CN101315644B - Part classification method based on developable clustering - Google Patents

Abstract

The invention relates to a parts classification method based on expansible cluster, which comprises the following steps: 1) the node domain of the structural characteristic of parts is extracted; 2) the characteristic variable numerical value v and the interval V of the parts are determined; 3) the distance formula of each part structure is established; 4) the structural relevancy T(i, j) of parts Ri and Rj is determined; 5) numerical analysis is carried out to a symmetric matrix M; 6) the average relevancy numerical value of cluster analysis is established; 7) the n minus 2 symmetric matrix is obtained by utilizing the algorithm process of step 5) to the newly formed n minus 1 symmetric matrix, the processes are circulated, and when the minimum relevancy is more than K value, the cluster process is terminated and a unique cluster result is obtained: the category of products and the parts of each category. The invention provides a part classification method based on the expansible cluster, which has clear relevancy analysis, high classification efficiency and high accuracy.

Description

Based on the part classification method that can open up cluster

Technical field

The present invention relates to a kind of sorting technique of part product.

Background technology

The notion of cluster is not a new notion, in building trade early in the twentieth century, building is just existed according to the notion that function is divided into building unit that can independent assortment, and building block is at this moment emphasized can realize on physical dimension to connect and exchange.Then, the notion of cluster is introduced into machinery manufacturing industry, people further together with the functional cohesion of institute's sub-module and physical product, module has had clear and definite functional definition feature, how much connecting interfaces, and function input, output interface feature.Development along with computer software technology, the notion of module has been used to non-physical product field again, notion in software industry module is being put into practice widely, and the modularization trend of large-scale software systems (such as the Windchill system of PTC S. A.) is more and more obvious.

The part classification technology is one of basic technology of product family's configuration design, is supporting product family's design and configuration design by forming effective dispensing unit.Traditional sorting technique is conceived to the user, and emphasis is the standardization of parts, and the component no longer of certain function is single parts, but some row performances are different with structure but interchangeable part that function is identical, and the user can assemble as required voluntarily.To decompose and to be combined as the basis, emphasize the interchangeability and the versatility of dispensing unit towards the sorting technique of large-scale customization.It has broken the viewpoint of " module is exactly universalization, the standardization of parts ", extensively take cad technique, make every effort to finish the existing design module division methods of product as much as possible with minimum module combinations and mainly comprise two classes: a class is a function analysis method, promptly divides design module according to the subfunction of product and mutual relation thereof; Another kind of is the part analysis method, promptly divides module according to the part of product and dependence each other thereof.The former emphasizes the functional independence of module, generally is used for demand analysis and conceptual phase.The latter considers the geometry and the physical independence of module, and be applied in the design of general industry product family keeping under the constant situation of basic function more, by the coupling or the different properties of product of modification acquisition of personalized part, to satisfy different customer demands.In the case, adopt the part analysis method to carry out design module and divide, the variation of part and how much of the functions between the part and physical interconnection are unified the high-level efficiency that consideration can guarantee to dispose restructuring procedure better.

By the Product Configuration Model generative theory as can be known, constraints such as function in the design and behavior can not be totally independent of structure and exist, and architectural feature also is the carrier of customer demand simultaneously, and therefore whole configuration module should launch around product family's inner structural features.Again because of client's the diversification of demand, obfuscation and the characteristics that are difficult to accurate expression, make the degree of association of shining upon between the structure reduce, its complicated contact is difficult to semantic statement, or even can't explain, and this will directly affect the accuracy and the customer satisfaction of configuration result.The product parts classification has following main standard:

(1) by designed product family and employed process variant, is combined into required product;

(2) in design with in making higher reusability is arranged, produce in batches, reduce cost so that form;

(3) suitable granularity of division so that combination and forming, does not increase the workload of part management again;

(4) consider economy and the convenience of designed module in product lifecycle (comprising design, manufacturing, sale, maintenance etc.).

In general, part classification should be followed following standard:

(1) keep classification element to have certain independence and integrality in function and configuration aspects;

(2) degree of correlation between classification element and the classification element is as far as possible little, i.e. weak coupling, and the degree of correlation between the inside modules will be tried one's best greatly, i.e. strong coupling;

(3) granularity of part decomposition is moderate, and the granularity of decomposition is too big, and the degree of coupling between the part must increase; It is more little to decompose granularity, causes division too scrappy, poor operability;

(4) part classification will be avoided one-sidedness and the blindness brought because of designer's subjective experience from the systematic analysis angle.

