CN103645248B - A kind of high-temperature alloy grain size evaluation method based on ultrasonic phase velocity - Google Patents

Abstract

The invention discloses a kind of high-temperature alloy grain size evaluation method based on ultrasonic phase velocity, comprise and extract high temperature alloy test block original A ripple information, calculate the longitudinal wave velocity of test block, obtain the phase velocity of test block, set up comprehensive evaluation model four steps, technique effect of the present invention is, by with two medium multivariate Gaussian sound-field models, obtain attenuation coefficient spectrum, again according to K-K relational expression, phase velocity spectrum is obtained by attenuation coefficient spectrum, phase velocity is resolved relative to use phase differential, this algorithm has higher stability, and provide possibility for the extraction of curved surface test block phase velocity, again because in the process using K-K relational expression, take longitudinal wave velocity as reference value, therefore in fact absorb the advantage of existing ultrasonic longitudinal wave sound velocity method, in addition, consider the dispersion degree of phase velocity and phase velocity simultaneously, make use of the velocity of sound information of diversification fully, establish comprehensive grain size evaluation model, improve the precision evaluating high-temperature alloy grain size with sound velocity method.

Description

A kind of high-temperature alloy grain size evaluation method based on ultrasonic phase velocity

Technical field

The present invention relates to a kind of measuring method of high-temperature alloy grain size, espespecially analyze hyperacoustic phase velocity in high temperature alloy, utilize ultrasonic phase velocity to carry out the method for Nondestructive Evaluation to high-temperature alloy grain size.

Background technology

High temperature alloy has alloy and has without outstanding advantages such as macrosegregations, be widely used in the high-end field such as space flight, military project, and grain size is the key factor affecting its mechanical property and serviceable life.

Current high-temperature alloy grain size mainly adopts metallographic examination to detect, the method has visual result and accuracy of detection comparatively advantages of higher, but detection efficiency is low, testing result can only reflect the crystal grain situation in the face of being observed, the professional standards impact of the examined personnel of accuracy of detection, and can damage material.

In harmless evaluation method, mainly contain eddy-current method and ultrasonic method.But the induction exchange current used in eddy-current method has kelvin effect, so the grain size of surface of test piece or nearly surface can only be reflected.And ultrasonic rule can reflect the microstructural characteristics at material internal deep layer place, wherein most popular with ultrasonic longitudinal wave sound velocity method, but existing ultrasonic longitudinal wave sound velocity method is for evaluating high-temperature alloy grain size, there is following deficiency:

1. longitudinal wave velocity is the speed that ultrasonic longitudinal wave ripple bag is transmitted in media as well, and ripple bag is formed by stacking by the sine wave of different frequency, and each frequency component in media as well velocity of propagation is called phase velocity.When ultrasound wave is propagated in media as well, some medium will play dispersion interaction to ultrasound wave, namely causes the difference of phase velocity.High temperature alloy belongs to polycrystalline material, it is exactly the large medium of dispersion degree, the velocity of propagation of each frequency component of ultrasound wave can be made to have larger difference, and existing method only does time-domain analysis to signal, do not consider the frequency domain characteristic of signal, the dispersion information relevant to grain size can not be provided, describe the impact of grain size on ultrasonic velocity all sidedly.

2. generally use repeatedly the advanced position of end ripple to determine sound path, calculate longitudinal wave velocity.From actual test case, the grain size of high temperature alloy is evaluated according to longitudinal wave velocity method, also exist because the decay of high temperature alloy to sound wave is serious, at the bottom of each time, the amplitude of ripple is less, end wavefront and noise signal are difficult to differentiate, thus it is inaccurate to cause longitudinal wave velocity to be measured, and finally makes the drawback that evaluation precision declines.

Summary of the invention

In order to improve the adverse effect that existing ultrasonic longitudinal wave sound velocity method is evaluated high-temperature alloy grain size, obtain grain size information accurately, the invention provides a kind of high-temperature alloy grain size evaluation method based on ultrasonic phase velocity, first attenuation coefficient spectrum is obtained, phase velocity spectrum is obtained again by K-K relational expression, by providing the phase velocity dispersion degree factor of phase velocity under certain frequency and alloy material, the grain size of comprehensive evaluation high temperature alloy simultaneously.

In order to realize above-mentioned technical purpose, technical scheme of the present invention is, a kind of high-temperature alloy grain size evaluation method based on ultrasonic phase velocity, comprises the following steps:

Step one: to the alloy test block needing to evaluate, measure the test block thickness L of position, measured point, then detected by pulse reflection method with supersonic detection device, position and the attitude of ultrasonic longitudinal wave probe need be adjusted when detecting, make axis and position, the measured point exact vertical of probe, original A wave datum is gone out by Detection and Extraction, an A wave datum array Awave_data stores, after the A wave datum of each alloy test block is all extracted, again metallographic examination is carried out to test block, the mean grain size in the multiple cross section of each test block is obtained with division lines method, be recorded in successively in array G, for the foundation of grain size evaluation model is prepared,

Step 2: the original A wave datum stored in the array Awave_data obtained by step one calculates longitudinal wave velocity v _gfirst manually gate 1, the reference position of gate 2 and gate length is adjusted, orient once ripple at the bottom of end ripple and secondary respectively, then apply rectangular window and intercept once wave number group BW2_data at the bottom of end wave number group BW1_data and secondary, again envelope extraction is carried out to array BW1_data and BW2_data Hilbert transform, and find forward first amplitude and be less than advanced position threshold value T the position that maximal value occurs from BW1_data and BW2_data _esampled point, obtain once the advanced position n of ripple at the bottom of end ripple, secondary ₁and n ₂, then sampling number Δ n=n apart between the advanced position of twice end ripple ₂-n ₁if the sampling time interval between two sampled points is dt, the test block thickness L recorded in integrating step one, according to formula v _g=2L/ (Δ ndt) calculates longitudinal wave velocity, repeatedly carries out this step to obtain the longitudinal wave velocity of all test blocks;

Step 3: the longitudinal wave velocity v that step 2 is obtained _g, and the variable relevant to the multivariate Gaussian sound-field model of two layer medium all substitutes in model, simulates the acoustic elasticity transport function t of surface wave and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω) the complex frequency spectrum V of the surface wave obtained by Fourier transform and once end ripple, is then combined _fW(ω) and V _bW1(ω), obtain attenuation coefficient spectrum α (ω), then according to K-K relational expression, in conjunction with attenuation coefficient spectrum α (ω), finally obtain phase velocity spectrum v _p(ω) this step, is repeatedly carried out to obtain the phase velocity spectrum of all test blocks;

Step 4: the contact between the mean grain size array G that each test block phase velocity spectrum obtained based on step 3 and step one obtain, sets up the high-temperature alloy grain size evaluation model based on ultrasonic phase velocity.

