CN106250527A - Alloy designations recognition methods based on pearson correlation coefficient - Google Patents

Abstract

The invention discloses a kind of alloy designations recognition methods based on pearson correlation coefficient, the method determines coupling element set for (1)；(2) test sample constituent content information to be identified by spectrogrph, and extract each constituent content value X in coupling element set_i；(3) each element upper content limit X in certain trade mark in coupling element set is extracted_maxiValue, lower limit X_miniValue；(4) structure actual match vector V_realWith perfect match vector V_ideal: (5) calculate actual match vector V_realWith perfect match vector V_idealBetween pearson correlation coefficient；(6) step (3) (5) is repeated, until calculating the pearson correlation coefficient of all trades mark in trade mark storehouse, the affiliated trade mark that the trade mark is sample to be identified that correlation coefficient is maximum.The method calculating process is simple and operation result is normalization data, dimensionless, its numerical value circle is between 1 to 1, and value is 1 to show that this trade mark fits like a glove with this measurement result, in the case of other, closer to 1, numerical value shows that matching degree is the highest, closer to 1, numerical value shows that matching degree is the lowest.

Description

Alloy designations recognition methods based on pearson correlation coefficient

Technical field

The present invention relates to alloy detection field, particularly relate to a kind of alloy designations identification based on pearson correlation coefficient Method.

Background technology

Trade mark identification, refers to contain the actual measurement constituent content information of detected sample with the element of several trades mark in trade mark storehouse Weight range information contrasts, thus it is speculated that go out to detect the relevant information such as the trade mark of object.Trade mark identification is as sending out of alloy detection industry The New function that exhibition gradually derives, this function is very easy to identification and the classification of alloy.

Existing trade mark recognition methods kind is less, generally divides the most several: (1) is according to whether surveying constituent content value Being positioned at the content range that the trade mark specifies and define whether detection object is this trade mark, the method shortcoming is, if certain element That even if content exceeds beyond this scope and few trade mark identification that also can affect detection object.(2) method improved has, and will contain Weight range does suitably to amplify and defines whether detection object is this trade mark with multiple " scopes ".(1), (2) two class algorithms have above Stronger occasionality, the impact of the easy examined error of recognition result, it is difficult to debugging and safeguards.(3) board based on membership function Number recognizer, constructs some membership functions according to different situations, will survey constituent content in the range of degree quantify, The method has more outstanding accuracy and fault-tolerance, but needs in the adjustment of parameter to consider the compatibility between different situations membership function Property, the most not easy care.

Trade mark identification problem also has another form: only provide the trade mark identification problem of constituent content reference value.The most just It is that trade mark storehouse does not provide the element content range information of the trade mark only to provide constituent content reference value.Existing trade mark identification side Method is not for this problem.

Summary of the invention

For above-mentioned deficiency, the present invention provides a kind of alloy designations recognition methods based on pearson correlation coefficient, by phase Closing coefficient to be incorporated in trade mark identification problem, solve traditional algorithm and calculate the problems such as complicated, more difficult maintenance, the most prominent is The method is applicable not only to provide the trade mark identification problem of element content range to be also applied for only providing constituent content reference value Trade mark identification problem.

In order to achieve the above object, the technical solution used in the present invention is as follows: a kind of conjunction based on pearson correlation coefficient Gold medal recognition methods, it is characterised in that specifically include following steps:

(1) coupling element set is determined；

Taking certain alloy sample is sample to be identified, if trade mark storehouse has n the trade mark, and the respectively trade mark 1, the trade mark 2 ... board Number n；

Take the whole elements contained by each trade mark, constitute the element set contained by this trade mark, finally give A1, A2……An；

Take A1, A2 ... the union of An i.e. obtains the element set A having content in trade mark storehouse；

Take the common factor of the element set B having the element set A of content and spectrogrph to be capable of identify that in trade mark storehouse, obtain mating element Collection C；

(2) test sample constituent content information to be identified by spectrogrph, and extract each constituent content in coupling element set Value X_i, and then obtain actual measurement constituent content vector X, X=[X₁,X₂,…,X_n], wherein, n is the individual of element in coupling element set C Number, i is positive integer, 1≤i≤n；