Cluster analysis is a kind of method of fuzzy problem in the reality being carried out classification analysis.Though it is cluster analysis has the application of a lot of data processing in practice, still rare at present in the research in product development field.Use that cluster thought is bottom-up establishes the effective ways that the entire product design platform is present research and design platform, a lot of experts and scholar study this.Erik J.Zamirowski[1] set up unified product family functional structure chart according to customer demand and product purpose, and function is carried out the similarity cluster according to function rule and variation rule, to identify product family's general module; William L.Moore, Jordan J[2] proposed to utilize the Conjoint Analysis theory to carry out the versatility analysis of product attribute and property value, and carry out the method for product family's design; Mitchell M.Tseng and Jianxin Jiao[3] etc. on the Axiomatic Design basis, utilize cluster analysis to realize that the module in the design of electronic products extracts; Wang Aimin [4] etc. has proposed the planning proposal of product family's constituent part based on the clustering method of the association analysis of process route and corresponding manufacturing cell.

A lot of scholars begins to carry out from quantitative aspect part and divides, and promptly in the product structure design space, by the degree of association between the definition part, and then the similar units cluster is divided in the same module.Using general clustering method mainly has the following disadvantages part:

(1), increased the uncertainty of classification by organizing the qualitative definite threshold value of expert;

(2) for the expression and the expansion of the degree of association, main method has following two kinds: the degree of association between each parts of definition product is several grades, according to the tightness degree that constitutes product parts, the degree of association between the artificial establishment part.For example, the product parts relationship type is defined as { extremely strong, intimate, moderate, general,, do not have six grades, each grade corresponding numerical value respectively is { 1.0,0.8,0.6,0.4,0.2,0}, the incidence relation between the part can be represented with table 1 like this, table 1 parts incidence relation.

Table 1

About the degree of association between the product parts, obtain the part incidence matrix according to above-mentioned.Divide incidence matrix by the definition threshold values, the degree of association of finishing between part cluster division detail of construction is sensitivity.So-called sensitivity is the numerical value that is caused other part variation by the variation of side's part.The change amount is little, and then sensitivity is little, otherwise also as the same.For the numerical value of sensitivity, then the same with situation in the said method, must manually in advance determine.For example for the deisgn product part, the definition sensitivity number is as shown in table 2, table 2 part incidence relation.

Table 2

The mutual numerical value of establishing between two parts is the basis of its sensitivity number, and in fact, sensitivity is the expansion of the degree of association, and main difference is to adopt sensitivity analysis, can see that the matrix of formation is also asymmetric, and this point is different with above-mentioned first method situation.Other processes are basic identical.(3) to classify also be the difficult point that part is divided to, the uncertain part of internal relations many for quantity.Present sorting technique mainly is at single feature, and in fact from the different physical functions of part, it has multiple characteristics, and under the therefore different characteristic of divisions, its classification results has a great difference, and this will directly influence classification effectiveness and precision.

Summary of the invention

For the correlation analysis that overcomes existing part product classification method is fuzzy, threshold value determines to be difficult to quantification, characteristic of division is single, classification effectiveness and the low deficiency of precision, the invention provides that a kind of correlation analysis is clear, classification effectiveness is high, precision is high based on the part classification method that can open up cluster.

The technical solution adopted for the present invention to solve the technical problems is:

A kind of based on the part classification method that can open up cluster, described part classification method may further comprise the steps:

1) extracts design of part characteristic node territory: establish and treat the sort product set A _n={ A ₁, A ₂..., A _n, i product A wherein _i={ R _1m..., R _Rm(1)

This model representation product A _iBy r part R _iForm, wherein part max architecture characteristic number is m, can use v _Rm=c (R _Rm) expression is used for the structure value of cluster analysis;

For i the architectural feature v that can generate profit maximized part _iSetting up its matter-element model is:

$R_i=\left[\begin{array}{ccc}{N}_i,& c_{i1},& v_{i1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{im}},& v_{\mathrm{im}}\end{array}\right]---\left(2\right)$

Im represents part R _iThe feature number;

For R _i,, obtain A by determining the constant interval of a certain architectural feature of part _iEigenwert nodes domains V=＜v _Min, v _Max, obtain the interval nodes domains X of character numerical value according to formula (2) _Njm=＜V _Min, V _Max＞, setting up the nodes domains model is R _Im:

$R_{\mathrm{mi}}=\left[\begin{array}{ccc}{N}_m,& c_{m1},& X_{m1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{mi}}& X_{\mathrm{mi}}\end{array}\right]---\left(3\right)$

X wherein _Nim(k=1 ..., mi) expression part R _iN character numerical value interval.