Described method, the step obtaining phase velocity spectrum in step 3 is:

Step 1: the reference position and the gate length that again manually adjust gate 2, orients surface wave, then intercepts out surface wave array FW_data;

Step 2: in the multivariate Gaussian sound-field model of two layer medium, the one way distance that ultrasound wave is propagated in water and alloy test block two media is respectively: the one way distance between ultrasonic probe to check point position is that the underwater sound is apart from L _w, by bubble sort method, find out the position that in FW_data, maximum absolute value value occurs, remember that this position is n ₀, and underwater sound speed v _wknown, then L _w=v _w(n ₀-1) dt/2, and the one way distance that ultrasound wave is propagated in alloy test block is the thickness L of test block, the variable also needed in computing simultaneously has, the longitudinal wave velocity v obtained in step 2 _g, the wafer radius r of circular ultrasonic longitudinal wave flat probe, the density p of water _w, the density p of high temperature alloy, waits the frequency sequence array f simulating frequency range;

Step 3: set up polar coordinates at the detecting head surface of circular ultrasonic longitudinal wave flat probe, carries out ring network to detecting head surface and formats discrete, the radial distance from the center of circle to edge is divided into N ₁equal portions, and make the center of circle be annulus, then there is N ₁+ 1 annulus, this N ₁+ 1 the pole footpath of annulus under polar coordinates is ascending is stored in array r, then rotates a circle with ray angle is divided into N ₂equal portions, the polar angle of every bar ray is stored in array θ successively, then pop one's head in by discrete as N ₁n ₂individual ring-type unit dimension, two then adjacent to array r annular radii averagings, can obtain each ring-type infinitesimal center point pole footpath array r _mid, finally footpath, the pole r of each ring-type infinitesimal central point _midwith polar angle θ, be converted into the two-dimensional discrete that rectangular coordinate x and y completes circular probe surface, namely use orderly real number to (x _j, y _k), j=1,2 ..., N ₁, k=1,2 ..., N ₂, record the coordinate of each ring-type infinitesimal central point, because acoustic axis line is vertical with position, measured point, therefore with (x _j, y _k) represent the coordinate of different sound path places acoustic beam radial section, and subtract each other the area ds that can obtain each ring-type infinitesimal by adjacent two sectorial areas;

Step 4: substituted into by each variable in step 2 and 3 in the multivariate Gaussian sound-field model of two layer medium and carry out computing, simulates the acoustic elasticity transport function t of surface wave and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω);

Step 5: the acoustic elasticity transport function t of the surface wave obtained by step 4 and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω), in conjunction with the complex frequency spectrum V of the surface wave obtained by Fourier transform and once end ripple _fW(ω) and V _bW1(ω) attenuation coefficient spectrum α (ω) is obtained;

Step 6: the centre frequency choosing ultrasonic probe used is reference frequency point ω ₀, and reference frequency point ω ₀the phase velocity v at place _p(ω ₀) the approximate longitudinal wave velocity v being taken as step 2 and obtaining _g, according to K-K relational expression, attenuation coefficient spectrum α (ω) obtained in integrating step 5, finally obtains phase velocity spectrum v _p(ω).

Described method, the step set up in step 4 based on the high-temperature alloy grain size evaluation model of ultrasonic phase velocity comprises:

Step 1: set up the grain size evaluation model based on single frequency phase velocity, first selected frequency is determined, pass through particle cluster algorithm, the frequency Selection Strategy of research phase velocity, select the optimum frequency the highest to grain size sensitivity: the total number of particles that note uses is M, wherein m particle (m=1,2,, M) positional representation be f _m, f _mnamely current frequency, note phase velocity spectrum lower dead center is f _min, top dead centre is f _maxtime, there is f _min<f _m<f _max, and the speed of particle m is expressed as v _m, individual optimal value is p _m, i.e. individual optimum frequency, colony's optimal value is p _g, i.e. the optimum frequency of colony, the function of speed and location updating is respectively v _m(t+1)=Ω v _m(t)+c ₁r ₁(p _m-f _i)+c ₂r ₂(p _g-f _m), f _m(t+1)=f _m(t)+v _m(t+1), getting inertial coefficient Ω is 0.729, gets learning coefficient c ₁and c ₂be 1.4962, r ₁and r ₂for the random number of [0,1], the highest iterations T _maxbe 100 times, total number of particles gets 40, and fitness function arranges as follows, and first selection substrate is with to at frequency f _mphase velocity array [the v of each test block lower _p] _mthe mean grain size array G obtained with step one, carries out quadratic fit by least square method

Separate above-mentioned normal equation system, obtain choosing frequency f _mtime matched curve [S ^*(v _p)] _m=[a ₀] _m+ [a ₁] _mv _p+ [a ₂] _mv _p ², operational symbol () wherein represents asks two inner product of vectors, as

Therefore the square error after selection matching is as fitness F _m, to weigh the sensitivity under this frequency between phase velocity and grain size

According to fitness F _msearch for, the optimum frequency f making fitness reach minimum value can be found _p, the phase velocity v under this frequency _p| _fpthe highest to the sensitivity of grain size, finally depict each test block at optimum frequency f _punder phase velocity-grain size matched curve, and it can be used as the grain size evaluation model g based on single frequency phase velocity ₁(v _p| _fp);

Step 2: set up the grain size evaluation model based on the dispersion degree factor, the dispersion degree factor is represented with the slope of phase velocity spectrum, first to the phase velocity spectrum least square method of each test block, linear fit is done to phase velocity and frequency, try to achieve the slope Δ v of each test block phase velocity spectrum _p/ Δ f; Then to the phase velocity spectrum slope Δ v of each test block _p/ Δ f and average grain size number group G, does quadratic fit by least square method again, finally depicts each test block phase velocity spectrum slope-grain size curve, and it can be used as the grain size evaluation model g based on the dispersion degree factor ₂(Δ v _p/ Δ f);

Step 3: the grain size evaluation model based on single frequency phase velocity that integrating step 1 obtains, the grain size evaluation model based on the dispersion degree factor obtained with step 2, the comprehensive high-temperature alloy grain size evaluation model based on ultrasonic phase velocity can be obtained, realized by weighted mean

g(Q)＝w ₁·g ₁([v _P| _fP] _Q)+w ₂·g ₂([Δv _P|Δf] _Q)

In formula, Q represents Q alloy test block, [v _p| _fp] _qbe that Q test block is at frequency f _punder phase velocity, [Δ v _p/ Δ f] _qbe the dispersion degree factor of Q test block, w ₁the weight of the grain size evaluation model based on single frequency phase velocity, w ₂it is the weight of the grain size evaluation model based on single frequency phase velocity.

Technique effect of the present invention is, by with two medium multivariate Gaussian sound-field models, obtain attenuation coefficient spectrum, again according to K-K relational expression, phase velocity spectrum is obtained by attenuation coefficient spectrum, resolve phase velocity relative to use phase differential, this algorithm has higher stability, and provides possibility for the extraction of curved surface test block phase velocity; Again because in the process using K-K relational expression, take longitudinal wave velocity as reference value, therefore in fact absorb the advantage of existing ultrasonic longitudinal wave sound velocity method; In addition, consider the dispersion degree of phase velocity and phase velocity simultaneously, make use of the velocity of sound information of diversification fully, establish comprehensive grain size evaluation model, improve the precision evaluating high-temperature alloy grain size with sound velocity method.

