causes an anomaly in the results of obtaining the grain size.

From the above analysis, when the ultrasonic detection parameter evaluation model is determined, an uncertainty phenomenon, a local phenomenon, a global phenomenon and the like are formed by taking an error as a target, so that the feasibility is lacked. The evaluation model is constructed from another angle (monotonicity) target, so that the mapped ultrasonic detection parameters not only keep a monotone increasing or monotone decreasing form in the condition of ordered arrangement along with the grain sizes, but also can realize the effectiveness of the test outside the sample set.

The primary alpha-phase grain size of the TC4 titanium alloy is sensitive to the response of ultrasonic parameters, and the ultrasonic nondestructive evaluation mode is constructed by forming a mapping relation that ultrasonic characteristic parameters can directly represent the grain size. Aiming at the uncertainty problem of the effectiveness of the evaluation model taking errors as targets, the monotonicity is the maximum of an optimization target, and the optimization problem is solved by combining an optimization algorithm to obtain a measurement model, wherein the process comprises the following steps: introducing a dimension reduction mapping function to form a single-dimensional effective parameter converted by the mapping function containing the undetermined coefficient; then normalizing the obtained single-dimensional effective parameters and introducing a fitting function, establishing an optimization target with maximum monotonicity, wherein the difference value of the ultrasonic parameters and the grain size sample points is positive or negative one by one, and converting the process of determining the single ultrasonic parameters by the undetermined coefficients of the dimension reduction mapping function and the fitting function into an optimization problem. And solving the optimization problem by taking the optimization target as a strategy and combining with the SADE algorithm, finally solving the problem, finding the optimal undetermined coefficient of a mapping function and a fitting function, and determining a Multi-parameter Ultrasonic Evaluation (MUEBM) model Based on Monotonicity so as to obtain a grain size soft measurement model.

In order to convert a plurality of ultrasonic characteristic parameters into a single-dimensional parameter form, the evaluation model corresponding to the internal sample is conveniently established by the single-dimensional ultrasonic characteristics and the grain size (average value and standard deviation). Using quadratic polynomial as initial form of mapping function, using Y' selected by correlation as input parameter, and combining SADE algorithm to optimize input parameter and find out optimum undetermined coefficient lambda forming mapping function _ij And obtaining a mapping function f to determine the single-dimensional ultrasonic characteristic parameter Z.

The form of the dimensionality reduction mapping function constructed in the embodiment is as follows:

in the formula (lambda) _i1 ,λ _i2 ,λ _i3 ) Representing the dimensionality reduction mapping function coefficients, wherein i is 1,2, …, m;

4 ultrasonic characteristic parameter vectors are selected according to the correlation; z represents a single-dimensional effective ultrasound parameter corresponding to the sample contained in the grain size.

Because the information content of the extracted ultrasonic characteristic parameters is different and is not in a unified dimension, a normalization method is introduced to make the ultrasonic parameters with dimension difference within a specified range so as to facilitate analysis and subsequent modeling. In this embodiment, the single-dimensional effective parameters reduced to the single dimension by the dimension reduction mapping function f are subjected to the characteristic scale control and normalization processing to obtain normalized single-dimensional effective parameters. The normalization range is controlled to be (N, M), and the calculation formula of the normalization processing is as follows:

wherein M, N is the maximum value and the minimum value of normalization, which are respectively 0.99 and 0.01, Z is the single ultrasonic parameter to be normalized, and the normalized single-dimensional effective parameter is

min (Z) represents the minimum component of the single-dimensional ultrasound feature vector Z, and max (Z) represents the maximum component of the single-dimensional ultrasound feature vector Z.

In order to obtain a relation model between the single-dimensional ultrasonic characteristic parameter Z and the grain size X constructed in the formula (11), a least square method fitting function is introduced to draw up a linear relation between Z and X. Under the action of the fitting function, the single-dimensional ultrasonic parameters Z and X form a linear expression form of a first-order polynomial, and various acoustic characteristics of the ultrasonic are reflected by grain sizes in the relation of ultrasonic nondestructive evaluation. Thus, the corresponding ultrasonic parameters can be reflected by inputting the corresponding grain sizes in the corresponding form of the fitting relation with X as the independent variable and Z as the dependent variable.