2) determine architectural feature variable numerical value v: generate profit maximized architectural feature trend based on the existing procucts on the market, utilize mid point gradually approximate algorithm obtain part optimum structure character numerical value v _i

When the requirement to architectural feature c belongs to more little dominance more, then

$v_{i+1}=\frac{v_i+v_{\mathrm{<figure-callout\; id="2"\; label="mim"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">mim</figure-callout>}}}{2},i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;---\left(4\right)$

$v_1=\frac{v_{\min }+v_{\mathrm{<figure-callout\; id="2"\; label="max"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">max</figure-callout>}}}{2}$

When the requirement to architectural feature c belongs to big more dominance more, then

$v_{i+1}=\frac{v_i+{\mathrm{<figure-callout\; id="2"\; label="v\; max"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">v</figure-callout>}}_{\max }}{2},i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;---\left(5\right)$

$v_1=\frac{v_{\min }+v_{\mathrm{<figure-callout\; id="2"\; label="max"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">max</figure-callout>}}}{2}$

3) set up each design of part apart from formula: according to formula (3) (4) and (5), calculate R _iDistance is worth the correlation degree ρ of its nodes domains:

$\mathrm{\&rho;}(v,v_i,X)=|v-v_i|-\frac{v_{\max }-v_{\min }}{2}---\left(6\right)$

The definition structure feature is apart from value: $K(R_i,R_{\mathrm{im}})=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i\frac{v_{\max }-\mathrm{\&rho;}(v,v_i,X)}{v_{\max }-v_{\min }}---\left(7\right)$

Apart from value K (R _i, R _Im) represented R _iThe related numerical value of architectural feature c, ω obtains by analytical hierarchy process, the weight of representation feature element, $\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i=1;$

4) determine R _i, R _jStructure connection degree T (i, j):

T(i，j)＝|K(R _i，R _im)-K(R _j，R _jm)| (8)

Wherein T (i, j)=T (j i), follows same computing method, for n architectural feature of deisgn product, calculate respectively T (i, j) (i=1 ..., n; J=1 ..., n), they constitute symmetric matrix M:

T (i, j)=0, or T (i j) levels off to 0, R _i, R _jBetween correlation degree big more; T (i, j)＞0; In this case, and T (i, value j) is big more, and expression correlation degree between the two is more little;

5) symmetric matrix M is carried out numerical analysis, because each row is represented the degree of association between a part and other n-1 the part, and T (i, i)=0 (i=1 ..., n), choose the minimum value in the matrix M, be made as T (p, q)=min{T (i, j), j=1 ..., n; I=1 ..., n; I ≠ j}, i.e. part R _p, R _qBetween the degree of association be minimum; R _p, R _qPut under in the same class, be designated as { p, q}, part R _p, R _qForm new new parts R _Pq, for any R _i, it and R _PqBetween the degree of association be: T (i, pq)=min{T (i, p), R (i, q) };

When n＞2, it is capable and q is capable to scratch p in the matrix M, adds R _Pq, they form a new symmetric matrix jointly, and its exponent number is n-1;

6) the average degree of association numerical value of establishment cluster analysis:

Set the average degree of association $K=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}T(i,j)}{{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">n</figure-callout>}}^2};$

7) the symmetrical incidence matrix on the n-1 rank of new formation is utilized algorithmic procedure in the step 5), obtain the symmetric matrix on n-2 rank, the circulation said process, when the minimum degree of association during greater than the K value, then stop cluster process, obtain unique cluster result: the classification of part, and the part of each classification.

Technical conceive of the present invention is: can open up is the new branch of science that China's Mainland scholar Cai Wenxian is born in the nineteen eighty-three foundation, through the exploration of many experts and scholars in theory and practice, formed featured theoretical frame at present, mainly be: 1, set up and to have opened up set and correlation function; 2, set up and to have opened up logic; 3, studied and changed contradictory problems for the basic skills of contradictory problems not etc.

From the degree of association that makes up element from the distance formula, by extracting the characteristic node territory, so establish cluster element apart from value.The square value has been represented the comprehensive related intensity of variation of element.By setting up the degree of association, use minimum degree of association method to the element cluster analysis apart from value.Establish product platform by the total of classification apart from value at last.Its building process meets the design sense of product platform, the designing requirement of having satisfied product platform that the implication of square value, the minimum degree of association is lucky.And the whole process of setting up is simpler and have efficient.

