CN102222139A - Method for quickly determining low-voltage motor protection device - Google Patents

Abstract

The invention relates to a method for quickly determining a low-voltage motor protection device, which comprises the following steps: on the basis of axiomatization theory, dividing function models of the low-voltage motor protection device to obtain a coupling-free or decoupling relationship matrix meeting design requirements; on the basis of a fuzzy cluster analysis method, mapping user requirements to the function models to obtain a relationship characteristic matrix; on the basis of module examples, reasoning the configuration of the low-voltage motor protection device, and setting up a product example search library; by applying a parameterization technology and utilizing dimension constraint and product device key identifiers, parameterizing a universal circuit model of the low-voltage motor protection device; calculating a low-voltage motor inverse time limit protection time parameter and timely carrying out protection actions; and searching the product example search library according to the user requirements to obtain corresponding information models for the low-voltage motor protection device or modifying product example search. The method meets the various user requirements, and simultaneously uses the designed existing process in the product design process so as to reduce the design expense and the production cycle.

Description

The method that motor Protective Device is determined fast

Technical field

The invention belongs to mechanical manufacturing field, what relate to is the method that a kind of low-tension motor is determined fast.

Background technology

In current international manufacture field, design develops to the advanced manufacturing technology direction from classic method with manufacturing technology, and manufacturing informationization and globalization have become an important trend of development of manufacturing.The core of manufacturing enterprise's informatization is Design of digital, manufacturing and correlated digital technology integrated of product.The product rapid development Study on Technology becomes the focus that enterprise's design is used just gradually, also is the necessary instrument of enterprise innovation capability improving.The design process of product is creatively to set up the active procedure that satisfies the functional requirement technological system.The product design of motor Protective Device is an important process that influences properties of product, quality, cost and business economic benefit; can product satisfy customer requirement; can respond turn of the market fast, depend on the quick design effort of product to a great extent.

Find through literature search prior art, gordian technique for the quick design of product, professor N.P.Suh of U.S. MIT department of mechanical engineering has proposed design theory, promptly the process of entire product design is divided into four territories, is respectively: client territory (Customer Domain), functional domain (Functional Domain), domain (Physical Domain) and process domain (Process Domain).The entire product design process is exactly the process of the mutual mapping of parameter between two territories adjacent one another are.Dumm has proposed the fuzzy C mean algorithm, through Bezdek it is promoted and development, and it is a kind of local iteration optimizing algorithm in essence.The Kritik design support system of development such as the Goel of Georgia Institute of Technology, what this system adopted is the module instance inference method.The Yang Ze of University Of Tianjin is flat to have studied microcomputer transformer module design proposal, and modular design makes the protection thinking of microcomputer transformer more clear, has improved the design efficiency and the reliability of protection software.Zhejiang University has studied in the Liao Huai continent and has made full use of existing parts and carry out the parametrization design proposal, increases the ratio of Variant Design parts and standard, with the least possible parts, is combined into the product of multiple as far as possible class.

Existing Fast design method can only effectively solve the design problem of local link, not integrating quick design key technological system.Do not use as yet at the motor Protective Device design aspect, lack the complete motor Protective Device scheme of design fast, have much room for improvement and improve to satisfy actual demand.

Summary of the invention

The objective of the invention is to propose the method that a kind of motor Protective Device is determined fast at the deficiency in the existing motor Protective Device design, satisfying the diversified demand of client, and the requirement of high product quality and agility.Make enterprise in product design process, use the ready-made technology that had designed simultaneously, reduce design cost and production cycle.

For reaching above purpose, solution of the present invention is:

The method that a kind of motor Protective Device is determined fast, it may further comprise the steps:

1) based on Axiomatic Theories the motor Protective Device functional module is divided

According to Axiomatic Theories, known current constraint and motor Protective Device design knowledge base, be met the relational matrix of independent axiom, decompose motor Protective Device functional module and domain up to the bottom, but the relational matrix of the nothing that obtains adhering to specification coupling or decoupling zero;

2) based on Fuzzy Cluster Analysis method customer demand is shone upon to functional module

Fuzzy cluster analysis is divided into different monoids to customer demand according to different features, feature can be client's essential characteristic, customer value feature and client's behavioural characteristic, by the relation property matrix of customer demand to the functional module mapping step 1) mesolow motor protecting device functional module and customer demand is connected;

3) dispose based on module instance reasoning motor Protective Device

Set up modular product case retrieval storehouse, it must be expressed in required various information in the configuration design process effectively, for determining that motor Protective Device provides case retrieval;

4) the utilization parametric technology utilizes dimension constraint and product components and parts key identifier with motor Protective Device universal circuit module parameterization;

5) low-tension motor inverse time-lag protection algorithm computation goes out low-tension motor inverse time-lag protection time parameter, in time protects action;

6) according to customer demand retrieval product example search library, obtain motor Protective Device, perhaps, if there is not then implementation step 1 of respective instance)-5) at least one revise the product example search library.