(3) from the constituent content upper and lower limit information that trade mark storehouse is recorded, in extraction coupling element set, each element is at certain board Upper content limit X in number_maxiValue, lower limit X_miniValue；

(4) structure actual match vector V_realWith perfect match vector V_ideal:

(4.1) each element upper content limit X in certain trade mark in the coupling element set obtained in step (3) is taken_maxiValue, Lower limit X_miniValue and actual measurement constituent content value X of each element of acquisition in step (2)_iCompare, calculate the preferable unit of each element Cellulose content value v_i, such as formula (1)；

$\left\{\begin{array}{c}{v}_i=X_{\max i},(X_i\≥X_{\max i})\\ v_i=X_{\min i},(X_i\≤X_{\min i})\\ v_i=X_i,(X_{\min i}\≤X_i\≤X_{\max i})\end{array}\right.---\left(1\right)$

(4.2) element at infinity content value v of each element obtained according to step (4.1)_i, build element at infinity content vector V, such as formula (2):

V=[v₁,v₂,…,v_n] (2)

(4.3) according to formula (3) and formula (4), actual measurement constituent content vector X step (2) obtained and step (4.2) obtain Element at infinity content vector V merge, obtain actual match vector V_real；By element at infinity content vector V and himself Merge, obtain perfect match vector V_ideal:

V_real=[X, V] ' (3)

V_ideal=[V, V] ' (4)

(5) actual match vector V is calculated_realWith perfect match vector V_idealBetween pearson correlation coefficient:

Computing formula such as formula (5):

$r_p=\frac{\mathrm{cov}(V_1,V_2)}{{\mathrm{\σ}}_{V_1}\·{\mathrm{\σ}}_{V_2}}=\frac{E\left(V_1{V}_2\right)-E\left(V_1\right)E\left(V_2\right)}{\sqrt{E\left(V_1^2\right)-E^2\left(V_1\right)}\sqrt{E\left(V_2^2\right)-E^2\left(V_2\right)}}---\left(5\right)$

In formula: r_pFor pearson correlation coefficient；Cov is covariance；σ is standard deviation；E is expectation；

(6) step (3)-(5) are repeated, until calculating the pearson correlation coefficient of all trades mark in trade mark storehouse, then by it The descending sequence of numerical value, the trade mark made number one is the affiliated trade mark of sample to be identified.

Beneficial effects of the present invention is as follows:

1) using the statistical method of maturation, Computing Principle is simple, easily realizes；

2) it is applicable not only to provide the trade mark identification problem of element content range to be also applied for only providing constituent content reference The trade mark identification problem of value；

3) the method calculating process is simple and operation result is normalization data, dimensionless, its numerical value circle in-1 to 1 it Between, value is 1 to show that this trade mark fits like a glove with this measurement result, and in the case of other, closer to 1, numerical value shows that matching degree is the highest, Closer to-1, numerical value shows that matching degree is the lowest, this processes just to necessary follow-up data and provides conveniently.And in traditional method It is usually correlation values the biggest (or the least) and shows that matching effect is the best, and the size of numerical value does not has a fixing scope, Later stage further data process and are difficult to judge the difference between two data.

Accompanying drawing explanation

The present invention is described further with embodiment below in conjunction with the accompanying drawings；

Fig. 1 is the calculation flow chart of the present invention.

Detailed description of the invention

The present embodiment takes any alloy sample S as detected sample, and those skilled in the art are according to detected sample S's Substantially classification chooses certain trade mark storehouse comprising the trade mark belonging to it (if known S is that steel alloy then selects steel alloy trade mark storehouse；As in more detail Ground, it is known that S is that steel alloy subordinate's special-purpose steel then selects this special-purpose steel trade mark storehouse).In the present embodiment, S is certain special-purpose steel, therefore institute Selecting trade mark storehouse is this special-purpose steel trade mark storehouse, has 3 trades mark, respectively P1, P2, P3 in trade mark storehouse.