Beneficial effect of the present invention mainly shows: (1) quantification is determined the design of part degree of association; (2) human intervention of minimizing cluster threshold value; (3) help the letter sorting of many features design of part feature; (4) simple and convenient, be easy to computer system and realize; (5), classification effectiveness height, precision height.

Description of drawings

Fig. 1 is based on the process flow diagram of the part classification method that can open up cluster.

Fig. 2 is part cluster analysis figure.

Fig. 3 is the system design process flow diagram of embodiment 3.

Embodiment

Below in conjunction with accompanying drawing the present invention is further described.

Embodiment 1

With reference to Fig. 1, a kind of based on the design of part sorting technique that can open up cluster, from the distance formula, the square value has been represented the comprehensive related intensity of variation of element.By setting up the degree of association, use minimum degree of association method to the structural element cluster analysis apart from value.Mainly comprise two steps, its flow process is seen Fig. 1:

The first step: set up correlation matrix between part

1) extracts design of part characteristic node territory.

If treat the sort product set A _n={ R ₁, R ₂..., R _n,

I product wherein $R_i=\{\left[\begin{array}{ccc}{N}_1,& c_{11},& v_{11}\\ & .& .\\ & .& .\\ & .& .\\ & c_{1m},& v_{1m}\end{array}\right],\&CenterDot;\&CenterDot;\&CenterDot;,\left[\begin{array}{ccc}{N}_r,& c_{r1},& v_{r1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{rm}},& v_{\mathrm{rm}}\end{array}\right]\}---\left(1\right)$

This model representation product R _iBy r part R _iForm, wherein part max architecture characteristic number is m, can use v _Nrm=c (R _Njm) expression is used for the structure value of cluster analysis;

For i the architectural feature v that can generate profit maximized part _iSetting up its matter-element model is:

$R_i=\left[\begin{array}{ccc}{N}_i,& c_{i1},& v_{i1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{im}},& v_{\mathrm{im}}\end{array}\right]---\left(2\right)$

Im represents part R _iThe feature number;

For R _i,, obtain A by determining the constant interval of a certain architectural feature of part _iEigenwert nodes domains V=＜v _Min, v _Max, obtain the interval nodes domains X of character numerical value according to formula (2) _Njm=＜V _Min, V _Max, setting up the nodes domains model is R _Im:

$R_{\mathrm{mi}}=\left[\begin{array}{ccc}{N}_m,& c_{m1},& X_{m1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{mi}}& X_{\mathrm{mi}}\end{array}\right]---\left(3\right)$

X wherein _Njm(k=1 ..., mi) expression part R _iN character numerical value interval.2) determine design of part characteristic variable numerical value v and interval V.

By determining the constant interval of a certain architectural feature of part, obtain A _nEigenwert nodes domains V=＜v _Min, v _Max, generate profit maximized architectural feature trend by the existing procucts of analyzing on the market, utilize mid point gradually approximate algorithm obtain part optimum structure character numerical value v _i

When the requirement to architectural feature c belongs to more little dominance more, then

$v_{i+1}=\frac{v_i+{\mathrm{<figure-callout\; id="2"\; label="v\; mim"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">v</figure-callout>}}_{\mathrm{mim}}}{2},i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;---\left(4\right)$

$v_1=\frac{v_{\min }+{\mathrm{<figure-callout\; id="2"\; label="v\; max"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">v</figure-callout>}}_{\max }}{2}$

When the requirement to feature c belongs to big more dominance more, then

$v_{i+1}=\frac{v_i+{\mathrm{<figure-callout\; id="2"\; label="v\; max"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">v</figure-callout>}}_{\max }}{2},i=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;---\left(5\right)$

$v_1=\frac{v_{\min }+{\mathrm{<figure-callout\; id="2"\; label="v\; max"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">v</figure-callout>}}_{\max }}{2}$

3) set up each design of part apart from formula.