Described step 2) the relation property matrix in should carry out decoupling zero if do not meet the coupling mapping of independent axiom, and the clustering algorithm decoupling zero that it can adopt best cluster numbers comprises:

Phase one is determined the best cluster numbers of clustering algorithm

Input: sample data

Output: sample cluster centre, the best cluster numbers of sample

Step 1 employing Euclidean distance formula calculates the Euclidean distance between all sample number strong points, generates the Euclidean distance matrix;

Step 2 adopts the sum of squares of deviations method, calculates the link distance between the sample data class, and generates the hierarchical cluster tree,

This step concretism is: be provided with N sample, N sample respectively is a class when initial, calculate distance between them according to sum of squares of deviations distance calculation formula then, to and be a new class apart from the class of minimum, the new class after calculating and the class and the distance of other classes will merge apart from two classes of minimum again, and so forth again, till N sample is merged into a class

Sum of squares of deviations distance calculation formula: $S=\underset{t=1}{\overset{k}{\mathrm{\Σ}}}\underset{\mathrm{\α}=1}{\overset{N_t}{\mathrm{\Σ}}}{(X_{\left(t\right)}^{\left(\mathrm{\α}\right)}-{\stackrel{\‾}{X}}_{\left(t\right)})}^{\′}(X_{\left(t\right)}^{\left(\mathrm{\α}\right)}-{\stackrel{\‾}{X}}_{\left(t\right)})$

N sample is divided into k class: C ₁, C ₂..., C _k, in the formula: use Expression C _tIn i sample, N _tExpression C _tIn number of samples,

Be C _tCenter of gravity;

Step 3 obtains different sample segmentation results according to link distance calculated value cutting hierarchical cluster tree;

Step 4 is added up value according to the F that F statistic formula calculates under the different segmentation results, and with it is calculated that the result, the corresponding optimal classification result of maximum F statistics value obtains best cluster numbers of sample data and corresponding cluster result thus;

F statistic formula: $F=\frac{\underset{j=1}{\overset{r}{\mathrm{\Σ}}}{n}_j{\left|\right|{\stackrel{\‾}{x}}^{\left(j\right)}-\stackrel{\‾}{x}\left|\right|}^2/(r-1)}{\underset{j=1}{\overset{r}{\mathrm{\Σ}}}\underset{i=1}{\overset{n_j}{\mathrm{\Σ}}}{\left|\right|x_i^{\left(j\right)}-{\stackrel{\‾}{x}}^{\left(f\right)}\left|\right|}^2/(n-r)}$

In the formula:

With Distance,

It is sample in the j class

With the center

Distance;

The clustering algorithm of the best cluster numbers of subordinate phase operative norm

Input: Weighting exponent m, sample data collection, algorithm iteration end condition ε, the best cluster numbers of sample

Output: cluster centre matrix V, fuzzy partition matrix U

The value that step 1 obtains after the phase one with execution algorithm is an initialization condition, and input sample data collection X, end condition ε＞0, Weighting exponent m;

Step 2 utilizes following formula to calculate sample and distances of clustering centers,

$d_A^2(x_j,v_i^{\left(l\right)})={(x_j-v_i^{\left(l\right)})}^{T}A(x_j-v_i^{\left(l\right)})$

In the formula: 1≤i≤c, 1≤j≤n;

Step 3 order

Revise U ^(l+1),

$u_{\mathrm{ij}}^{\left(l\right)}=\left\{\begin{array}{cc}\frac{1}{\underset{r=1}{\overset{c}{\mathrm{\&Sigma;}}}{d}_A^2(x_j,v_i)/{(x_j,v_r)}^{2m-1}}& {\mathrm{<figure-callout\; id="2"\; label="d\; ij"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">d</figure-callout>}}_{\mathrm{ij}}^{}>0\\ 1& {\mathrm{<figure-callout\; id="2"\; label="d\; ij"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">d</figure-callout>}}_{\mathrm{ij}}^{}=0\\ 0& k\&NotEqual;i,{\mathrm{<figure-callout\; id="2"\; label="d\; jk"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">d</figure-callout>}}_{\mathrm{jk}}^{}=0\end{array}\right.;$

Step 4 couple U ^(l)Upgrade the cluster centre battle array

$v_i^{(l+1)}=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(l\right)}\right)}^m{x}_j/\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(l\right)}\right)}^m$

In the formula: 1≤i≤c, U ^(l)Represent the degree of membership matrix after the iteration l time,

Be the cluster center of gravity after l+1 iteration;

Step 5 if || V ^(l+1)-V ^(l)||≤ε, then stop iteration, otherwise carry out l=l+1, jump procedure 2 is determined low-voltage motor protector merit structure unit cluster feature coupling incidence matrix,

Clustering algorithm based on best cluster numbers carries out obtaining non-coupling mapping after the decoupling zero,

Described step 3) comprises:

Example in the product example search library; Use C _k={ a _K1, a _K2..., a _Ki..., a _Kn, k=1,2 ..., m. represents that wherein m represents the instance number in the current case library, a _Ki(i=1,2 ..., n) be example C _kAttribute, represent a with a ternary Vector Groups _Ki=(p _i, w _i, x _i), p wherein _iThe representation attribute name, w _iThe weighted value of representation attribute, x _iThe representation attribute value;

Client's product demand; Use Q ₀={ a ₀₁, a ₀₂..., a _0iExpression, according to customer demand Q ₀With example C _kCarry out similarity and calculate, retrieval meets the example of customer demand most;

Similarity is calculated,

S (a wherein _0i.a _Ki) be object a _0iAnd a _KiSimilarity, similarity is 1 when their property value is identical, similarity is 0 in the time of different;

Relatively similarity is selected the most similar module instance, deposits the module library after the modification again in.

Described step 4) comprises:

Motor Protective Device universal circuit to same function is concluded, and is divided into several classes, and in same building mode, components and parts model difference, the circuit module that shape is identical is classified as a class, determines to revise parameter with this;

Divide the result according to Axiomatic Design Theory, each universal circuit module that will be relevant with functional domain is carried out related effectively, constitute the motor Protective Device circuit, need to select the low-tension motor local circuit module of modification, according to aforementioned definite modification parameter the universal circuit module is made amendment, make that original localized design is reused;

After the instance parameter modification, the optimum example that the CAD (computer aided design) system was revised saves as new example, expands the product example search library.

Owing to adopted such scheme, the present invention to have following characteristics: the present invention can satisfy the diversified demand of client, makes enterprise use the ready-made technology that had designed simultaneously in product design process, reduces design cost and production cycle.