In the embodiment, a first fitting function for fitting the average grain size and the normalized single-dimensional effective parameters is shown in a formula (13), wherein a dependent variable is the grain size, and an independent variable is the normalized single-dimensional effective parameters;

in the formula, X ^* Representing the grain size obtained after fitting according to the normalized single-dimensional effective parameters; lambda [ alpha ] _a And λ _b Is the fitting coefficient to be determined. The inverse transform of equation (13) into a second fitting function can represent the ultrasonic testing parameters as a function of the grain size X, as follows:

Z ^* ＝F′(X)＝ξ ₁ X+ξ ₂ (14)

in the formula, Z ^* Shows a normalized single-dimensional effective parameter, xi, obtained by fitting the grain size of an experimental sample as input ₁ ＝1/λ _a 、ξ ₂ ＝-λ _b /λ _a Respectively representing the undetermined coefficients of the fitting function.

An ultrasonic evaluation model which is obtained by collecting and initially calculating the ultrasonic characteristic parameters shown in the table 1 as input and is output by taking the average value and the standard deviation of the grain sizes in the table 2 and can reflect the conditions of the sizes of the microscopic grains and the uniformity of discrete distribution is established as a final grain size soft measurement model. In order to determine the evaluation model with the maximum and the best optimization target, each undetermined coefficient of a dimensionality reduction mapping function and a fitting function needs to be accurately obtained, so that higher measurement accuracy is achieved.

This embodiment is as follows

The maximum number of the differences between the sequential samples in the relation curve formed by the X and the X is preferably the same positive number or the same negative numberThe goal of quantization, i.e., monotonicity. When the optimization target is close to the maximum value, the obtained relation curve forms a monotone increasing or monotone decreasing trend and even can reach complete monotony, so that regular formation is realized under the condition that X is unchanged

And (4) arranging. The undetermined coefficients of the corresponding 2 groups of functions are determined through the optimization method, so that the grain size soft measurement model can be determined, the error between the fitting value and the calculated value is small, and the measurement precision of the model is improved.

FIG. 5 is a diagram illustrating the calculation monotonicity and non-monotonicity provided in embodiment 3 of the present invention. The curve in fig. 5 is a relationship curve of the second fitting function, and as can be seen from fig. 5, the abscissa in the graph shows an increasing trend of completely presenting a rule, while the ordinate does not completely present a decreasing trend. The difference value of the vertical coordinates between the point 2 and the point 1 is a negative number, namely 1 monotone section is presented, and the 10 points are sequentially calculated to obtain 9 monotone decreasing sections, namely the complete monotone is satisfied. The dotted line frame areas in the graph show that the

points

5 and 8 show ascending trends, the overall monotonous trend is violated, so that only 7 monotonous decreasing sections are formed in the graph 7, and the evaluation result is abnormal and deviated when a fitted line is formed in the violated area. The calculation formula for constructing the monotonicity target is as follows:

wherein the content of the first and second substances,

l′ _num representing the number of positive differences, n being the total number of normalized single-dimensional significant parameters in the monotonicity sample, k _i Normalized single-dimensional effective parameter Z representing second fitting function within monotonicity sample ^* The difference of the successive terms of (a); l _max-num The number of the difference values which are the successive terms is the maximum value of positive or negative.

Number phase of monotone section when searchingAt the same time, i.e. the number of differences of successive terms is a positive or negative maximum value l _max-num As such, the formula for considering the minimum error is as follows:

in the formula (I), the compound is shown in the specification,

it is indicated that when the found monotonicity numbers are the same, the optimal monotonicity target number with the minimum average absolute error is preferentially selected.

Monotonicity obtained according to a monotonicity strategy shows that the larger the monotonicity number is, the smaller the corresponding error is, and the better the obtained grain size soft measurement model is; the smaller the monotonic number, the larger the corresponding error, and the worse the grain size soft measurement model. Aiming at the optimization target problem, in order to ensure that the error of the grain size soft measurement model is restricted by good monotonicity, 2 groups of optimization coefficients are searched by combining the dimension reduction of a dimension reduction mapping function and a first fitting function, optimization and transformation in the fitting process and an SADE algorithm. The formula of the resulting optimization objective is as follows:

wherein λ ═ λ (λ) _i1 ,λ _i2 ,λ _i3 )、ξ＝(ξ ₁ ,ξ ₂ ) λ and ξ respectively represent the reduced-dimension mapping function coefficient and the second fitting function coefficient, l _max-num Normalized single-dimensional effective parameter Z representing second fitting function within monotonicity sample ^* The number of the difference values of the successive terms is the maximum value of positive or negative.

