CN106326610A - Design network based design alterationpropagation predicting method and system - Google Patents

Abstract

The invention disclosesa design network based design alteration propagation predicting method and system. For eliminating the negative influence on the cost and efficiency of frequent design alteration in the product design process, predicting an alteration propagation path in advance and inhibiting unnecessary alteration before design implementation, and the purpose of narrowing an alteration propagation influence range is achieved by utilizing the optimal propagation path for technical analysis design of complicated network analysisbased on an established relational network model about design attributes. By the adoption of the method, a fairly objective designalteration analysis network model can be established, and thepropagation predicting accuracy is ensured fundamentally. In addition, an alteration propagation strength mathematic model is quantitatively established based on a complicated network analysis technology.

Description

A kind of design alteration propagation prediction method based on planned network and system

Technical field

The present invention relates to a kind of design alteration propagation prediction method based on planned network and system.

Background technology

In R & D of complex, design requirement may change, and the availability of resource and assembly can change, design Losing efficacy and erroneous decision can often occur, design information is it may happen that careless mistake in exchange process simultaneously, thus causes design to become More inevitable.Product design change management is significant for manufacturing enterprise.Product design change is to technological artifacts In structure (including cooperation, shape, size, surface, material etc.), performance (including stability, intensity, corrosion resistance etc.), merit Energy (including speed, efficiency etc.) or performance-function (i.e. design principle) relation, performance-structure (i.e. physics law) relation are done Amendment.Design alteration can frequently occur in the development of design continuously to product particularly complex product, and determines The ultimate cost of product 70-80%.

Design alteration management can help the probability of business forcast change propagation, reduces unnecessary change, has selected Effect change propagation path and as early as possible implement change.The core of design alteration management is change propagation prediction and control method.If Meter change propagation is a process, and a certain parts or unit generation during i.e. current system configures or designs are changed thus caused and be System changes generation further or more.It is predicted change propagation supporting effective design decision and engineering rule Draw, it is to avoid the subsystem that variable cost is high, it is possible in design initial stage quantitative predication workload and the variation of product cost.

The existing research relating to design alteration propagation forecast remains in following both sides limitation and weak point:

(1) need to be manually specified change propagation sensitivity and influence degree between design attributes during modeling, there is the strongest master The property seen, thus reduce the precision of prediction of change propagation.Such as, it is left that the most frequently used change propagation method precision of prediction is only 30% Right.

(2) lack in attribute granularity, carry out design alteration Communication Research.

Summary of the invention

The purpose of the present invention is contemplated to solve the problems referred to above, it is provided that a kind of design alteration based on planned network is propagated pre- Survey method and system, set up change propagation intensity mathematical model by Complex Networks Analysis technology, improve design alteration and propagate pre- The accuracy surveyed, the change propagation probability between design attributes can be defined, consider design simultaneously by the method objectively Nargin in parameter, reduces change propagation coverage while analyzing optimum change propagation path.

To achieve these goals, the present invention adopts the following technical scheme that

A kind of design alteration propagation prediction method based on planned network, comprises the steps:

Step (1): set up design alteration and analyze network model；

Step (2): calculate the change propagation probability between the design attributes of storage in product design change data base；To set Change propagation probability between meter attribute analyzes the weight connected in network model as design alteration；Thus foundation is more objective Design alteration analyze network model；

Step (3): utilize probability of spreading, the number of degrees of design attributes, design attributes change nargin, weight and long-chain to connect Penalty coefficient set up design alteration transmission intensity model, calculate design attributes change propagation intensity；

Step (4): carry out change propagation path optimization: using change propagation intensity as object function, pass through ant group optimization Algorithm obtains design alteration and analyzes the change propagation path of the corresponding largest cumulative change propagation intensity minimized on network, described Change propagation path is optimum propagation path.

Described step (3) change propagation intensity is the biggest, and the connected amount of change needed for Subsequent attributes is the biggest.

The step of described step (1) is:

It is expressed as internodal connection as network node, the annexation between father and son's design attributes using design attributes, Gradually refine according to design attributes is top-down, ultimately form a design alteration and analyze network model；Belong to Same Part Annexation between design attributes is defined as short chain, and the annexation between the design attributes of part is defined as long-chain.

Described step (1):

With set V={v₁,v₂,...,v_nCorresponding design attributes collection, and to define E be internodal to connect collection, design alteration Analyze network model and be denoted as G={V, E}；

Design attributes is expressed as circle, and annexation is expressed as the line of band arrow, and design attributes is clustered according to affiliated part, Every class is distinguished by dotted line；Annexation includes parameter annexation and constraint annexation two class.