Accompanying drawing explanation

Fig. 1 is overview flow chart of the present invention;

Fig. 2 is supersonic detection device figure of the present invention;

Fig. 3 is A waveform schematic diagram of the present invention;

Fig. 4 is that wave envelope at the bottom of the once end of the present invention ripple, secondary extracts schematic diagram;

Fig. 5 is attenuation coefficient of the present invention spectrum schematic diagram;

Fig. 6 is phase velocity of the present invention spectrum schematic diagram;

Fig. 7 is the test block pictorial diagram of this example;

Fig. 8 is the metallographic examination system of this example;

The high temperature alloy metallograph of this example of Fig. 9;

Figure 10 is the phase velocity spectrum of each test block of the different grain size of this example;

Figure 11 is the grain size evaluation model of this example based on single frequency phase velocity;

Figure 12 is the grain size evaluation model of this example based on the dispersion degree factor.

Wherein, 1 be Ultrasound Instrument, 2 for ultrasonic probe, 3 for 6-dof motion platform, 4 for motion control card, 5 for tank,

6 be high-speed data acquisition card, 7 be computing machine, 8 be probe holder, 9 be control circuit, 10 for detected object.

Embodiment

See Fig. 2-Figure 12, the high temperature alloy that this embodiment is In718 for a collection of trade mark, illustrates evaluation method of the present invention.First detected object is fixed on and fills in the tank of water by the present invention, ultrasonic longitudinal wave is encouraged to pop one's head in ultrasonic pulse generation/receiver (also known as Ultrasound Instrument), ultrasonic longitudinal wave probe is held on 6-dof motion platform by probe holder, the motion of 6-dof motion platform is controlled by the motion control card connection control circuit that computing machine is installed, adjustment ultrasonic probe pose in the sink, and obtain with the high-speed data acquisition card on computing machine and store the ultrasonic signal that Ultrasound Instrument receives, finally carry out further treatment and analysis on computers.The step of test and repair is as follows:

Step one: first to the alloy test block needing to evaluate, with supersonic detection device as shown in Figure 2, detected by pulse reflection method, extract original A wave datum, and the mean grain size that metallographic examination obtains the multiple cross section of each test block is carried out to test block, be recorded in successively in array G, for the foundation of grain size evaluation model is prepared;

Step 1: in the centre of test block, chooses position, a measured point, and the upper mark of note, the test block thickness L of position, measured point is measured with micrometer caliper; Alloy test block is placed in tank detect;

Step 2: with probe holder clamping probe, be connected on 6-dof motion platform, by the position of motion control card adjustment probe, close to position, measured point, depart from for preventing actual sound path and measured test block thickness, affect ultrasonic velocity information to extract accurately, probe axis must with the section exact vertical of position, measured point, need the A of controlled motion platform, the attitude of B axial adjustment probe aims at position, measured point, by can realize the automatic adjustment of probe attitude in conjunction with ultrasonic A waveform, when surface wave reflected amplitudes is maximum, can think that the acoustic axis line of probe reaches vertical with the section of position, test block measured point, A can be recorded in, during B axle moves continuously, the surface wave reflected amplitudes that different attitude obtains, finally navigate to the attitude that a surface wave reflected amplitudes is the strongest, complete the automatic adjustment of attitude, the automatic adjustment of attitude ensure that the simplicity of operation and the repeatability of test,

Step 3: adjustment underwater sound distance, A ripple is made to present surface wave and at least twice end ripple, as shown in Figure 3, finally store the raw ultrasound A wave datum under this some position rf-mode lower, in fact an A ripple signal can store with an array Awave_data, change next test block, repeat step 1 and step 2;

Step 4: metallographic examination is carried out to each high temperature alloy, choose three representational cross sections, the mean grain size in three cross sections is obtained with resection, the mean grain size of whole test block is represented with the grain size mean value in three cross sections, after obtaining the mean grain size in the multiple cross section of each test block, be recorded in successively in array G, for the foundation of grain size evaluation model is prepared;

Step 2: calculate longitudinal wave velocity v by the original A wave datum that step one obtains _g, repeatedly carry out the longitudinal wave velocity that this step can obtain all test blocks, for computing is below prepared, comprise following steps:

Step 1: according to the feature of A waveform, surface wave can be told artificially, once ripple at the bottom of end ripple and secondary, first gate 1, the reference position of gate 2 and gate length will manually be adjusted, orient once ripple at the bottom of end ripple and secondary respectively, then apply rectangular window and intercept once wave number group BW2_data at the bottom of end wave number group BW1_data and secondary;

Step 2: general measure longitudinal wave velocity needs to use envelope pattern, and gather with rf-mode due to A ripple signal, need to carry out envelope extraction to array BW1_data and BW2_data, can select by Hilbert transform, then delivery realizes (can directly use these two system functions of hilbert () and abs () to realize in Matlab);

Step 3: find out by bubble sort method the position that in BW1_data and BW2_data, maximal value occurs, an advanced position threshold value T is set _e, get T _ebe 0.02, from the position that maximal value occurs, forwards search for, get the advanced position that sequence number that first amplitude be less than the sampled point of 0.02 is ripple at the bottom of this time, remember that once the advanced position of end ripple is n ₁, the advanced position of ripple at the bottom of secondary is n ₂, as shown in Figure 4;

Step 4: use once the advanced position of ripple at the bottom of end ripple and secondary to determine sound path, sampling number Δ n=n apart between the advanced position of twice end ripple ₂-n ₁, the sampling time interval between two sampled points is dt, in conjunction with the thickness L of test block, and can according to following formula v _g=2L/ (Δ ndt) calculates longitudinal wave velocity.

Step 3: the longitudinal wave velocity v that step 2 is obtained _g, and the variable relevant to the multivariate Gaussian sound-field model of two layer medium all substitutes in model, simulates the acoustic elasticity transport function t of surface wave and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω) the complex frequency spectrum V of the surface wave obtained by Fourier transform and once end ripple, is then combined _fW(ω) and V _bW1(ω), obtain attenuation coefficient spectrum α (ω), then according to K-K relational expression, in conjunction with attenuation coefficient spectrum α (ω), finally obtain phase velocity spectrum v _p(ω), repeatedly carry out the phase velocity spectrum that this step can obtain all test blocks, comprise following steps:

Step 1: the reference position and the gate length that again manually adjust gate 2, orients surface wave, then intercepts out surface wave array FW_data;

Step 2: in the multivariate Gaussian sound-field model of two layer medium, the one way distance that ultrasound wave is propagated in water and alloy test block two media is respectively: the one way distance between ultrasonic probe to check point position is that the underwater sound is apart from L _w, by bubble sort method, find out the position that in FW_data, maximum absolute value value occurs, remember that this position is n ₀, and underwater sound speed v _wknown, then L _w=v _w(n ₀-1) dt/2, and the one way distance that ultrasound wave is propagated in alloy test block is the thickness L of test block, the variable also needed in computing simultaneously has, the longitudinal wave velocity v obtained in step 2 _g, the wafer radius r of circular ultrasonic longitudinal wave flat probe, the density p of water _w, the density p of high temperature alloy, waits the frequency sequence array f simulating frequency range;