According to formula (3), (4), (5) 6), calculate R _iDistance is worth the correlation degree ρ of its nodes domains:

$\mathrm{\&rho;}(v,v_i,X)=|v-v_i|-\frac{v_{\max }-v_{\min }}{2}---\left(6\right)$

The definition part is apart from value: $K(R_i,R_{\mathrm{im}})=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i\frac{v_{\max }-\mathrm{\&rho;}(v,v_i,X)}{v_{\max }-v_{\min }}---\left(7\right)$

Apart from value K (R _i, R _Im) expression R _iThe related numerical value of architectural feature c, ω obtains by analytical hierarchy process, the weight of representation feature element, $\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i=1.$

4) determine part R _i, R _jStructure connection degree T (i, j):

T(i，j)＝|K(R _i，R _im)-K(R _j，R _jm)| (8)

Wherein T (i, j)=T (j, i).Follow same principle, for n architectural feature of deisgn product, T (i, j) (i=1 ..., n; J=1 ..., n) can calculate respectively, they constitute symmetric matrix M.

Because M is symmetric matrix, promptly available its last triangle or time triangle represented global matrix.After seeking symmetric matrix M, for each part R _i(i=1 ..., n), seek about its degree of association with n-1 the part that is left.According to T (i, implication j), the minimum degree of association have been described the correlation degree between two parts, the correlation degree between its value condition and two parts is closely related.T (i, j)=0, or T (i j) levels off to 0, part R _i, R _jBetween correlation degree big more.T (i, j)＞0; In this case, and T (i, value j) is big more, and expression correlation degree between the two is more little.

Second step: utilize minimum degree of association method classification

5) symmetric matrix M is carried out numerical analysis, because each row is represented the degree of association between a part and other n-1 the part individuality, and T (i, i)=0 (i=1 ..., n).Choose the minimum value in the matrix M, be made as T (p, q)=min{T (i, j), j=1 ..., n; I=1 ..., n; I ≠ j}, i.e. part R _p, R _qBetween the degree of association be minimum.R _p, R _qPut under in the same class, be designated as { p, q}, part R _p, R _qForm new part R _PqFor any R _i, it and new parts R _PqBetween the degree of association be: T (i, pq)=min{T (i, p), T (i, q) };

When n＞2, it is capable and q is capable to scratch p in the matrix M, adds R _Pq, they form a new symmetric matrix jointly, and its exponent number is n-1.

6) the average degree of association numerical value of establishment cluster analysis:

Set the average degree of association $K=\frac{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}T(i,j)}{{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="S2008100623488E00012.png"\; state="\{\{state\}\}">n</figure-callout>}}^2},$ In case when the minimum correlation distance of selecting surpassed the K value, then cluster process stopped.

7) symmetric matrix on the n-1 rank of new formation is utilized algorithmic procedure in the step 5), obtain the symmetric matrix on n-2 rank, the circulation said process.Because it is big that the minimum degree of association constantly becomes, and when finally greater than the K value, then stops cluster process.Obtain unique cluster result: the classification of part, and the part of each classification.

Embodiment 2

See figures.1.and.2, a kind of based on the design of part sorting technique that can open up cluster, the treatment scheme of present embodiment is identical with embodiment 1.This method mainly may further comprise the steps: at above-mentioned about opening up the theoretical analysis of cluster, here the design of electric drill product family is carried out opening up cluster analysis, in the structural design space, setting electric drill basic composition structure is respectively: drill chuck, shell, auger spindle, fan, motor, bearing, these 7 parts of gear are formed.According to the correlation theory that can open up clustering method, at demand, analyze existing deisgn product, extract the characteristic of each part, according to formula (1) (5), set up the nodes domains matter-element model R of part _Mi(i=1 ..., 7).If the correlated parts characteristic node territory numerical value that obtains is:

According to the above-mentioned relevant optimized individual v that chooses _iTheoretical analysis and the requirement of formula (3), any class that is chosen in the design value interval has designed the analysis that the electric drill of finishing is done the degree of association.Theoretical according to the above analysis, what select here is at general customer base demand, in designing the electric drill product, and the electric drill product individuality of profit on sales the best.Establishing design platform on based on such product category, also is to meet the principle that design platform is set up.At above-mentioned part feature, in design knowledge base, read a certain electric drill part feature numerical value that satisfies condition.Shown in table 3.The a certain electric drill product feature of table 3 numerical value

Table 3

The order of the characteristic parameter in the above-mentioned table 3 is consistent with the part nodes domains of setting up early stage.Wherein, drill chuck, motor, fan, gear have 3 features, and auger spindle, bearing, shell have 2 features.Data value is the character numerical value of critical part (only having chosen above-mentioned 7 parts) in this electric drill design process, certainly, is diversified for the selection of this class best product individuality, only chooses the product individuality of a certain class profit the best here.Finish in foundation on the basis of best product personal feature numerical value, in conjunction with the character numerical value nodes domains of above-mentioned foundation.Application of formula (6), calculate part apart from formula value ρ (v, v _i, X) (i=1 ..., 7), result of calculation sees Table 4.Table 4 design of part feature apart from value table (i=1 ..., 7)

Table 4

The feature of corresponding above-mentioned each part about the electric drill product, the utilization analytical hierarchy process obtains the weighted value of each attribute of component feature, as table 5.Each feature weight value of table 5 part.