In this embodiment, when establishing a TC4 primary α -phase grain size ultrasonic soft measurement model based on monotonicity with respect to ultrasonic characteristic parameters and grain sizes, an implementation flow of constructing a primary α -phase grain size ultrasonic soft measurement model taking monotonicity strategy as an optimization target and simultaneously considering multi-parameter ultrasonic responses is as follows:

(1) redundant information aiming at a plurality of ultrasonic characteristic parameters Y is excessive, and the ultrasonic characteristic parameters Y' required by preliminary screening are limited by correlation analysis and intervals while the grain size is determined

Strong correlation with ultrasound parameters.

(2) And (3) reducing the dimension of the selected Y' through a quadratic polynomial dimension reduction mapping function f constructed by the formula 11 to obtain a new single-dimensional effective ultrasonic parameter Z.

(3) The normalization process shown in the formula (12) is adopted, so that Z is converted into Z in the same scale

And the model is convenient to build.

(4) Will be provided with

And

establishing an evaluation model, introducing a first fitting function which is a first-order fitting function F shown in a formula (13), and fitting by adopting the formula (13) and the formula (14) to obtain a newly-fitted single-dimensional effective ultrasonic parameter Z ^* 。

(5) According to the monotonicity target shown in equations (15) and (16), i.e. the one-dimensional effective ultrasound parameter Z ^* The maximum number of monotones for increasing or decreasing the included sample points determines the mode of calculating the monotonicity.

(6) Constructing an optimization target shown in a formula (17), and performing monotonicity on a monotonicity target l by using a SADE algorithm _max-num And optimizing, and searching an ideal dimensionality reduction mapping function coefficient lambda and a fitting function coefficient xi so as to determine a corresponding dimensionality reduction mapping function F and a corresponding fitting function F.

Step four: experimental validation and analysis

(1) Model experiments and result analysis in sample set

Ultrasonic testing parameters (containing 10 samples) extracted from tables 1 and 2Originally) and grain size as experimental basis, based on the selection

For input, the undetermined coefficient of the function f can be obtained according to the formula (11), and the fitting function coefficient is obtained by solving the optimization problem to establish a grain size soft measurement model.

When the monotonicity-based Evaluation model is established, all characteristic parameter sample values are newly sequenced along with the grain sizes, the Ultrasonic parameters Y' are input while 10 samples are selected as the basis, and a Multi-parameter Ultrasonic Evaluation (MUE) model which aims at minimizing errors is calculated. The ultrasonic wave and the grain size (average value and standard deviation) are plotted and expressed as f ═ xi ₁ X+ξ ₂ The corresponding dimensionality reduction mapping function coefficients are obtained as shown in tables 5 and 7. Taking the sample data in table 1 as a basis, respectively corresponding the extracted sound velocity average value, attenuation coefficient average value and nonlinear coefficient average value to the grain size and drawing a relation curve, directly establishing a fitting model for each ultrasonic parameter and the grain size by a least square method, and recording the obtained fitting evaluation model as gamma (tau) xi ₁ X+ξ ₂ Where (τ ═ 1,2,3) represents the category of conventional 3 measurement evaluation models. The values of the undetermined coefficients, the monotonic numbers, the error values, and the fitted correlation coefficients of the resulting model are shown in tables 6 and 8. Fig. 6 is a model and a fitting relationship curve of 5 types of average values of the evaluated grain sizes provided in example 3 of the present invention, in which part (a) of fig. 6 shows a MUEBM evaluation relationship curve, part (b) of fig. 6 shows a MUE evaluation relationship curve, part (c) of fig. 6 shows a relationship curve of sound velocity and average value of the grain sizes, part (d) of fig. 6 shows a relationship curve of attenuation coefficient and average value of the grain sizes, and part (e) of fig. 6 shows a relationship curve of nonlinear coefficient and average value of the grain sizes. FIG. 7 is a model and a fitting relationship curve of 5 types of standard deviations of evaluation grain sizes provided in example 3 of the present invention, in which part (a) of FIG. 7 shows a MUEBM evaluation relationship curve, part (b) of FIG. 7 shows a MUE evaluation relationship curve, part (c) of FIG. 7 shows a relationship curve of sound velocity and standard deviation of grain sizes, and part (d) of FIG. 7 shows a relationship curve of sound velocity and standard deviation of grain sizesThe attenuation coefficient is plotted against the grain size standard deviation, and part (e) of fig. 7 shows the nonlinear coefficient is plotted against the grain size standard deviation.