Description of drawings

Fig. 1 is the synoptic diagram in four territories of axiom design of the present invention.

Fig. 2 is the mapping process synoptic diagram between functional domain of the present invention and the domain.

Fig. 3 shines upon synoptic diagram between functional module of the present invention and domain.

Fig. 4 is the anti-time limit characteristic curve map after the present invention divides.

Fig. 5 is a motor Protective Device fast determination method process flow diagram of the present invention.

Embodiment

The present invention is further illustrated below in conjunction with the accompanying drawing illustrated embodiment.

The axiom design is divided into four territories to the process of entire product design, is respectively: user domain (Customer Domain), functional domain (Functional Domain), domain (Physical Domain) and process domain (Process Domain).The entire product design process is exactly the process of the mutual mapping of parameter between two territories adjacent one another are, as shown in Figure 1.In fact the design process of product is exactly " Z " font mapping process between four territories, from the process progressively launching of high-level low level to detail of design abstract concept.

Hierarchical relationship between functional domain and the domain, as shown in Figure 2.The designer must analyze the product design demand, determines the requirement of product general function and maps out the overall design parameter, it is decomposed into all kinds of subfunctions again and requires to reach with it design parameter one to one.Be divided into the subfunction requirement from the general function requirement, a functional hierarchy model can be decomposed and form to functional requirement just, and corresponding design parameter also can be divided into different ranks.Each layer all has the design object of oneself, and the decision-making of high-rise design influences the state of finding the solution of low level problem.

In " Z " font mapping process of axiomatization design, make correct decision-making and must follow independent axiom and information axiom.The basic intension of independent axiom will remain the independence of functional requirement (FRs) exactly, and FRs is defined as designing the minimal set of necessary need for independence.Axiom design come representation function to require with the form of formula and design parameter between relation as follows:

{FR} _m*l＝[A] _m*n{DP} _n*l

{ FR} wherein _M*lBe function vector, { DP} _N*lBe the design parameter vector, [A] _M*nBe design matrix.Wherein [A] _M*nCan followingly represent:

$A_{m*n}=\left|\begin{array}{ccccc}{A}_{11}& A_{12}& A_{13}& \&CenterDot;\&CenterDot;\&CenterDot;& A_{1n}\\ A_{21}& A_{22}& A_{23}& \&CenterDot;\&CenterDot;\&CenterDot;& A_{2n}\\ A_{31}& A_{32}& A_{33}& \&CenterDot;\&CenterDot;\&CenterDot;& A_{3n}\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ A_{m1}& A_{m2}& A_{\mathrm{<figure-callout\; id="3"\; label="m"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">m</figure-callout>}3}& \&CenterDot;\&CenterDot;\&CenterDot;& A_{\mathrm{mn}}\end{array}\right|$

Wherein each element in the design matrix can be by following formulate:

The present invention includes following steps:

Step 1 is based on the division of the motor Protective Device functional module of Axiomatic Theories

The deviser uses Axiomatic Theories; according to known current constraint (CAs) and motor Protective Device design knowledge base etc.; be met the relational matrix of independent axiom, the motor Protective Device functional requirement as shown in Figure 3 and the decomposable process of domain.FRs and DPs vector are decomposed, up to the FRs and the DPs of the bottom.Whole decomposable process is one and abstractly is summarized into low-level detailed description process from high level, but finally obtains the nothing coupling that adheres to specification or the relational matrix of decoupling zero, guarantees the independence between the low-tension motor defencive function module.

The mapping relations of setting up each layer between low-tension motor functional module and domain are as follows:

The decomposition of (1) first level

The general function of motor Protective Device requires FR for realizing the various failure protection functions of low-tension motor, and corresponding structure territory DP is the domain of each functional module of motor Protective Device.Function FR ₁, FR ₂And FR ₅Can be by domain DP ₁, DP ₂And DP ₅Independently finish, and function FR ₃, FR ₄And FR ₆At domain DP own ₃, DP ₄And DP ₆Influence under, also to be subjected to DP ₁And DP ₂The influence of domain.Can obtain following relational matrix by the relation of analyzing them.Wherein, X represents between motor Protective Device defencive function and the corresponding defencive function universal circuit module parameter relevant (value of X is according to the industry experience) that 0 expression is separate or related very little between the two.

$\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="FR"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">FR</figure-callout>}}_1\\ {\mathrm{<figure-callout\; id="2"\; label="FR"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">FR</figure-callout>}}_2\\ {\mathrm{FR}}_5\\ {\mathrm{<figure-callout\; id="3"\; label="FR"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">FR</figure-callout>}}_3\\ {\mathrm{FR}}_4\\ {\mathrm{FR}}_6\end{array}\right]=\left|\begin{array}{cccccc}X& 0& 0& 0& 0& 0\\ 0& X& 0& 0& 0& 0\\ 0& 0& X& 0& 0& 0\\ X& X& 0& X& 0& 0\\ X& X& 0& 0& X& 0\\ X& X& X& X& \mathrm{<figure-callout\; id="1"\; label="X\; X\; DP"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">X</figure-callout>}& X\\ \end{array},\right|\left[\begin{array}{c}{\mathrm{DP}}_{}\end{array}\right]\left[\begin{array}{c}{}_1\\ {\mathrm{<figure-callout\; id="2"\; label="DP"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">DP</figure-callout>}}_2\\ {\mathrm{DP}}_5\\ {\mathrm{<figure-callout\; id="3"\; label="DP"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">DP</figure-callout>}}_3\\ {\mathrm{DP}}_4\\ {\mathrm{DP}}_6\end{array}\right]$