Step 3: set up polar coordinates at the detecting head surface of circular ultrasonic longitudinal wave flat probe, carries out ring network to detecting head surface and formats discrete, the radial distance from the center of circle to edge is divided into N ₁equal portions, and make the center of circle be annulus, then there is N ₁+ 1 annulus, this N ₁+ 1 the pole footpath of annulus under polar coordinates is ascending is stored in array r, then rotates a circle with ray angle is divided into N ₂equal portions, the polar angle of every bar ray is stored in array θ successively, then pop one's head in by discrete as N ₁n ₂individual ring-type unit dimension, two then adjacent to array r annular radii averagings, can obtain each ring-type infinitesimal center point pole footpath array r _mid, finally footpath, the pole r of each ring-type infinitesimal central point _midwith polar angle θ, be converted into the two-dimensional discrete that rectangular coordinate x and y completes circular probe surface, namely use orderly real number to (x _j, y _k), j=1,2 ..., N ₁, k=1,2 ..., N ₂, record the coordinate of each ring-type infinitesimal central point, because acoustic axis line is vertical with position, measured point, then can use (x _j, y _k) represent and the coordinate of different sound path places acoustic beam radial section in addition, subtract each other the area ds that can obtain each ring-type infinitesimal by adjacent two sectorial areas;

Step 4: substituted into by each variable in step 2 and 3 in the multivariate Gaussian sound-field model of two layer medium and carry out computing, simulates the acoustic elasticity transport function t of surface wave and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω), first the Relative Vibration velocity expression of surface wave under different acoustic beam radial section coordinates is

$\begin{array}{c}\frac{v_{FW}(x_j,y_k,\mathrm{\&omega;})}{v_0\left(\mathrm{\&omega;}\right)}=R_{12}\underset{q=1}{\overset{15}{\&Sigma;}}{A}_q\frac{\sqrt{\det {\&lsqb;M_2\&rsqb;}_q}}{\sqrt{\det {\&lsqb;M_{2;0}\&rsqb;}_q}}\frac{\sqrt{\det {\&lsqb;M_1\&rsqb;}_q}}{\sqrt{\det {\&lsqb;M_{1:0}\&rsqb;}_q}}\\ \&times;\exp \&lsqb;i\mathrm{\&omega;}(\frac{2L_w}{v_w}+\frac{1}{2}\left(\begin{array}{cc}{x}_j& y_k\end{array}\right){\&lsqb;M_2\&rsqb;}_q\left(\begin{array}{c}{x}_j\\ y_k\end{array}\right))\&rsqb;\end{array}---\left(1\right)$

In formula, angular frequency=2 π f, f ∈ f, v _fWfor the vibration velocity of surface wave, v ₀(ω) be the vibration velocity that detecting head surface is initial, orderly real number is to (x _j, y _k) obtained by step 3, j=1,2 ..., N ₁, k=1,2 ..., N ₂, R ₁₂for the reflection coefficient between water-alloy interface, be specially

$R_{12}=\frac{{\mathrm{\&rho;v}}_G-{\mathrm{\&rho;}}_w{v}_w}{{\mathrm{\&rho;v}}_G+{\mathrm{\&rho;}}_w{v}_w}$

Four Metzler matrix when superposing for the q time in formula (1) are respectively

${\&lsqb;M_{1;0}\&rsqb;}_q=\left[\begin{array}{cc}{\mathrm{iB}}_q/\left(v_w{D}_R\right)& 0\\ 0& {\mathrm{iB}}_q/\left(v_w{D}_R\right)\end{array}\right]---\left(2.1\right)$

${\&lsqb;M_1\&rsqb;}_q=\&lsqb;D^{G_1}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_1}\&rsqb;{\&lsqb;A^{G_1}+B^{G_1}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(2.2\right)$

${\&lsqb;M_{2;0}\&rsqb;}_q=\&lsqb;D^{G_{2;0}}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_{2;0}}\&rsqb;{\&lsqb;A^{G_{2;0}}+B^{G_{2;0}}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(2.3\right)$

${\&lsqb;M_2\&rsqb;}_q=\&lsqb;D^{G_2}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_2}\&rsqb;{\&lsqb;A^{G_2}+B^{G_2}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(2.4\right)$

A in formula (1) and formula (2.1) _qand B _q, be 15 groups of multivariate Gaussian superposition complex coefficients, its value is as shown in table 1

Table 1 15 groups of multivariate Gaussian superposition complex coefficients

D in formula (2.1) _rfor Rayleigh distance, expression is

$D_R=\frac{{\mathrm{\&omega;r}}^2}{2v_w}$

Formula (2.2) is respectively to three overall A, B, C, D matrix in formula (2.4)

$\left[\begin{array}{cc}{A}^{G_1}& B^{G_1}\\ C^{G_1}& D^{G_1}\end{array}\right]=\left[\begin{array}{cc}{A}_{p1}& B_{p1}\\ C_{p1}& D_{p1}\end{array}\right]---\left(3.1\right)$

$\left[\begin{array}{cc}{A}^{G_{2;0}}& B^{G_{2;0}}\\ C^{G_{2;0}}& D^{G_{2;0}}\end{array}\right]=\left[\begin{array}{cc}{A}_{r1}& B_{r1}\\ C_{r1}& D_{r1}\end{array}\right]\left[\begin{array}{cc}{A}^{G_1}& B^{G_1}\\ C^{G_1}& D^{G_1}\end{array}\right]---\left(3.2\right)$

$\left[\begin{array}{cc}{A}^{G_2}& B^{G_2}\\ C^{G_2}& D^{G_2}\end{array}\right]=\left[\begin{array}{cc}{A}_{p2}& B_{p2}\\ C_{p2}& D_{p2}\end{array}\right]\left[\begin{array}{cc}{A}^{G_{2;0}}& B^{G_{2;0}}\\ C^{G_{2;0}}& D^{G_{2;0}}\end{array}\right]---\left(3.3\right)$

Each A, B, C, D matrix in above formula are respectively

$A_{p1}=D_{p1}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{p1}=v_w{L}_w\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],C_{p1}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

$A_{r1}=D_{r1}=\left[\begin{array}{cc}-1& 0\\ 0& 1\end{array}\right],B_{r1}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right],C_{r1}=\frac{2}{v_w}\left[\begin{array}{cc}-h_{11}^1& 0\\ 0& h_{22}^1\end{array}\right]$

$A_{p2}=D_{p2}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{p2}=v_w{L}_w\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],C_{p2}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

H in formula ¹for the curvature matrix of water-alloy interface, its subscript refers to the 1st interface, h ¹ ₂₂subscript refer to the 2nd row the 2nd column element

$h^1=\left[\begin{array}{cc}1/R_x^1& 0\\ 0& 1/R_y^1\end{array}\right]$

R in formula _x ¹and R _y ¹, be respectively the radius-of-curvature of two orthogonal directionss, if alloy upper surface is plane, then R _x ¹and R _y ¹for infinity, so far, the Relative Vibration velocity expression of surface wave can be obtained, the acoustic elasticity transport function of surface wave under certain frequency can be obtained by the Relative Vibration speed of surface wave