Table 5

Foundation is finished after each weight of product parts, in conjunction with the part feature nodes domains numerical value interval of above-mentioned foundation, according to formula (7), calculates the distance value K (R of each part _i, R _Im) (i=1 ..., 7).Result of calculation can see Table 6.Table 6 part each apart from value.

Table 6

Calculate apart from value, for 7 parts of deisgn product, T (i, j) (i=1 ..., 7; J=1 ..., 7) can calculate respectively.Obtain symmetrical incidence matrix M.As follows, numeral 1 ..., 7 corresponding dash numbers of 7 expressions:

$\begin{array}{ccccccc}1& 2& 3& 4& 5& 6& 7\end{array}$

$M=\left[\begin{array}{ccccccc}0.000& 0.284& 0.141& 0.109& 0.186& 0.171& 0.349\\ & 0.000& 0.143& 0.175& 0.098& 0.113& 0.633\\ & & 0.000& 0.032& 0.045& 0.030& 0.490\\ & & & 0.000& 0.077& 0.062& 0.458\\ & & & & 0.000& 0.015& 0.535\\ & & & & & 0.000& 0.520\\ & & & & & & 0.000\end{array}\right]$

Each degree of association numerical value according to incidence matrix M obtains K=0.095 according to above-mentioned computing method about the average degree of association.It is T (5 that the analyzing and associating matrix M obtains the minimum degree of association, 6)=0.015, like this, should be part R5, R6 is divided into a class, and promptly part bearing and auger spindle should belong in the same class, and both individualities just are combined into new individuality and are made as R56, calculate the degree of association between each individual and new individuality, form new incidence matrix M (1).

Seeking to obtain the 2nd the related numerical value of minimum in new matrix is T (3,56)=0.03, and like this, part fan and auger spindle, bearing are combined into new combination individuality, should all divide fan, auger spindle, bearing in same class.The above-mentioned algorithm of cycle applications.Obtain corresponding minimum degree of association value and be followed successively by 0.015,0.03,0.032, this stylish symmetric matrix is:

Under this state of i.e. expression elder generation, individual R ₃, R ₄, R ₅, R ₆In same class, individual R ₁, R ₇, R ₂Also constitute a class by itself.Can continue to apply above-mentioned cluster process, obtaining the minimum degree of association successively is 0.098,0.109,0.349.But when minimum correlation distance was 0.098, because 0.098 numerical value has surpassed average degree of association K=0.095, cluster process stopped.Should select minimum correlation distance is cluster result under 0.032 state.Set out by original symmetric matrix like this, the dendrogram that obtains as shown in Figure 2.Use 1,2 respectively ..., 7 represent 7 design elements that constitute it.

According to the division result of Fig. 2, can see that the classification of division should be 4 classes, and the individuality in each class is A ₁={ R ₃, R ₄, R ₅, R ₆, A ₂={ R ₂, L ₃={ R ₁, A ₃={ R ₇.What application was maximum in the establishment market is auger spindle, fan, gear, bearing, and the part of customization part should be drill chuck, shell and motor, and on the basis of forming as product platform based on these four kinds of parts, development family releases market.

Embodiment 3

With reference to Fig. 1, Fig. 2 and Fig. 3, and the method among the embodiment 1, a kind of based on the design of part sorting technique that can open up cluster, use corresponding product design modeling software, set up the prototype system of having developed electric drill product family design platform.The system design flow process is seen Fig. 3.