The decomposition of (2) second levels

FR1 can further be decomposed into: analog quantity detects FR ₁₁, quantity of state is measured FR ₁₂, on off state monitoring FR ₁₃The corresponding structure territory is: analog quantity load module DP ₁₁, quantity of state load module DP ₁₂, switching input module DP ₁₃Function FR ₁₁, FR ₁₂And FR ₁₃All can be by domain DP ₁₁, DP ₁₂And DP ₁₃Independently finish, relational matrix is as follows:

$\left[\begin{array}{c}{\mathrm{<figure-callout\; id="11"\; label="FR"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">FR</figure-callout>}}_{11}\\ {\mathrm{<figure-callout\; id="12"\; label="FR"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">FR</figure-callout>}}_{12}\\ {\mathrm{FR}}_{13}\end{array}\right]=\left|\begin{array}{ccc}X& 0& 0\\ 0& X& 0\\ 0& 0& X\end{array}\right|\left[\begin{array}{c}{\mathrm{<figure-callout\; id="11"\; label="DP"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">DP</figure-callout>}}_{11}\\ {\mathrm{<figure-callout\; id="12"\; label="DP"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">DP</figure-callout>}}_{12}\\ {\mathrm{DP}}_{13}\end{array}\right]$

Similar FR ₂Also can further be decomposed into: monitor operation or participate in industrial automation process FR ₂₁, carry out the corresponding command action FR ₂₂The corresponding structure territory is: analog quantity transformation output module DP ₂₁, relay output module DP ₂₂, relational matrix is as follows:

$\left[\begin{array}{c}{\mathrm{<figure-callout\; id="21"\; label="FR"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">FR</figure-callout>}}_{21}\\ {\mathrm{<figure-callout\; id="22"\; label="FR"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">FR</figure-callout>}}_{22}\end{array}\right]=\left|\begin{array}{cc}X& 0\\ \mathrm{<figure-callout\; id="21"\; label="X\; X\; DP"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">X</figure-callout>}& X\\ \end{array},\right|\left[\begin{array}{c}{\mathrm{DP}}_{}\end{array}\right]\left[\begin{array}{c}{}_{21}\\ {\mathrm{<figure-callout\; id="22"\; label="DP"\; filenames="HDA0000066282260000012.png"\; state="\{\{state\}\}">DP</figure-callout>}}_{22}\end{array}\right]$

FR ₃Further be decomposed into: most common failure protection FR ₃₁, TE guardtime (allowing the stall time during 7 times of rated current) FR ₃₂, earth leakage protection FR ₃₃, thermal overload protection FR ₃₄The corresponding structure territory is: basic defencive function module DP ₃₁, TE guardtime module DP ₃₂, leakage protection module DP ₃₃, thermal overload protection module DP ₃₄The failure mode of motor is a lot, and corresponding motor protecting device also can design various emergency protection modules, only some common faults protections has been assembled a basic defencive function module among the present invention, and concrete detailed design is given an example no longer one by one.Relational matrix is as follows:

$\left[\begin{array}{c}{\mathrm{FR}}_{31}\\ {\mathrm{FR}}_{32}\\ {\mathrm{FR}}_{33}\\ {\mathrm{FR}}_{34}\end{array}\right]=\begin{array}{c}\left|\begin{array}{cccc}X& 0& 0& 0\\ 0& X& 0& 0\\ 0& 0& X& 0\\ 0& 0& 0& X\end{array}\right|\end{array}\left[\begin{array}{c}{\mathrm{DP}}_{31}\\ {\mathrm{DP}}_{32}\\ {\mathrm{DP}}_{33}\\ {\mathrm{DP}}_{34}\end{array}\right]$

FR in like manner ₄Can further be decomposed into: network remote control control function FR ₄₁, dead electricity is reset function FR automatically ₄₂The corresponding structure territory is: network remote control divide-shut brake module DP ₄₁, restart control function DP ₄₂Relational matrix is as follows:

$\left[\begin{array}{c}{\mathrm{FR}}_{41}\\ {\mathrm{FR}}_{42}\end{array}\right]=\left|\begin{array}{cc}X& 0\\ 0& X\end{array}\right|\left[\begin{array}{c}{\mathrm{DP}}_{41}\\ {\mathrm{DP}}_{42}\end{array}\right]$

The decomposition of (3) tri-layers

With function FR ₁₁For example further is decomposed into: three-phase voltage is measured FR ₁₁₁, three-phase current detection FR ₁₁₂, leakage current numerical measuring FR ₁₁₃, temperature monitoring FR ₁₁₄The corresponding structure territory is: three-phase voltage load module DP ₁₁₁, three-phase current load module DP ₁₁₂, leakage current load module DP ₁₁₃, temperature and resistance and 4-20mA load module DP ₁₁₄FR wherein ₁₁₁, FR ₁₁₂, FR ₁₁₃And FR ₁₁₄Domain DP can be arranged ₁₁₁, DP ₁₁₂, DP ₁₁₃And DP ₁₁₄Independently finish, concrete relational matrix is as follows:

$\left[\begin{array}{c}{\mathrm{FR}}_{111}\\ {\mathrm{FR}}_{112}\\ {\mathrm{FR}}_{113}\\ {\mathrm{FR}}_{114}\end{array}\right]=\begin{array}{c}\left|\begin{array}{cccc}X& 0& 0& 0\\ 0& X& 0& 0\\ 0& 0& X& 0\\ 0& 0& 0& X\end{array}\right|\end{array}\left[\begin{array}{c}{\mathrm{DP}}_{111}\\ {\mathrm{DP}}_{112}\\ {\mathrm{DP}}_{113}\\ {\mathrm{DP}}_{114}\end{array}\right]$

Other parameters are by this decomposition.