TABLE 5 undetermined coefficients of MUEBM, MUE model mapping functions targeting grain size averages

TABLE 6 evaluation model parameters targeting average grain size

TABLE 7 undetermined coefficients of MUEBM, MUE model mapping functions targeting grain size standard deviation

TABLE 8 evaluation model parameters targeting grain size standard deviation

Several model-related parameters were established for the ultrasonic parameters and the mean values of grain size, see table 6. With the first order polynomial fitting model established,

the mean absolute error value of the model was 0.0729,

the model and the other 3 models are 0.0969, 0.2068, 0.1999 and 0.1983 respectively, and it can be seen that

Model for the average between calculated and fitted valuesThe absolute difference is minimal, wherein

The model is also superior to the other 3 models. From the statistical condition of the maximum number of sample points contained in the calculated value, which is monotonically increased or decreased in sequence, the number of the sample points used by the model is 10,

models and

the models satisfy the maximum monotonic numbers of 7 and 5, while the other 3 models are 5, and 3, respectively, as is apparent

The model has the advantage of monotonicity, and meanwhile, the correlation coefficient of the model provided by the invention is 0.9307, so that the model has the advantage over other 4 models. Although it is used for

And

fitting can be performed by using a polynomial of degree 2, and the obtained average absolute error value is smaller than that of the fitting model of order 1, but it can be seen from parts (c) and (e) of fig. 7 that the fitting condition of order 2 causes evaluation failure, so it is appropriate to consider the fitting relation model of order 1.

As can be seen from the view in figure 7,

the fitting degree of (A) is very strong and the satisfied monotonicity performance is good, the discrimination of other 4 models is very large, and the state of monotonicity performance is poor, in addition

The overlap ratio of the model is also strong, but the displayed monotonous trend is inferior to that

The evaluation model is evaluated from the angles of monotonicity, correlation, discrimination, average absolute error and the like,

the model is superior to other 4 models, the presented error is minimum, the correlation is strongest and the monotonous performance is best, and the soft measurement model of the grain size provided by the invention has stronger characterization capability in a sample set through comprehensive analysis.

Several model-related parameters were established for the ultrasonic parameters and the standard deviation of the grain size, see table 8. According to the analysis of relevant parameters in the table, for establishing the fitting model of the first-order polynomial,

the mean absolute error of the model was 0.0342, while the other 3 models were 0.0820, 0.2325, 0.1912, respectively, as evident

The model has the smallest error value for statistics between the calculated value and the fitted value, wherein the error value of the attenuation coefficient model is the largest and the characterization capability is weak. From the statistical situation of the number of sample points contained in the calculated value which are monotonically increased or decreased in sequence,

the number of models is 9, and the monotonicity of the model satisfies the complete monotonicity, and ^s (1)、Γ ^s (2) and Γ ^s (3) The number of the monotone is 6, 5 and 4 respectively, and the monotone performance of the model is optimal, so that the model is close to a good condition, the correlation coefficient is-0.9868, is also a correlation coefficient better than other models, and is basically equal to the correlation coefficient

The models are equal. And

the model is compared with the model to determine,

the mean absolute error value of the model is slightly less than 0.0247, but is slightly more advantageous

The monotonic segment represented by the model is only 6, appearing to be not completely monotonic and inferior to

Complete monotonicity of the model. Similarly, in the case of the fitting of order 2 in fig. 7 (d), (e) and table 8, although the average absolute error of the obtained model becomes smaller, the evaluation result is invalid due to an inestimable phenomenon, and therefore, only the fitting model of order 1 is considered.

As can be seen from the view in figure 7,

the fitting line and the optimizing line of the model are basically in a superposition state and show a monotonous decreasing trend. While the lines of the actual values of the attenuation coefficient and nonlinear coefficient models and the fitted values are very different. For the other 3 models, only the two lines of the sound velocity model are slightly close. In addition, the

The models are more consistent but show a non-perfect monotonically decreasing trend. The evaluation model is judged from the angles of monotonicity, correlation, discrimination, average absolute error and the like, and the method provided by the invention

The model has stronger representation capability, and the result graph display of the comprehensive performance is superior to other 3 models.