$t_{A;FW}\left(\mathrm{\&omega;}\right)=\frac{2}{S}\underset{S}{\&Integral;}\frac{v_{FW}(xj,yk,\mathrm{\&omega;})}{v_0\left(\mathrm{\&omega;}\right)}\&CenterDot;ds---\left(4\right)$

In formula, S is the total area of detecting head surface, S=π r ²ds is the area of each ring-type infinitesimal, and integration item is just in formula (1), the Relative Vibration speed of surface wave under different acoustic beam radial section coordinates, on the other hand, once the Relative Vibration velocity expression of end ripple under different acoustic beam radial section coordinates is

$\begin{array}{c}\frac{v_{{\mathrm{BW}}_1}(x_j,y_k,\mathrm{\&omega;})}{v_0\left(\mathrm{\&omega;}\right)}=T_{12}{R}_{21}{T}_{21}\underset{q=1}{\overset{15}{\&Sigma;}}{A}_q\frac{\sqrt{\det {\&lsqb;M_4\&rsqb;}_q}}{\sqrt{\det {\&lsqb;M_{4;0}\&rsqb;}_q}}\frac{\sqrt{\det {\&lsqb;M_3\&rsqb;}_q}}{\sqrt{\det {\&lsqb;M_{3;0}\&rsqb;}_q}}\frac{\sqrt{\det {\&lsqb;M_2\&rsqb;}_q}}{\sqrt{\det {\&lsqb;M_{2;0}\&rsqb;}_q}}\frac{\sqrt{\det {\&lsqb;M_1\&rsqb;}_q}}{\sqrt{\det {\&lsqb;M_{1;0}\&rsqb;}_q}}\\ \&times;\exp \&lsqb;i\mathrm{\&omega;}(\frac{2L_w}{v_w}+\frac{2L}{v_G}+\frac{1}{2}\left(\begin{array}{cc}{x}_j& y_k\end{array}\right){\&lsqb;M_4\&rsqb;}_q\left(\begin{array}{c}{x}_j\\ y_k\end{array}\right))\&rsqb;\end{array}---\left(5\right)$

In formula, v _bW1for the vibration velocity of once end ripple, v ₀(ω) be still the initial vibration velocity of detecting head surface, orderly real number is to (x _j, y _k) obtained by step 3, j=1,2 ..., N ₁, k=1,2 ..., N ₂, T ₁₂for the transmission coefficient between water-alloy interface, R ₂₁for the reflection coefficient between alloy-water termination, T ₂₁for the transmission coefficient between alloy-water termination, be specially

$T_{12}=\frac{2{\mathrm{\&rho;v}}_G}{{\mathrm{\&rho;v}}_G+{\mathrm{\&rho;}}_w{v}_w},R_{21}=\frac{{\mathrm{\&rho;}}_w{v}_w-{\mathrm{\&rho;v}}_G}{{\mathrm{\&rho;v}}_G+{\mathrm{\&rho;}}_w{v}_w},T_{21}=\frac{2{\mathrm{\&rho;}}_w{v}_w}{{\mathrm{\&rho;v}}_G+{\mathrm{\&rho;}}_w{v}_w}$

Eight Metzler matrix during in formula (8) the q time superposition, front four completely the same in form with formula (2.1) in formula (2.4), then four are respectively

${\&lsqb;M_{3;0}\&rsqb;}_q=\&lsqb;D^{G_{3;0}}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_{3;0}}\&rsqb;{\&lsqb;A^{G_{3;0}}+B^{G_{3;0}}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(6.1\right)$

${\&lsqb;M_3\&rsqb;}_q=\&lsqb;D^{G_3}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_3}\&rsqb;{\&lsqb;A^{G_3}+B^{G_3}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(6.2\right)$

${\&lsqb;M_{4;0}\&rsqb;}_q=\&lsqb;D^{G_{4;0}}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_{4;0}}\&rsqb;{\&lsqb;A^{G_{4;0}}+B^{G_{4;0}}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(6.3\right)$

${\&lsqb;M_4\&rsqb;}_q=\&lsqb;D^{G_4}{\&lsqb;M_{1;0}\&rsqb;}_q+C^{G_4}\&rsqb;{\&lsqb;A^{G_4}+B^{G_4}{\&lsqb;M_{1;0}\&rsqb;}_q\&rsqb;}^{-1}---\left(6.4\right)$

Seven overall A, B, C, D matrix using in Metzler matrix are respectively

$\left[\begin{array}{cc}{A}^{G_1}& B^{G_1}\\ C^{G_1}& D^{G_1}\end{array}\right]=\left[\begin{array}{cc}{A}_{p1}& B_{p1}\\ C_{p1}& D_{p1}\end{array}\right]---\left(7.1\right)$

$\left[\begin{array}{cc}{A}^{G_{2;0}}& B^{G_{2;0}}\\ C^{G_{2;0}}& D^{G_{2;0}}\end{array}\right]=\left[\begin{array}{cc}{A}_{t1}& B_{t1}\\ C_{t1}& D_{t1}\end{array}\right]\left[\begin{array}{cc}{A}^{G_1}& B^{G_1}\\ C^{G_1}& D^{G_1}\end{array}\right]---\left(7.2\right)$

$\left[\begin{array}{cc}{A}^{G_2}& B^{G_2}\\ C^{G_2}& D^{G_2}\end{array}\right]=\left[\begin{array}{cc}{A}_{p2}& B_{p2}\\ C_{p2}& D_{p2}\end{array}\right]\left[\begin{array}{cc}{A}^{G_{2;0}}& B^{G_{2;0}}\\ C^{G_{2;0}}& D^{G_{2;0}}\end{array}\right]---\left(7.3\right)$

$\left[\begin{array}{cc}{A}^{G_{3;0}}& B^{G_{3;0}}\\ C^{G_{3;0}}& D^{G_{3;0}}\end{array}\right]=\left[\begin{array}{cc}{A}_{r2}& B_{r2}\\ C_{r2}& D_{r2}\end{array}\right]\left[\begin{array}{cc}{A}^{G_2}& B^{G_2}\\ C^{G_2}& D^{G_2}\end{array}\right]---\left(7.4\right)$

$\left[\begin{array}{cc}{A}^{G_3}& B^{G_3}\\ C^{G_3}& D^{G_3}\end{array}\right]=\left[\begin{array}{cc}{A}_{p3}& B_{p3}\\ C_{p3}& D_{p3}\end{array}\right]\left[\begin{array}{cc}{A}^{G_{3;0}}& B^{G_{3;0}}\\ C^{G_{3;0}}& D^{G_{3;0}}\end{array}\right]---\left(7.5\right)$

$\left[\begin{array}{cc}{A}^{G_{4;0}}& B^{G_{4;0}}\\ C^{G_{4;0}}& D^{G_{4;0}}\end{array}\right]=\left[\begin{array}{cc}{A}_{t3}& B_{t3}\\ C_{t3}& D_{t3}\end{array}\right]\left[\begin{array}{cc}{A}^{G_3}& B^{G_3}\\ C^{G_3}& D^{G_3}\end{array}\right]---\left(7.6\right)$