Step 2 is shone upon customer demand based on Fuzzy Cluster Analysis method to functional module

Fuzzy cluster analysis is divided into the discussion customer demand according to different features the problem of different monoids.The angle difference that different segmentation features is paid close attention to, the content of different feature reflections is also different.Feature can be client's essential characteristic, customer value feature and client's behavioural characteristic.By the relation property matrix of customer demand step I mesolow motor protector functional module and customer demand are connected to the functional module mapping.

Wherein A refers to the characteristic index matrix with n client's sample, a _iFor a sample of k characteristic index, a are arranged _IkK the characteristic index that refers to i sample.If do not meet the coupling mapping of independent axiom, should carry out decoupling zero, will introduce utilization D-FCM (clustering algorithm of best cluster numbers) decoupling algorithm below and obtain uncoupled mapping relations.

D-FCM (clustering algorithm of best cluster numbers) decoupling algorithm basic step is as follows:

Phase one is determined the best cluster numbers of FCM algorithm

Input: sample data

Output: sample cluster centre, the best cluster numbers of sample

Step 1 employing Euclidean distance formula calculates the Euclidean distance between all sample number strong points, generates the Euclidean distance matrix.

Step 2 adopts the sum of squares of deviations method, calculates the link distance between the sample data class, and generates the hierarchical cluster tree.

This step concretism is: be provided with N sample, N sample respectively is a class when initial, calculates distance between them according to sum of squares of deviations distance calculation formula (1) then, will and be a new class apart from the class of minimum.The new class after calculating and the class and the distance of other classes will merge apart from two classes of minimum again again, and so forth, and till N sample is merged into a class.

$S=\underset{t=1}{\overset{k}{\mathrm{\&Sigma;}}}\underset{\mathrm{\&alpha;}=1}{\overset{N_t}{\mathrm{\&Sigma;}}}{(X_{\left(t\right)}^{\left(\mathrm{\&alpha;}\right)}-{\stackrel{\&OverBar;}{X}}_{\left(t\right)})}^{\&prime;}(X_{\left(t\right)}^{\left(\mathrm{\&alpha;}\right)}-{\stackrel{\&OverBar;}{X}}_{\left(t\right)})---\left(1\right)$

N sample is divided into k class: C ₁, C ₂..., C _k, in the formula: use

Expression C _tIn i sample, N _tExpression C _tIn number of samples,

Be C _tCenter of gravity.

Step 3 obtains different sample segmentation results according to link distance calculated value cutting hierarchical cluster tree.

Step 4 is added up value according to the F that F statistic formula (2) calculates under the different segmentation results, and with it is calculated that the result, the corresponding optimal classification result of maximum F statistics value can obtain best cluster numbers of sample data and corresponding cluster result thus.

$F=\frac{\underset{j=1}{\overset{r}{\mathrm{\&Sigma;}}}{n}_j{\left|\right|{\stackrel{\&OverBar;}{x}}^{\left(j\right)}-\stackrel{\&OverBar;}{x}\left|\right|}^2/(r-1)}{\underset{j=1}{\overset{r}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{n_j}{\mathrm{\&Sigma;}}}{\left|\right|x_i^{\left(j\right)}-{\stackrel{\&OverBar;}{x}}^{\left(f\right)}\left|\right|}^2/(n-r)}---\left(2\right)$

In the formula:

For

With Distance,

It is sample in the j class With the center

Distance.

Subordinate phase operative norm FCM clustering algorithm

Input: Weighting exponent m, sample data collection, algorithm iteration end condition ε, the best cluster numbers of sample

Output: cluster centre matrix V, fuzzy partition matrix U

The value that step 1 obtains after the phase one with execution algorithm is an initialization condition, and input sample data collection X, end condition ε＞0, Weighting exponent m.

Step 2 utilizes formula (3) to calculate sample and distances of clustering centers.

$d_A^2(x_j,v_i^{\left(l\right)})={(x_j-v_i^{\left(l\right)})}^{T}A(x_j-v_i^{\left(l\right)})---\left(3\right)$

In the formula: 1≤i≤c, 1≤j≤n.

Step 3 order

Revise U ^(l+1).

$u_{\mathrm{ij}}^{\left(l\right)}=\left\{\begin{array}{cc}\frac{1}{\underset{r=1}{\overset{c}{\mathrm{\&Sigma;}}}{d}_A^2(x_j,v_i)/{(x_j,v_r)}^{2m-1}}& {\mathrm{<figure-callout\; id="2"\; label="d\; ij"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">d</figure-callout>}}_{\mathrm{ij}}^{}>0\\ 1& {\mathrm{<figure-callout\; id="2"\; label="d\; ij"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">d</figure-callout>}}_{\mathrm{ij}}^{}=0\\ 0& k\&NotEqual;i,{\mathrm{<figure-callout\; id="2"\; label="d\; jk"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">d</figure-callout>}}_{\mathrm{jk}}^{}=0\end{array}\right.$

Step 4 couple U ^(l)Upgrade the cluster centre battle array

$v_i^{(l+1)}=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(l\right)}\right)}^m{x}_j/\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(l\right)}\right)}^m$

In the formula: 1≤i≤c, U ^(l)Represent the degree of membership matrix after the iteration l time,

Be the cluster center of gravity after l+1 iteration.

Step 5 if || V ^(l+1)-V ^(l)||≤ε then stops iteration, otherwise carries out l=l+1, jump procedure 2.

Determine low-voltage motor protector merit structure unit cluster feature coupling incidence matrix,

Carry out obtaining non-coupling mapping after the decoupling zero based on D-FCM,

Step 3 disposes based on module instance reasoning motor Protective Device

Carry out the retrieval of modular product example, need set up the corresponding information model, its model must be expressed in required various information in the configuration design process effectively, for finding the solution of design problem provides support.Concrete case retrieval step is as follows.