(1) coupling element set is determined

The acquisition methods having the element set of content in trade mark storehouse is:

1. take the whole elements contained by each trade mark, constitute the element set contained by this trade mark, finally give A1, A2、A3；

2. the union taking A1, A2, A3 i.e. obtains the element set A having content in trade mark storehouse, has and contain in the present embodiment in trade mark storehouse Amount element set A be: 13 (Al), 14 (Si), 15 (P), 16 (S), 22 (Ti), 23 (V), 24 (Cr), 25 (Mn), 26 (Fe), 27 (Co), 28 (Ni), 29 (Cu), 40 (Zr), 41 (Nb), 42 (Mo), 46 (Pd), 72 (Hf), 73 (Ta), 74 (W), totally 19 kinds of units Element；

The element set B that the present embodiment Instrumental is capable of identify that: from No. 22 element tis to No. 92 element U, totally 71 kinds of elements；

Take the common factor of A and B, obtain mate element set C:22 (Ti), 23 (V), 24 (Cr), 25 (Mn), 26 (Fe), 27 (Co), 28 (Ni), 29 (Cu), 40 (Zr), 41 (Nb), 42 (Mo), 46 (Pd), 72 (Hf), 73 (Ta), 74 (W), totally 15 kinds of units Element；

Owing to Fe is " surplus " element (its content range does not does in the trade mark a kind of element of clear stipulaties), in trade mark storehouse Data deficiencies thus it is not done and mates, in like manner, detecting alloy aluminum or during alloyed copper, if database data the most entirely can not to Al and Cu mates.Otherwise, if " surplus " element can also be added coupling element set by all information, participate in pearson phase relation Number calculates.Therefore in this example, final coupling element set is: 22 (Ti), 23 (V), 24 (Cr), 25 (Mn), 27 (Co), 28 (Ni), 29 (Cu), 40 (Zr), 41 (Nb), 42 (Mo), 46 (Pd), 72 (Hf), 73 (Ta), 74 (W), totally 14 kinds of elements；

(2) test sample constituent content information to be identified by spectrogrph, and extract each constituent content in coupling element set Value X_i, thus obtain actual measurement constituent content vector X, X=[X₁,X₂,…,X_n]；Wherein, during n is coupling element set C, element is individual Number, i is positive integer, 1≤i≤n；

(3) from the constituent content upper and lower limit information that trade mark storehouse is recorded, in extraction coupling element set, each element is at certain board Upper content limit X in number_maxiValue, lower limit X_miniValue；Constituent content upper and lower limit information is according to national standard；Now initial data is received Collecting complete, details are shown in Table 1:

(4) structure actual match vector V_realWith perfect match vector V_ideal:

(4.1) each element is taken in the coupling element set obtained in step (3) in certain trade mark (the most first taking trade mark P1) Upper content limit X_maxiValue, lower limit X_miniValue and actual measurement constituent content value X of each element of acquisition in step (2)_iCompare, Calculate element at infinity content value v of each element_i, such as formula (1)；

$\left\{\begin{array}{c}{v}_i=X_{\max i},(X_i\≥X_{\max i})\\ v_i=X_{\min i},(X_i\≤X_{\min i})\\ v_i=X_i,(X_{\min i}\≤X_i\≤X_{\max i})\end{array}\right.---\left(1\right)$

(4.2) element at infinity content value v of each element obtained according to step (4.1)_i, build element at infinity content vector V, such as formula (2):

V=[v₁,v₂,…,v_n] (2)

(4.3) according to formula (3) and formula (4), actual measurement constituent content vector X step (2) obtained and step (4.2) obtain Element at infinity content vector V merge, obtain actual match vector V_real；By element at infinity content vector V and himself Merge, obtain perfect match vector V_ideal:

V_real=[X, V] ' (3)

V_ideal=[V, V] ' (4)

(5) actual match vector V is calculated_realWith perfect match vector V_idealBetween pearson correlation coefficient:

Computing formula such as formula (5):

$r_p=\frac{\mathrm{cov}(V_1,V_2)}{{\mathrm{\σ}}_{V_1}\·{\mathrm{\σ}}_{V_2}}=\frac{E\left(V_1{V}_2\right)-E\left(V_1\right)E\left(V_2\right)}{\sqrt{E\left(V_1^2\right)-E^2\left(V_1\right)}\sqrt{E\left(V_2^2\right)-E^2\left(V_2\right)}}---\left(5\right)$