$\left[\begin{array}{cc}{A}^{G_4}& B^{G_4}\\ C^{G_4}& D^{G_4}\end{array}\right]=\left[\begin{array}{cc}{A}_{p4}& B_{p4}\\ C_{p4}& D_{p4}\end{array}\right]\left[\begin{array}{cc}{A}^{G_{4;0}}& B^{G_{4;0}}\\ C^{G_{4;0}}& D^{G_{4;0}}\end{array}\right]---\left(7.7\right)$

Each A, B, C, D matrix in above formula are respectively

$A_{p1}=D_{p1}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{p1}=v_w{L}_w\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],C_{p1}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

$A_{t1}=D_{t1}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{t1}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right],C_{t1}=(\frac{1}{v_w}-\frac{1}{v_G})\left[\begin{array}{cc}{h}_1^1& 0\\ 0& h_{22}^1\end{array}\right]$

$A_{p2}=D_{p2}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{p2}=v_{G}L\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],C_{p2}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

$A_{r2}=D_{r2}=\left[\begin{array}{cc}-1& 0\\ 0& 1\end{array}\right],B_{r2}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right],C_{r2}=\frac{2}{v_G}\left[\begin{array}{cc}-h_1^2& 0\\ 0& h_{22}^2\end{array}\right]$

$A_{p3}=D_{p3}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{p3}=v_{G}L\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],C_{p3}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

$A_{t3}=D_{t3}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{t3}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right],C_{t3}=(\frac{1}{v_G}-\frac{1}{v_w})\left[\begin{array}{cc}{h}_1^3& 0\\ 0& h_{22}^3\end{array}\right]$

$A_{p4}=D_{p4}=\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],B_{p4}=v_w{L}_w\left[\begin{array}{cc}1& 0\\ 0& 1\end{array}\right],C_{p4}=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]$

H in formula ³with h ¹be all the curvature matrix of water-alloy interface, but due to method direction contrary, then h ³=-h ¹, h ²for the curvature matrix of alloy-water termination

$h^2=\left[\begin{array}{cc}1/R_x^2& 0\\ 0& 1/R_y^2\end{array}\right]$

R in formula _x ²and R _y ², be respectively the radius-of-curvature of two orthogonal directionss, if alloy upper surface is plane, then R _x ²and R _y ²for infinity, so far, the Relative Vibration velocity expression of end ripple can be obtained once, the acoustic elasticity transport function of end ripple under certain frequency can be obtained once by the Relative Vibration speed of once end ripple

$t_{A;BW1}\left(\mathrm{\&omega;}\right)=\frac{2}{S}\underset{S}{\&Integral;}\frac{v_{BW1}(xj,yk,\mathrm{\&omega;})}{v_0\left(\mathrm{\&omega;}\right)}\&CenterDot;ds---\left(8\right)$

In formula, S is still the total area of detecting head surface, S=π r ²ds is still the area of each ring-type infinitesimal, and integration item is also just in formula (5), the once Relative Vibration speed of end ripple under different acoustic beam radial section coordinates, now by formula (4) and formula (8), travel through whole waiting and simulate the frequency sequence of frequency range, the acoustic elasticity transport function t of complete surface wave in a frequency range and once end ripple can be simulated _{a; FW}(ω) and t _{a; BW1}(ω);

Step 5: the acoustic elasticity transport function t of the surface wave obtained by step 4 and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω), in conjunction with the complex frequency spectrum V of the surface wave obtained by Fourier transform and once end ripple _fW(ω) and V _bW1(ω) attenuation coefficient spectrum α (ω) is obtained

$\mathrm{\&alpha;}\left(\mathrm{\&omega;}\right)=-\frac{1}{2L}ln\left(\right|\frac{t_{A;FW}\left(\mathrm{\&omega;}\right)}{t_{A;BW1}\left(\mathrm{\&omega;}\right)}\&CenterDot;\frac{V_{BW1}\left(\mathrm{\&omega;}\right)}{V_{FW}\left(\mathrm{\&omega;}\right)}\left|\right)$

In formula, V _fW(ω) and V _bW1(ω) be respectively, surface wave array FW_data and once end wave number group BW1_data, do Fast Fourier Transform (FFT) and obtain complex frequency spectrum (can directly use system function fft () to realize Fast Fourier Transform (FFT) in Matlab), the attenuation coefficient that can be obtained as shown in Figure 5 by the method composes α (ω);

Step 6: the centre frequency first choosing ultrasonic probe used when carrying out the collection of A ripple is reference frequency point ω ₀, and reference frequency point ω ₀the phase velocity v at place _p(ω ₀) the approximate longitudinal wave velocity v being taken as step 2 and obtaining _g, according to K-K relational expression, attenuation coefficient spectrum α (ω) obtained in integrating step 5, finally obtains phase velocity spectrum v _p(ω).

$v_P\left(\mathrm{\&omega;}\right)={\&lsqb;{\left(v_P\right({\mathrm{\&omega;}}_0\left)\right)}^{-1}-\frac{2}{\mathrm{\&pi;}}{\&Integral;}_{{\mathrm{\&omega;}}_0}^{\mathrm{\&omega;}}\frac{\mathrm{\&alpha;}\left(\mathrm{\&omega;}\right)}{{\mathrm{\&omega;}}^2}\&CenterDot;d\mathrm{\&omega;}\&rsqb;}^{-1}$

Owing to waiting that the frequency sequence simulating frequency range is discrete, integration in above-mentioned is actually numerical integration, can be calculated by newton-Ke Tesi formula or Long Beige algorithm, here select comparatively simple and precision meets user demand trapezoid formula calculates (can directly use system function trapz () to realize in Matlab), the phase velocity spectrum obtained by the method, phase velocity spectrum as shown in Figure 6.

Step 4: the contact between the mean grain size array G that each test block phase velocity spectrum obtained based on step 3 and step one obtain, set up the high-temperature alloy grain size evaluation model based on ultrasonic phase velocity, comprise following steps:

Step 1: set up the grain size evaluation model based on single frequency phase velocity, first selected frequency is determined, pass through particle cluster algorithm, the frequency Selection Strategy of research phase velocity, select the optimum frequency the highest to grain size sensitivity: the total number of particles that note uses is M, wherein m particle (m=1,2,, M) positional representation be f _m, f _mnamely current frequency, note phase velocity spectrum lower dead center is f _min, top dead centre is f _maxtime, there is f _min<f _m<f _max, and the speed of particle m is expressed as v _m, individual optimal value is p _m, i.e. individual optimum frequency, colony's optimal value is p _g, i.e. the optimum frequency of colony, the function of speed and location updating is respectively v _m(t+1)=Ω v _m(t)+c ₁r ₁(p _m-f _i)+c ₂r ₂(p _g-f _m), f _m(t+1)=f _m(t)+v _m(t+1), getting inertial coefficient Ω is 0.729, gets learning coefficient c ₁and c ₂be 1.4962, r ₁and r ₂for the random number of [0,1], the highest iterations T _maxbe 100 times, total number of particles gets 40, and fitness function arranges as follows, and first selection substrate is with to at frequency f _mphase velocity array [the v of each test block lower _p] _mthe mean grain size array G obtained with step one, carries out quadratic fit by least square method