(1) the example C in the case library _k={ a _K1, a _K2..., a _Ki..., a _Kn, k=1,2 ..., m. represents that wherein m represents the instance number in the current case library.a _Ki(i=1,2 ..., n is example C _kAttribute, can represent with a ternary Vector Groups: a _Ki=(p _i, w _i, x _i), p wherein _iThe representation attribute name, w _iThe weighted value of representation attribute, x _iThe representation attribute value.

(2) client's product demand can be used Q ₀={ a ₀₁, a ₀₂..., a _0iExpression, according to customer demand Q ₀With example C _kCarry out similarity and calculate, retrieval meets the example of customer demand most.

(3) similarity computational algorithm,

S (a wherein _0i.a _Ki) be object a _0iAnd a _KiSimilarity, similarity is 1 when their property value is identical, similarity is 0 in the time of different.

(4) compare similarity, select the most similar module instance, deposit the module library after the modification again in.

Step 4, motor Protective Device universal circuit module parameterization

(1) the low-voltage motor protector universal circuit of same function is concluded, be divided into several classes.In same building mode, components and parts model difference, the circuit module that shape is identical is classified as a class, determines to revise parameter with this.

(2) divide the result according to Axiomatic Design Theory, each universal circuit module that will be relevant with functional domain is carried out related effectively, constitutes the motor Protective Device circuit.The designer can select its low-tension motor local circuit module that need revise, according to the modification parameter of step (1) the universal circuit module is made amendment, and makes that original localized design is reused.

(3) after the instance parameter modification, the optimum example that CAD (CAD (computer aided design)) system was revised saves as new example, expands case library.

Step 5 is calculated the low-tension motor inverse time-lag protection

The algorithm of low-tension motor inverse time-lag protection specifically comprises as follows:

(1) the universal expression formula of inverse-time overcurrent protection characteristic equation

In the formula, t is a delay time, and K is the relay design constant, T _pBe client's constant of adjusting, I is a sample rate current, I _pBe the inverse time lag current setting, r gets different values according to the different occasions of protection.

(2) according to low-tension motor heat accumulation principle low-tension motor anti-time limit characteristic curve is divided, Fig. 4 works as I for the anti-time limit characteristic curve after dividing _A/ I _N〉=20 o'clock, with specified time t _MinThe overcurrent protection characteristic, and cooperate with the syllogic overcurrent protection time limit.When electric current hour because the influence of measuring accuracy, the time delay error after causing adding up is bigger, therefore blocks with a maximum straight line 2 of definiting time-lag, the electric current multiple at both intersection point places is M.An other straight line 3 that electric current multiple K can adjust is worked as I _A/ I _NDuring＜K, protected element can steady in a long-term move.As M≤I _A/ I _NDuring≤K, with specified time time-delay t _MaxAction; As M≤I _A/ I _N, represent the anti-time limit characteristic deferred action that adds up at≤20 o'clock with the universal expression formula.

(3) to (I/I in the formula _P) ^rTry to achieve, adopt the MacLaurin deployment algorithm to do following expansion:

$f(I_A/I_N)={(I_A/I_N)}^r={[1+(I_A/I_N-1)]}^r=1+r(I_A/I_N-1)+\frac{r(r-1)}{2}{(I_A/I_N-1)}^2+\&CenterDot;\&CenterDot;\&CenterDot;$

$+\frac{r(r-1)\&CenterDot;\&CenterDot;\&CenterDot;(r-n+1)}{n!}{(I_A/I_N-1)}^n+R_n,$

$R_n=\frac{r(r-1)\&CenterDot;\&CenterDot;\&CenterDot;(r-n)}{(n+1)!}{(I_A/I_N-1)}^{\mathrm{<figure-callout\; id="1"\; label="n\; +"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">n</figure-callout>}+1}$

R in the formula _n(x) be the truncation error that n rank MacLaurin launches

(4) use Q ₀Be defined as critical relatively heat, i.e. the largest cumulative heat that can bear when burning of object of protection (low-tension motor).Along with the heat accumulation of object of protection, heat increases gradually, when its heat history reaches Q ₀, the protection tripping operation.If the relative heat that object of protection produces in constant duration Δ t is q.If altogether object of protection is sampled N time before the trip protection, then

Nq＝Q ₀

Trip protection delay time constantly is

t＝MΔt

Can get by following formula, $t=\frac{Q_0}{q}\mathrm{\&Delta;t}$

The final relative heat that must produce the Δ t time is

$q=\frac{Q_0\mathrm{\&Delta;t}}{K}|{(I_A/I_N)}^r-1|$

If q _iBe electric current I _iAt Δ t _iThe relative heat that produces in time is because Δ t _iTime is shorter, and it is constant to be set in the Δ t time electric current, then

${\mathrm{\&Delta;t}}_1:q=\frac{Q_0\mathrm{\&Delta;t}}{K}|{(I_{A1}/I_N)}^r-1|$

${\mathrm{\&Delta;t}}_2:q=\frac{Q_0\mathrm{\&Delta;t}}{K}|{(I_{A2}/I_N)}^r-1|$

${\mathrm{\&Delta;t}}_3:q=\frac{Q_0\mathrm{\&Delta;t}}{K}|{(I_{A3}/I_N)}^r-1|$

$\begin{array}{c}\&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\end{array}$

${\mathrm{\&Delta;t}}_N:q=\frac{Q_0\mathrm{\&Delta;t}}{K}|{(I_{\mathrm{AN}}/I_N)}^r-1|$

In the formula, I _AiFor at Δ t _iThe relative heat that time produces, then

Q ₁＝0+q ₁

Q ₂＝Q ₁+q ₂

.

.

.