System development background and running environment: the native system algorithm realizes that the main VC6.0++ of employing finishes.As the main flow SDK (Software Development Kit) that Microsoft provides, VC6.0++ is that range of application or execution efficient all have impayable superiority.Realization for engineering drawing, select UG NX4.0 as Software Development Platform, can make full use of UG in product level modeling technique and knowledge fusion techniques, make up the robotization of product design, also expand to the upper design of product to a great extent simultaneously, the real concept design is integrated with follow-up detailed design and technological design.System development platform is selected as follows:

Minimum hardware configuration: PC586

The 32M internal memory

The 16M video memory

4 times of above CD drive of speed

Software minimalist configuration: Window9X/NT4/2000/XP operating system

Unigraphics NX 4.0 platforms

Microsoft?Visual?C++6.0

Microsoft?Access?6.0

The foundation of product knowledge database

Product data sheet: the individuality of analyzing existing deisgn product family is formed, and by consulting electric tool handbook and design specifications, the design specification of each building block of electric drill product is summarized as follows, and sets up following corresponding product data sheet.Individual products data knowledge storehouse must be along with the generation of new product category, and constantly upgrades and replace.The data knowledge storehouse just can constantly be replenished.Analyze the product category individuality and the principal character of existing electric drill knowledge data base, set up part electric drill product database form, see Table 5-1.Table 5-1 is the relevant main feature of electric drill product database.

Model specification	Maximum drilling diameter Φ/mm	Rated voltage/V	Nominal torque/(N m)	Power/W	Rated speed/(r/m in)	Physical dimension (length * wide * height)/mm	Weight/kg
								J1Z-CD3-6A	6	220	0.85	250	1200	236X62X166	1.2
J1Z-HU-6A	6	220	0.85	300	2300	248X65X160	1.35
								J1Z-KL-6A	6	220	0.86	230	2400	210X65X175	1.4
J1Z-KW-6(6.5) A	6.5	220	0.56	240	2000	220X64X215	1.0
								J1Z-CD-6C	6	220	0.5	320	2100	202X63X148	1.13
J1Z-MH-6C	6	220	0.55	220	2200	234X65X158	1.8
								J1Z-SF1-6C	6	220	0.5	210	2000	179X63X145	0.9
J1Z-YD2-6C	6	220	0.5	230	2010	205X61X153	0.8
								J1Z-CD-8C	8	220	1.0	360	2400	205X61X175	1.52
J1Z-IIU05-10 A	10	220	2.2	450	820	225X55X125	1.6
								J1Z-KW02-10 A	10	220	2.3	450	820	210X80X218	1.8
J1Z-MH-10A	10	220	2.2	430	1200	288X72X171	2.0
								J1Z-SD04-10A	10	220	2.2	320	900	204X68X160	1.53
J1Z-SF1-10A	10	220	1.0	280	1800	235X70X195	1.5
								J1Z-CD3-13A	13	220	4.0	500	920	295X70X190	2.0
J1Z-FQ-13	13	120	7.8	800	800	300X90X220	2.7
								J1Z-FR-13	13	230	7.8	800	800	300X90X220	2.7
J1Z-KW02-13 A	13	220	4.0	520	920	295X113X258	3.0
								J1Z-MH-13A	13	220	4.0	480	1300	359X72X140	2.0
J1Z-SF2-13A	13	220	4.0	440	880	290X100X220	4.5
								J1Z-LD01-19A	16	220	4.0	810	820	410X104X345	5.7
J1Z-LS2-19A	16	220	4.1	880	820	475X118X337	5.7
								J1Z-SF1-19A	16	220	4.5	880	860	340X121X315	5.9
J1Z-LD01-23A	16	220	5.0	810	820	410X104X345	5.7
								J1Z-LS2-23A	14	220	5.0	880	880	475X118X337	5.7
J1Z-SF1-23A	16	220	5.0	880	800	340X121X315	5.9
								…	…	…	…	…	…	…

Table 5-1

Above table has been analyzed each electric drill kind of existing database, just principal character is showed, and the part feature data of each electric drill all do not show.

System's detailed design: in the detailed design process of system, mainly divide three aspects to launch, 1, to database according to existing deisgn product, each parts list is shown as the form of design matter-element.2,, establish product design platform part and form, and then establish the composition of customization part according to the method that can open up cluster.3, analyze the demand properties of customers, determine whether and to upgrade conversion to the part in the structure storehouse.

Product family's design platform part constitutes: enter platform design system of electric drill product family, check that the part of electric drill constitutes.Check each part nodes domains of forming electric drill.For example about motor and drill chuck nodes domains partly; According in the distance formula for the seeking of the best profit individuality, obtain some body characteristics numerical value; According to above-mentioned data, obtain the part composition that product family's platform constitutes.