Separate above-mentioned normal equation system, obtain choosing frequency f _mtime matched curve [S ^*(v _p)] _m=[a ₀] _m+ [a ₁] _mv _p+ [a ₂] _mv _p ², operational symbol () wherein represents asks two inner product of vectors, as

Therefore the square error after selection matching is as fitness F _m, to weigh the sensitivity under this frequency between phase velocity and grain size

According to fitness F _msearch for, the optimum frequency f making fitness reach minimum value can be found _p, the phase velocity v under this frequency _p| _fpthe highest to the sensitivity of grain size, finally depict each test block at optimum frequency f _punder phase velocity-grain size matched curve, and it can be used as the grain size evaluation model g based on single frequency phase velocity ₁(v _p| _fp);

Step 2: set up the grain size evaluation model based on the dispersion degree factor, the dispersion degree factor is represented with the slope of phase velocity spectrum, first to the phase velocity spectrum least square method of each test block, linear fit is done to phase velocity and frequency, try to achieve the slope Δ v of each test block phase velocity spectrum _p/ Δ f; Then to the phase velocity spectrum slope Δ v of each test block _p/ Δ f and average grain size number group G, does quadratic fit by least square method again, finally depicts each test block phase velocity spectrum slope-grain size curve, and it can be used as the grain size evaluation model g based on the dispersion degree factor ₂(Δ v _p/ Δ f);

Step 3: the grain size evaluation model based on single frequency phase velocity that integrating step 1 obtains, the grain size evaluation model based on the dispersion degree factor obtained with step 2, the comprehensive high-temperature alloy grain size evaluation model based on ultrasonic phase velocity can be obtained, realized by weighted mean

g(Q)＝w ₁·g ₁([v _P| _fP] _Q)+w ₂·g ₂([Δv _P|Δf] _Q)

In formula, Q represents Q alloy test block, [v _p| _fp] _qbe that Q test block is at frequency f _punder phase velocity, [Δ v _p/ Δ f] _qbe the dispersion degree factor of Q test block, w ₁the weight of the grain size evaluation model based on single frequency phase velocity, w ₂it is the weight of the grain size evaluation model based on single frequency phase velocity.

The CYS-1100 type 6-dof motion platform that this example adopts Shanghai Cytrix Electrical Technology Co., Ltd. to produce, adopt the PCI-9820 Data Acquisition Card of Taiwan Ling Hua, adopt the 5072PR type ultrasonic pulse generation/receiver of Olympus, and the circular high resolving power water logging ultrasonic longitudinal wave flat probe of I3-1008-R-SU type, the MetaServ250 type double plate polisher lapper of Buehler is used during metallographic examination, and the DM4000M type metaloscope of Leica.This example adopts a collection of trade mark to be that the alloy test block of In718 is tested, and as shown in Figure 7, as shown in Figure 8, the relevant parameters of test block is as shown in table 2 for metallographic examination system for test block pictorial diagram.

The relevant parameters of table 2 test block

Adopt the high-temperature alloy grain size evaluation method based on ultrasonic phase velocity of the present invention, first the A wave datum of five test blocks is gathered respectively, next step carries out metallographic examination, with metaloscope acquisition metallograph as shown in Figure 9, the crystal grain angle value obtained with division lines method is as shown in table 2, ask the longitudinal wave velocity of five test blocks again, next the attenuation coefficient spectrum of five test blocks is obtained respectively by two medium multivariate Gaussian sound-field models, then with the nominal center frequency 10MHz of ultrasonic probe I3-1008-R-SU for reference frequency point, the phase velocity at reference frequency point place is similar to and is taken as respective longitudinal wave velocity, obtain the phase velocity spectrum of five test blocks according to the phase velocity of reference frequency point by K-K relational expression, the phase velocity spectrum of five test blocks, as shown in Figure 10.

Finally obtain the grain size evaluation model based on single frequency phase velocity, as shown in figure 11, concrete expression formula is image

g ₁(v _P| _fP)＝824.4913(v _P| _fP) ²-9447.4144(v _P| _fP)+27064.6250

With the grain size evaluation model based on the dispersion degree factor, as shown in figure 12, concrete expression formula is image

g ₂(Δv _P|Δf)＝1.0894(Δv _P|Δf) ²-5.8256(Δv _P|Δf)+9.0174

Suitably weight selection, makes w ₁=0.2584764, w ₂=0.7415236, finally can set up the comprehensive high-temperature alloy grain size evaluation model based on ultrasonic phase velocity, concrete expression formula is

g(Q)＝0.2584764g ₁([v _P| _fP] _Q)+0.7415236g ₂([Δv _P|Δf] _Q)

Table 3 illustrates evaluation result and the error analysis of this example.

The evaluation result of this example of table 3 and error analysis

Claims (3)

1. based on a high-temperature alloy grain size evaluation method for ultrasonic phase velocity, it is characterized in that, comprise the following steps:

Step one: to the alloy test block needing to evaluate, measure the test block thickness L of position, measured point, then detected by pulse reflection method with supersonic detection device, position and the attitude of ultrasonic longitudinal wave probe need be adjusted when detecting, make axis and position, the measured point exact vertical of probe, original A wave datum is gone out by Detection and Extraction, an A wave datum array Awave_data stores, after the A wave datum of each alloy test block is all extracted, again metallographic examination is carried out to test block, the mean grain size in the multiple cross section of each test block is obtained with division lines method, be recorded in successively in array G, for the foundation of grain size evaluation model is prepared,

Step 2: the original A wave datum stored in the array Awave_data obtained by step one calculates longitudinal wave velocity v _gfirst manually gate 1, the reference position of gate 2 and gate length is adjusted, orient once ripple at the bottom of end ripple and secondary respectively, then apply rectangular window and intercept once wave number group BW2_data at the bottom of end wave number group BW1_data and secondary, again envelope extraction is carried out to array BW1_data and BW2_data Hilbert transform, and find forward first amplitude and be less than advanced position threshold value T the position that maximal value occurs from BW1_data and BW2_data _esampled point, obtain once the advanced position n of ripple at the bottom of end ripple, secondary ₁and n ₂, then sampling number Δ n=n apart between the advanced position of twice end ripple ₂-n ₁if the sampling time interval between two sampled points is dt, the test block thickness L recorded in integrating step one, according to formula v _g=2L/ (Δ ndt) calculates longitudinal wave velocity, repeatedly carries out this step to obtain the longitudinal wave velocity of all test blocks;

Step 3: the longitudinal wave velocity v that step 2 is obtained _g, and the variable relevant to the multivariate Gaussian sound-field model of two layer medium all substitutes in model, simulates the acoustic elasticity transport function t of surface wave and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω) the complex frequency spectrum V of the surface wave obtained by Fourier transform and once end ripple, is then combined _fW(ω) and V _bW1(ω), obtain attenuation coefficient spectrum α (ω), then according to K-K relational expression, in conjunction with attenuation coefficient spectrum α (ω), finally obtain phase velocity spectrum v _p(ω) this step, is repeatedly carried out to obtain the phase velocity spectrum of all test blocks;

Step 4: the contact between the mean grain size array G that each test block phase velocity spectrum obtained based on step 3 and step one obtain, sets up the high-temperature alloy grain size evaluation model based on ultrasonic phase velocity.