Q _N+1＝Q _N+q _N+1

Work as Q _N+1〉=Q ₀Be that heat accumulative Q is incremented to Q ₀The time, trip protection, trip protection delay time be,

t＝Δt ₁+Δt ₂+…++Δt _N＝NΔt

(5) genetic algorithm optimization inverse time protection characteristic curve mathematic model, the low-tension motor method realizes the target mathematical model that family curve is optimized:

$\max \mathrm{\&delta;}\left(n\right)=\frac{r(r-1)\&CenterDot;\&CenterDot;\&CenterDot;(r-n)}{(n+1)!}{(I/I_P-1)}^{\mathrm{<figure-callout\; id="1"\; label="n\; +"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">n</figure-callout>}+1}$

Constraint condition: $\left\{\begin{array}{c}\mathrm{\&delta;}\left(n\right)\&le;0.5\%,2\&le;I/I_P\&le;20\\ t=\mathrm{N\&Delta;t}\&le;\mathrm{3,2}\&le;I/I_P\&le;20\\ Q_{N+1}\&le;{\mathrm{<figure-callout\; id="2"\; label="Q"\; filenames="HDA0000066282260000021.png"\; state="\{\{state\}\}">Q</figure-callout>}}_0,2\&le;I/I_P\&le;20\end{array}\right.$

To above-mentioned Optimization Model, fitness function is $\mathrm{\&sigma;}=|\frac{f(I_A/I_N)-1}{{(I_A/I_N)}^r-1}-1|$

Genetic algorithm parameter is set, optimizes above-mentioned mathematical model, obtained the mathematical model that microprocessor can computing, in program, can calculate inverse time-lag protection actuation time so very easily.

The invention provides a kind of Fast design method of motor Protective Device design process, its flow process as shown in Figure 5.Its detailed step is as follows:

(1) according to the demand inquiry product example storehouse (store the parameters allocation plan of motor Protective Device) of client to aspects such as performance, function and the price proposition of product, if the module that satisfies the demands is arranged in the case library, then direct output products allocation plan, design finishes;

(2) if the module that does not satisfy the demands in the case library is extracted similar product from case library.Each example in customer demand and the product example storehouse is compared, calculate the similarity of all examples, extract the fundamental mode of the example of similarity maximum as innovative design;

(3) customer demand is mapped to the product function module by fuzzy cluster analysis, compare with the functional module of the example that similarity is the highest in the step (2), find out the demand different and be mapped to the module that is influenced, thereby to need in the example to determine the module adjusted with example;

(4) if the module that does not satisfy the demands in the module library, but only need just can satisfy the demands by revising some existing module, then revising some existing module by the parametrization designing technique just can satisfy the demands, then carrying out module revises, deposit in the module library then with the corresponding module in the amended new module replacement example, and new module;

(5) if the module that does not satisfy the demands in the module library, and can not be by revising the module that existing module is met customer demand, then need to create new module, deposit in the module library then with the corresponding module in the new module replacement example of creating, and new module;

(6) the new product configuration scheme that produces of audit.Similar example has formed the new product scheme through after revising, and whether this product solution is met customer need and whether had the also further evaluation of needs of suitable technical parameter.Wherein technical parameter obtains by low-tension motor inverse time-lag protection new algorithm.

(7) output products scheme.The new product scheme is by after the audit, then can printout so that the guide product design.

(8) store the new product allocation plan into case library.Storage new product allocation plan can constantly expand the case library of product.

The above-mentioned description to embodiment is can understand and apply the invention for ease of those skilled in the art.The person skilled in the art obviously can easily make various modifications to these embodiment, and needn't pass through performing creative labour being applied in the General Principle of this explanation among other embodiment.Therefore, the invention is not restricted to the embodiment here, those skilled in the art are according to announcement of the present invention, and not breaking away from the improvement that category of the present invention makes and revise all should be within protection scope of the present invention.

Claims (4)

1. the method determined fast of a motor Protective Device, it is characterized in that: it may further comprise the steps:

1) based on Axiomatic Theories the motor Protective Device functional module is divided

According to Axiomatic Theories, known current constraint and motor Protective Device design knowledge base, be met the relational matrix of independent axiom, decompose motor Protective Device functional module and domain up to the bottom, but the relational matrix of the nothing that obtains adhering to specification coupling or decoupling zero;

2) based on Fuzzy Cluster Analysis method customer demand is shone upon to functional module

Fuzzy cluster analysis is divided into different monoids to customer demand according to different features, feature is client's essential characteristic, customer value feature and client's behavioural characteristic, by the relation property matrix of customer demand to the functional module mapping step 1) mesolow motor protecting device functional module and customer demand is connected;

3) dispose based on module instance reasoning motor Protective Device

Set up modular product case retrieval storehouse, it must be expressed in required various information in the configuration design process effectively, for determining that motor Protective Device provides case retrieval;

4) the utilization parametric technology utilizes dimension constraint and product components and parts key identifier with motor Protective Device universal circuit module parameterization;

5) low-tension motor inverse time-lag protection algorithm computation goes out low-tension motor inverse time-lag protection time parameter, in time protects action;

6) according to customer demand retrieval product example search library, obtain motor Protective Device, perhaps, if there is not then implementation step 1 of respective instance)-5) at least one revise the product example search library.