(2) Model verification and result analysis outside sample set

And (3) carrying out verification outside the sample set and result analysis on the model established by the average value and standard deviation of the corresponding ultrasonic characteristic parameters and the grain sizes. To be provided with

And

2 test specimens, 11.67 μm and 13.64 μm, respectively, as the average of the grain sizes, to

And

2 test specimens of standard deviation of grain size, 3.64 μm and 5.79 μm, respectively. Will T ₁ 、T ₂ Inputting ultrasonic characteristic parameters of a test sample, calculating related ultrasonic parameters of the test sample by a MUEBM method, a MUE method, a sound velocity method, an attenuation coefficient method and a nonlinear coefficient method respectively to obtain corresponding grain sizes (average value and standard deviation), and comparing and analyzing the obtained result with the grain sizes (average value and standard deviation) measured by a metallographic method. The accuracy and the relative error value of the evaluation result are taken as the basis for judging the effectiveness of the model, and the evaluation results of the average value and the standard deviation of the grain sizes of the models are respectively shown in tables 9 and 10.

TABLE 9 comparison of evaluation results of 5 models targeting average grain size

TABLE 10 comparison of evaluation results of 5 models targeting standard deviation of grain size

To is directed at

The sample and the non-linear coefficient method model evaluation result have the maximum deviation which reaches 4.14 mu m, and the MUEBM model evaluation result has the minimum deviationThe grain size soft measurement model (MUEBM model) provided by the invention has the best evaluation result precision of 0.26 mu m. For the

In the samples, the evaluation result of the MUEBM model was 0.06 μm, which is superior to the general 3 models, and the evaluation result of the MUE model was inferior to the MUEBM model. From the relative error analysis, the MUEBM models have relative error values of-2.23% and 0.44% for 2 samples tested, respectively, and perform best compared to the other 4 models. Therefore, the MUEBM model established for the average grain size has the advantages of high evaluation result precision and small relative error.

In the same way, aim at

The evaluation result of the attenuation coefficient method is the best, the evaluation result of the MUEBM model is better than that of the other 3 models, and the evaluation result of the MUE model is the worst to reach 3.65 μm. For the

The sample has the largest deviation of the evaluation results of the attenuation coefficient model, which reaches 4.64 mu m, and the deviation of the evaluation results of the sound velocity method model is the smallest, which is only 0.52 mu m, while the evaluation results of the MUEBM method are superior to other methods and are basically equal to the sound velocity method. Analysis of the correlation in terms of relative error, in which the model of the attenuation coefficient method (A) ((B))

Sample), sonic velocity method model (

Sample) is slightly smaller, but the evaluation result is less than the evaluation result, and the test results of 2 samples cannot be guaranteed to be all superior. The evaluation result of the MUEBM model on 2 test samples is the most stable and reliable and better than the traditional 3 models, wherein the result presented by the verification result precision and relative error of the MUE model on 2 samples is inferior to the model. It can be seen that the MUEBM model overall evaluation established on the standard deviation of the grain sizeThe price result has high precision and small relative error value, which shows that the MUEBM model verifies the effectiveness of the test outside the sample set.

(3) MUEBM model validity analysis

Comprehensive analysis shows that the accuracy of the evaluation result of the MUEBM model constructed according to the grain sizes (average value and standard deviation) is far better than that of the MUE model, and the reliability of the model in a sample can only be ensured and the effectiveness of the evaluation result cannot be ensured due to the fact that the structure of the microscopic grain size of the TC4 titanium alloy is complex and the optimization model taking the error as the target. Therefore, the optimization target of monotonicity is newly drawn up, so that the optimizing line and the fitting line can both guarantee the monotonous ordered trend, the validity of the measurement result is guaranteed, and the result also shows advantages compared with 3 models with single parameter as the main part. This is because the change of ultrasonic parameters is caused by the difference of microscopic grain size and defect interaction of the material, and the obtained ultrasonic characteristic parameters contain different information of primary alpha phase grain size. The MUEBM evaluation method comprehensively considers the responses of the grain size (average value) to 4 ultrasonic characteristic parameters of sound velocity standard deviation, attenuation coefficient average value, secondary bottom wave frequency peak value and relative nonlinear coefficient standard deviation, simultaneously considers the influences of the grain size (average value) to the sound velocity standard deviation, the attenuation coefficient average value, secondary bottom wave frequency offset and the relative nonlinear coefficient standard deviation and takes monotonicity as an important strategy target to optimize the model, and eliminates the interference of redundant ultrasonic parameters through dimension reduction, fitting and optimization processing. For the construction of a measurement model, an optimized evaluation model which is searched with the minimum error has a small error, but forms a non-monotonous model or even an invalid model; on the contrary, the optimized evaluation model MUEBM which is searched by taking monotonicity as a target has small internal error, good monotonicity and good evaluation result outside a sample set, and the model presents regular monotonous trend and has small precision and good robustness of the evaluation result.