2. method according to claim 1, is characterized in that, the step obtaining phase velocity spectrum in step 3 is:

Step 1: the reference position and the gate length that again manually adjust gate 2, orients surface wave, then intercepts out surface wave array FW_data;

Step 2: in the multivariate Gaussian sound-field model of two layer medium, the one way distance that ultrasound wave is propagated in water and alloy test block two media is respectively: the one way distance between ultrasonic probe to check point position is that the underwater sound is apart from L _w, by bubble sort method, find out the position that in FW_data, maximum absolute value value occurs, remember that this position is n ₀, and underwater sound speed v _wknown, then L _w=v _w(n ₀-1) dt/2, and the one way distance that ultrasound wave is propagated in alloy test block is the thickness L of test block, the variable also needed in computing simultaneously has, the longitudinal wave velocity v obtained in step 2 _g, the wafer radius r of circular ultrasonic longitudinal wave flat probe, the density p of water _w, the density p of high temperature alloy, waits the frequency sequence array f simulating frequency range;

Step 3: set up polar coordinates at the detecting head surface of circular ultrasonic longitudinal wave flat probe, carries out ring network to detecting head surface and formats discrete, the radial distance from the center of circle to edge is divided into N ₁equal portions, and make the center of circle be annulus, then there is N ₁+ 1 annulus, this N ₁+ 1 the pole footpath of annulus under polar coordinates is ascending is stored in array r, then rotates a circle with ray angle is divided into N ₂equal portions, the polar angle of every bar ray is stored in array θ successively, then pop one's head in by discrete as N ₁n ₂individual ring-type unit dimension, two then adjacent to array r annular radii averagings, can obtain each ring-type infinitesimal center point pole footpath array r _mid, finally footpath, the pole r of each ring-type infinitesimal central point _midwith polar angle θ, be converted into the two-dimensional discrete that rectangular coordinate x and y completes circular probe surface, namely use orderly real number to (x _j, y _k), j=1,2 ..., N ₁, k=1,2 ..., N ₂, record the coordinate of each ring-type infinitesimal central point, because acoustic axis line is vertical with position, measured point, therefore with (x _j, y _k) represent the coordinate of different sound path places acoustic beam radial section, and subtract each other the area ds that can obtain each ring-type infinitesimal by adjacent two sectorial areas;

Step 4: substituted into by each variable in step 2 and 3 in the multivariate Gaussian sound-field model of two layer medium and carry out computing, simulates the acoustic elasticity transport function t of surface wave and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω);

Step 5: the acoustic elasticity transport function t of the surface wave obtained by step 4 and once end ripple _{a; FW}(ω) and t _{a; BW1}(ω), in conjunction with the complex frequency spectrum V of the surface wave obtained by Fourier transform and once end ripple _fW(ω) and V _bW1(ω) attenuation coefficient spectrum α (ω) is obtained;

Step 6: the centre frequency choosing ultrasonic probe used is reference frequency point ω ₀, and reference frequency point ω ₀the phase velocity v at place _p(ω ₀) the approximate longitudinal wave velocity v being taken as step 2 and obtaining _g, according to K-K relational expression, attenuation coefficient spectrum α (ω) obtained in integrating step 5, finally obtains phase velocity spectrum v _p(ω).

3. method according to claim 1, is characterized in that, the step set up in step 4 based on the high-temperature alloy grain size evaluation model of ultrasonic phase velocity comprises:

Step 1: set up the grain size evaluation model based on single frequency phase velocity, first selected frequency is determined, pass through particle cluster algorithm, the frequency Selection Strategy of research phase velocity, select the optimum frequency the highest to grain size sensitivity: the total number of particles that note uses is M, wherein m particle (m=1,2,, M) positional representation be f _m, f _mnamely current frequency, note phase velocity spectrum lower dead center is f _min, top dead centre is f _maxtime, there is f _min<f _m<f _max, and the speed of particle m is expressed as v _m, individual optimal value is p _m, i.e. individual optimum frequency, colony's optimal value is p _g, i.e. the optimum frequency of colony, the function of speed and location updating is respectively v _m(t+1)=Ω v _m(t)+c ₁r ₁(p _m-f _i)+c ₂r ₂(p _g-f _m), f _m(t+1)=f _m(t)+v _m(t+1), getting inertial coefficient Ω is 0.729, gets learning coefficient c ₁and c ₂be 1.4962, r ₁and r ₂for the random number of [0,1], the highest iterations T _maxbe 100 times, total number of particles gets 40, and fitness function arranges as follows, and first selection substrate is with to at frequency f _mphase velocity array [the v of each test block lower _p] _mthe mean grain size array G obtained with step one, carries out quadratic fit by least square method

Separate above-mentioned normal equation system, obtain choosing frequency f _mtime matched curve [S ^*(v _p)] _m=[a ₀] _m+ [a ₁] _mv _p+ [a ₂] _mv _p ², operational symbol () wherein represents asks two inner product of vectors, as

Therefore the square error after selection matching is as fitness F _m, to weigh the sensitivity under this frequency between phase velocity and grain size

According to fitness F _msearch for, the optimum frequency f making fitness reach minimum value can be found _p, the phase velocity v under this frequency _p| _fpthe highest to the sensitivity of grain size, finally depict each test block at optimum frequency f _punder phase velocity-grain size matched curve, and it can be used as the grain size evaluation model g based on single frequency phase velocity ₁(v _p| _fp);

Step 2: set up the grain size evaluation model based on the dispersion degree factor, the dispersion degree factor is represented with the slope of phase velocity spectrum, first to the phase velocity spectrum least square method of each test block, linear fit is done to phase velocity and frequency, try to achieve the slope Δ v of each test block phase velocity spectrum _p/ Δ f; Then to the phase velocity spectrum slope Δ v of each test block _p/ Δ f and average grain size number group G, does quadratic fit by least square method again, finally depicts each test block phase velocity spectrum slope-grain size curve, and it can be used as the grain size evaluation model g based on the dispersion degree factor ₂(Δ v _p/ Δ f);

Step 3: the grain size evaluation model based on single frequency phase velocity that integrating step 1 obtains, the grain size evaluation model based on the dispersion degree factor obtained with step 2, the comprehensive high-temperature alloy grain size evaluation model based on ultrasonic phase velocity can be obtained, realized by weighted mean

g(Q)＝w ₁·g ₁([v _P| _fP] _Q)+w ₂·g ₂([Δv _P|Δf] _Q)

In formula, Q represents Q alloy test block, [v _p| _fp] _qbe that Q test block is at frequency f _punder phase velocity, [Δ v _p/ Δ f] _qbe the dispersion degree factor of Q test block, w ₁the weight of the grain size evaluation model based on single frequency phase velocity, w ₂it is the weight of the grain size evaluation model based on single frequency phase velocity.