2. the method that motor Protective Device as claimed in claim 1 is determined fast; it is characterized in that: the relation property matrix described step 2) is not if meet the coupling mapping of independent axiom; should carry out decoupling zero, the clustering algorithm decoupling zero that it adopts best cluster numbers comprises:

Phase one is determined the best cluster numbers of clustering algorithm

Input: sample data

Output: sample cluster centre, the best cluster numbers of sample

Step 1 employing Euclidean distance formula calculates the Euclidean distance between all sample number strong points, generates the Euclidean distance matrix;

Step 2 adopts the sum of squares of deviations method, calculates the link distance between the sample data class, and generates the hierarchical cluster tree,

This step concretism is: be provided with N sample, N sample respectively is a class when initial, calculate distance between them according to sum of squares of deviations distance calculation formula then, to and be a new class apart from the class of minimum, the new class after calculating and the class and the distance of other classes will merge apart from two classes of minimum again, and so forth again, till N sample is merged into a class

Sum of squares of deviations distance calculation formula: $S=\underset{t=1}{\overset{k}{\mathrm{\&Sigma;}}}\underset{\mathrm{\&alpha;}=1}{\overset{N_t}{\mathrm{\&Sigma;}}}{(X_{\left(t\right)}^{\left(\mathrm{\&alpha;}\right)}-{\stackrel{\&OverBar;}{X}}_{\left(t\right)})}^{\&prime;}(X_{\left(t\right)}^{\left(\mathrm{\&alpha;}\right)}-{\stackrel{\&OverBar;}{X}}_{\left(t\right)})$

N sample is divided into k class: C ₁, C ₂..., C _k, in the formula: use

Expression C _tIn i sample, N _tExpression C _tIn number of samples,

Be C _tCenter of gravity;

Step 3 obtains different sample segmentation results according to link distance calculated value cutting hierarchical cluster tree;

Step 4 is added up value according to the F that F statistic formula calculates under the different segmentation results, and with it is calculated that the result, the corresponding optimal classification result of maximum F statistics value obtains best cluster numbers of sample data and corresponding cluster result thus;

F statistic formula: $F=\frac{\underset{j=1}{\overset{r}{\mathrm{\&Sigma;}}}{n}_j{\left|\right|{\stackrel{\&OverBar;}{x}}^{\left(j\right)}-\stackrel{\&OverBar;}{x}\left|\right|}^2/(r-1)}{\underset{j=1}{\overset{r}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{n_j}{\mathrm{\&Sigma;}}}{\left|\right|x_i^{\left(j\right)}-{\stackrel{\&OverBar;}{x}}^{\left(f\right)}\left|\right|}^2/(n-r)}$

In the formula:

For

With

Distance,

It is sample in the j class

With the center

Distance;

The clustering algorithm of the best cluster numbers of subordinate phase operative norm

Input: Weighting exponent m, sample data collection, algorithm iteration end condition ε, the best cluster numbers of sample

Output: cluster centre matrix V, fuzzy partition matrix U

The value that step 1 obtains after the phase one with execution algorithm is an initialization condition, and input sample data collection X, end condition ε＞0, Weighting exponent m;

Step 2 utilizes following formula to calculate sample and distances of clustering centers,

$d_A^2(x_j,v_i^{\left(l\right)})={(x_j-v_i^{\left(l\right)})}^{T}A(x_j-v_i^{\left(l\right)})$

In the formula: 1≤i≤c, 1≤j≤n;

Step 3 order

Revise U ^(l+1),

$u_{\mathrm{ij}}^{\left(l\right)}=\left\{\begin{array}{cc}\frac{1}{\underset{r=1}{\overset{c}{\mathrm{\&Sigma;}}}{d}_A^2(x_j,v_i)/{(x_j,v_r)}^{2m-1}}& d_{\mathrm{ij}}^2>0\\ 1& d_{\mathrm{ij}}^2=0\\ 0& k\&NotEqual;i,d_{\mathrm{jk}}^2=0\end{array}\right.;$

Step 4 couple U ^(l)Upgrade the cluster centre battle array

$v_i^{(l+1)}=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(l\right)}\right)}^m{x}_j/\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{\left(u_{\mathrm{ij}}^{\left(l\right)}\right)}^m$

In the formula: 1≤i≤c, U ^(l)Represent the degree of membership matrix after the iteration l time,

Be the cluster center of gravity after l+1 iteration;

Step 5 if || V ^(l+1)-V ^(l)||≤ε, then stop iteration, otherwise carry out l=l+1, jump procedure 2,

Determine low-voltage motor protector merit structure unit cluster feature coupling incidence matrix,

Clustering algorithm based on best cluster numbers carries out obtaining non-coupling mapping after the decoupling zero,

3. the method that motor Protective Device as claimed in claim 1 is determined fast, it is characterized in that: described step 3) comprises:

Example in the product example search library; Use C _k={ a _K1, a _K2..., a _Ki..., a _Kn, k=1,2 ..., m. represents that wherein m represents the instance number in the current case library, a _Ki(i=1,2 ..., n) be example C _kAttribute, represent with a ternary Vector Groups: a _Ki=(p _i, w _i, x _i), p wherein _iThe representation attribute name, w _iThe weighted value of representation attribute, x _iThe representation attribute value;

Client's product demand; Use Q ₀={ a ₀₁, a ₀₂..., a _0iExpression, according to customer demand Q ₀With example C _kCarry out similarity and calculate, retrieval meets the example of customer demand most;

Similarity is calculated, S (a wherein _0i.a _Ki) be object a _0iAnd a _KiSimilarity, similarity is 1 when their property value is identical, similarity is 0 in the time of different;

Relatively similarity is selected the most similar module instance, deposits the module library after the modification again in.

4. the method that motor Protective Device as claimed in claim 1 is determined fast, it is characterized in that: described step 4) comprises:

Motor Protective Device universal circuit to same function is concluded, and is divided into several classes, and in same building mode, components and parts model difference, the circuit module that shape is identical is classified as a class, determines to revise parameter with this;

Divide the result according to Axiomatic Design Theory, each universal circuit module that will be relevant with functional domain is carried out related effectively, constitute the motor Protective Device circuit, need to select the low-tension motor local circuit module of modification, according to aforementioned definite modification parameter the universal circuit module is made amendment, make that original localized design is reused;

After the instance parameter modification, the optimum example that the CAD (computer aided design) system was revised saves as new example, expands the product example search library.

Images