The MUEBM measurement model integrates effective multi-parameter ultrasonic characteristics with strong correlation of grain size and has good anti-interference performance. Although the 3 methods constructed by using the single ultrasonic parameter select the ultrasonic parameter with strong correlation with the grain size, the single ultrasonic parameter contains less acoustic information of grain size detection and cannot cover all the contained acoustic information, and the directly established fitting model has weak anti-interference capability and extremely unstable characterization results, so that the detection and evaluation are not ideal when the method is applied to actual detection and evaluation. On the contrary, the representation capability of the MUEBM evaluation method is ideal, and the grain size and the distribution uniformity of the discrete degree are comprehensively analyzed and represented.

The beneficial effect that this embodiment can realize is:

1) the structural complexity information aiming at the primary alpha phase grain size of the TC4 titanium alloy is increased, and the MUE model and the MUEBM model have error similarity. The former evaluation results are not ideal and good evaluation effects cannot be guaranteed. The monotonicity presented by the evaluation model with the monotonicity as the target is most ideal and even complete monotonicity, and the evaluation result has high precision and small relative error. The method shows that the method has the advantages that the method has no effectiveness for taking errors into consideration for the aviation titanium alloy material with complex information grain size as a modeling target, and the evaluation effect obtained by introducing a monotonicity target and ensuring the minimization of the model errors is the best.

2) When an optimization model with monotonicity as a target is constructed, multi-parameter responses need to be considered at the same time. Namely: preferably, a plurality of ultrasonic characteristic parameters are processed by means of mapping, normalization, fitting and the like, monotonicity is formulated as an optimization target, the target is optimized by combining an SADE algorithm, an optimal mapping function and fitting function coefficients are determined, and an ideal MUEBM evaluation model can be obtained.

3) Through experimental analysis, the MUEBM measurement model highlights superior performance. Compared with the MUE method, MUEBM shows a good evaluation result in consideration of an influence factor which is important for monotonicity; compared with a sound velocity model, an attenuation coefficient model and a nonlinear coefficient model, the new method integrates global information responding to the grain size, can represent the true value of the grain size, and can analyze the discrete degree of the grain size deviating from the true value, namely the uniformity distribution condition. An evaluation model with high precision, good monotonicity and small error is constructed by using the optimal characteristic parameters, and a result graph of the obtained model is visual and complete and has a stable evaluation effect.

The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.

The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the foregoing, the description is not to be taken in a limiting sense.

Claims

1. A method for determining alloy grain size based on mapping monotonicity, the method comprising:

establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set by taking the monotonicity maximum optimization target, and specifically comprising the following steps of:

substituting the optimal fitting coefficient into the first fitting function to obtain a grain size soft measurement model;

2. The method according to claim 1, wherein the determining a selection interval corresponding to the selection time according to the interval step and the selection time specifically includes:

according to the formula:

determining the selectionA selection interval corresponding to a time, wherein t represents a selection time,

3. The method according to claim 2, wherein determining the effective ultrasonic testing parameters at each selection time using a correlation metric criterion according to the minimum threshold, the selection interval, the respective ultrasonic testing parameter values and the grain size values specifically comprises:

4. A system for determining alloy grain size based on mapping monotonicity, the system comprising:

the grain size soft measurement model establishing module is used for establishing a grain size soft measurement model according to the final effective ultrasonic detection parameter set by taking monotonicity as a maximum optimization target, and specifically comprises the following steps:

the soft measurement model determining unit is used for substituting the optimal fitting coefficient into the first fitting function to obtain a grain size soft measurement model;

5. The determination system of claim 4, wherein the selection interval determination module is configured to:

6. The determination system according to claim 5, wherein the effective ultrasound detection parameter determination module specifically comprises:

and the selection time updating unit is used for updating the selection time.