In formula: r_pFor pearson correlation coefficient；Cov is covariance；σ is standard deviation；E is expectation；

(6) step (3)-(5) are repeated, until calculating the pearson correlation coefficient of all trades mark in trade mark storehouse, then by it The descending sequence of numerical value, it is clear that the trade mark of detected sample S be most possibly P1, secondly for P2, P3, as shown in table 2:

Calculating process of the present invention is simple and operation result is normalization data, dimensionless, its numerical value circle between-1 to 1, Value is 1 to show that this trade mark fits like a glove with this measurement result, and in the case of other, closer to 1, numerical value shows that matching degree is the highest, number Closer to-1, value shows that matching degree is the lowest, this processes just to necessary follow-up data and provides conveniently.And in traditional method one As be that correlation values the biggest (or the least) shows that matching effect is the best, and the size of numerical value does not has a fixing scope, after Phase further data process and are difficult to judge the difference between two data.

Claims (1)

1. an alloy designations recognition methods based on pearson correlation coefficient, it is characterised in that specifically include following steps:

(1) coupling element set is determined；

Taking certain alloy sample is sample to be identified, if trade mark storehouse has n the trade mark, and the respectively trade mark 1, the trade mark 2 ... trade mark n；

Take the whole elements contained by each trade mark, constitute the element set contained by this trade mark, finally give A1, A2 ... An；

Take A1, A2 ... the union of An i.e. obtains the element set A having content in trade mark storehouse；

Take the common factor of the element set B having the element set A of content and spectrogrph to be capable of identify that in trade mark storehouse, obtain mating element set C.

(2) test sample constituent content information to be identified by spectrogrph, and extract each constituent content value X in coupling element set_i, And then obtain actual measurement constituent content vector X, X=[X₁,X₂,…,X_n], wherein, n is the number of element in coupling element set C, and i is Positive integer, 1≤i≤n.

(3) from the constituent content upper and lower limit information that trade mark storehouse is recorded, in extraction coupling element set, each element is in certain trade mark Upper content limit X_maxiValue, lower limit X_miniValue.

(4) structure actual match vector V_realWith perfect match vector V_ideal:

(4.1) each element upper content limit X in certain trade mark in the coupling element set obtained in step (3) is taken_maxiValue, lower limit X_miniValue and actual measurement constituent content value X of each element of acquisition in step (2)_iComparing, the element at infinity calculating each element contains Value v_i, such as formula (1)；

$\left\{\begin{array}{c}{v}_i=X_{\max i},(X_i\≥X_{\max i})\\ v_i=X_{\min i},(X_i\≤X_{\min i})\\ v_i=X_i,(X_{\min i}\≤X_i\≤X_{\max i})\end{array}\right.---\left(1\right)$

(4.2) element at infinity content value v of each element obtained according to step (4.1)_i, build element at infinity content vector V, as Formula (2):

V=[v₁,v₂,…,v_n] (2)

(4.3) according to formula (3) and formula (4), the reason that actual measurement constituent content vector X step (2) obtained and step (4.2) obtain Think that constituent content vector V merges, obtain actual match vector V_real；Element at infinity content vector V is carried out with himself Merge, obtain perfect match vector V_ideal:

V_real=[X, V] ' (3)

V_ideal=[V, V] ' (4)

(5) actual match vector V is calculated_realWith perfect match vector V_idealBetween pearson correlation coefficient: computing formula such as formula (5):

$r_p=\frac{\mathrm{cov}(V_1,V_2)}{{\mathrm{\σ}}_{V_1}\·{\mathrm{\σ}}_{V_2}}=\frac{E\left(V_1{V}_2\right)-E\left(V_1\right)E\left(V_2\right)}{\sqrt{E\left(V_1^2\right)-E^2\left(V_1\right)}\sqrt{E\left(V_2^2\right)-E^2\left(V_2\right)}}---\left(5\right)$

In formula: r_pFor pearson correlation coefficient；Cov is covariance；σ is standard deviation；E is expectation.

(6) step (3)-(5) are repeated, until calculating the pearson correlation coefficient of all trades mark in trade mark storehouse, then by its numerical value Descending sequence, the trade mark made number one is the affiliated trade mark of sample to be identified.