Claims (1)

1. one kind based on the electric drill part classification method that can open up cluster, and it is characterized in that: described electric drill part classification method may further comprise the steps:

1) extracts electric drill design of part characteristic node territory: establish and wait to classify the set of electric drill part

I product A wherein _i={ R _1m..., R _Rm(1)

The expression product A _iBy r electric drill part R _iForm, wherein electric drill part max architecture characteristic number is m, uses v _Rm=c (R _Rm) expression is used for the structure value of cluster analysis;

For i the architectural feature v that can generate profit maximized electric drill part _iSetting up its matter-element model is:

$R_i=\left[\begin{array}{ccc}{N}_i,& c_{i1},& v_{i1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{im}},& v_{\mathrm{im}}\end{array}\right]---\left(2\right)$

Im represents electric drill part R _iThe feature number;

For R _i,, obtain A by determining the constant interval of a certain architectural feature of electric drill part _iEigenwert nodes domains V=＜v _Min, v _Max, obtain the interval nodes domains X of character numerical value according to formula (2) _Njm=＜V _Min, V _Max, setting up the nodes domains model is R _Mi:

$R_{\mathrm{mi}}=\left[\begin{array}{ccc}{N}_m,& c_{m1},& X_{m1}\\ & .& .\\ & .& .\\ & .& .\\ & c_{\mathrm{mi}}& X_{\mathrm{mi}}\end{array}\right]---\left(3\right)$

X wherein _Njm(k=1 ..., mi) expression electric drill part R _iN character numerical value interval;

2) determine architectural feature variable numerical value v: generate profit maximized architectural feature trend based on the existing procucts on the market, utilize mid point gradually approximate algorithm obtain electric drill part optimum structure feature value v _i

When the requirement to architectural feature c belongs to more little dominance more, then

$v_{i+1}=\frac{v_i+v_{\mathrm{mim}}}{2},i=\mathrm{1,2},...---\left(4\right)$

$v_1=\frac{v_{\min }+v_{\max }}{2}$

When the requirement to architectural feature c belongs to big more dominance more, then

$v_{i+1}=\frac{v_i+v_{\max }}{2},i=\mathrm{1,2},...---\left(5\right)$

$v_1=\frac{v_{\min }+v_{\max }}{2}$

3) set up each electric drill design of part apart from formula: according to formula (3) (4) and (5), calculate R _iDistance is worth the correlation degree ρ of its nodes domains:

$\mathrm{\&rho;}(v,v_i,X)=|v-v_i|-\frac{v_{\max }-v_{\min }}{2}---\left(6\right)$

The definition structure feature is apart from value:

Apart from value K (R _i, R _Im) represented R _iThe related numerical value of architectural feature c, ω _iObtain by analytical hierarchy process, the weight of representation feature element,

4) determine R _i, R _jStructure connection degree T (i, j):

T(i，j)＝|K(R _i，R _im)-K(R _j，R _jm)| (8)

Wherein T (i, j)=T (j i), follows same computing method, for n architectural feature of deisgn product, calculate respectively T (i, j) (i=1 ..., n; J=1 ..., n), they constitute symmetric matrix M:

Or T (i j) levels off to 0, R _i, R _jBetween correlation degree big more;

(i, j)＞0, in this case, (i, value j) is big more, expression R for T for T _i, R _jBetween correlation degree more little;

5) symmetric matrix M is carried out numerical analysis, because each row is represented the degree of association between an electric drill part and other n-1 the electric drill part, and T (i, i)=0 (i=1 ..., n), choose the minimum value in the matrix M, be made as T (p, q)=min{T (i, j), j=1 ..., n; I=1 ..., n; I ≠ j}, i.e. electric drill part R _p, R _qBetween the degree of association be minimum; R _p, R _qPut under in the same class, be designated as { p, q}, electric drill part R _p, R _qForm new electric drill part R _Pq, for any R _i, it and R _PqBetween the degree of association be: T (i, pq)=min{T (i, p), T (i, q) };

When n＞2, scratch the capable and q row of p in the matrix M, according to R _PqWith any R _i, between the degree of association, they form a new symmetric matrix jointly, its exponent number is n-1;

6) the average degree of association numerical value of establishment cluster analysis:

Set the average degree of association

7) the symmetrical incidence matrix on the n-1 rank of new formation is utilized algorithmic procedure in the step 5), obtain the symmetric matrix on n-2 rank, the circulation said process, when the minimum degree of association during greater than the K value, then stop cluster process, obtain unique cluster result: the classification of electric drill part, and the electric drill part of each classification.

Images