Described parameter annexation is expressed as functional form y=f (x₁,x₂,...,x_n), wherein y is father's design attributes, x_i Represent the sub-design attributes of i-th.In network model is analyzed in design alteration, arrow is pointed to father's design attributes by sub-design attributes.One Item parameter annexation is made up of father's design attributes and many sub-design attributes, and the value of father's design attributes is by separate Sub-design attributes value determine,

$dy=(\∂f/\∂x_i)dx;$

Analyze in network model in design alteration, separate due between sub-design attributes, the change of sub-design attributes The most directly change the fraternal attribute of sub-design attributes, but indirectly change sub-design attributes by change father's design attributes Brother's attribute；The change of father's design attributes realizes by changing its sub-design attributes, and the size of amount of change is by being manually set；Cause This, in parameter annexation, father's design attributes changes along with the change of arbitrary sub-design attributes；When father's design attributes changes Time, if father's design attributes has a sub-design attributes, sub-design attributes changes, if father's design attributes has many height and sets Meter attribute, changing according to different setting change weights of every sub-design attributes.

Described constraint annexation is that the assembly constraint in order to meet between parts, function coordinate and ensure performance and people For formulate design rule, with function f (x₁,x₂,...,x_n) >=0 represents；

Each design attributes in constraint annexation has identical dimension, adopts in network model is analyzed in design alteration Represent with four-headed arrow；

Constraint annexation is when the change of arbitrary constrained attributes, and other constrained attributes are also made change and could be continued to ensure Connector closes restriction relation；

In constraint annexation, the change of any attribute all can cause the change of other connection attributes, changes weight and is It is manually set.

Described step (2):

Design alteration is analyzed in network model, and each connection has a weight, i.e. probability of spreading P_ij。P_ijIt is current Drive follow-up attribute when attribute i occurs to change and the probability of change occurs.P_ijIt is equal in design alteration data base at a record Middle design attributes i occurs on the premise of the conditional probability of attribute j occurs, i.e.

$P_{ij}=P\left(v_j\right|v_i)=\frac{P(v_i\∩v_j)}{P\left(v_i\right)}=P\left(v_i\right|v_j)\frac{P\left(v_j\right)}{P\left(v_i\right)}=P_{ji}\frac{P\left(v_j\right)}{P\left(v_i\right)}---\left(1\right)$

In formula (1), due to P (v_i) and P (v_j) unequal, P_ijAnd P_jiUnequal.If do not connected between two design attributes Connect, P_ijEqual to 0.And the probability of spreading sum of design attributes i adjacent attribute all with design attributes i is 1.

$\underset{j\∈F}{\Σ}{P}_{ij}=1---\left(2\right)$

P(v_j|v_i) meaning of parameters be in design alteration data base one record in design attributes i occurs on the premise of The conditional probability of attribute j occurs；P(v_i∩v_j) meaning of parameters be design attributes i and j simultaneously appears in design alteration data base Article one, the probability in record；P(v_i) meaning of parameters be that design alteration data base occurs the probability of design attributes i；P(v_j) Meaning of parameters is to occur the probability of design attributes j in design alteration data base；P(v_i|v_j) meaning of parameters be design alteration data The conditional probability of attribute i occurs on the premise of storehouse design attributes j occurs in a record；P_jiMeaning of parameters be current Drive follow-up attribute i when attribute j occurs to change and the probability of change occurs.

Described step (3):

When kth time is propagated, design attributes change propagation intensityIt is defined as:

$I_{ij}^k=\left\{\begin{array}{cc}0& {\mathrm{\Δ\ρ}}_i^k\≤{\mathrm{\ρ}}_i\\ {\mathrm{\ω}}_s\[{\mathrm{\ω}}_p(1-P_{ij})+{\mathrm{\ω}}_d\frac{d_j}{\underset{j\∈F^k}{\Σ}{d}_j}\](1-\frac{{\mathrm{\ρ}}_i}{{\mathrm{\Δ\ρ}}_i^k})& {\mathrm{\Δ\ρ}}_i^k>{\mathrm{\ρ}}_i\end{array}\right.,i\∈F^{k-1}---\left(3\right)$

Wherein, ρ_iIt is the change nargin when kth time is propagated of design attributes i,Passing in kth time for design attributes i Required change sowing time value；F^kRepresent the design attributes set affected by kth time change propagation；ω_pIt it is the weight of probability of spreading Value, ω_dIt is the weighted value of the design attributes number of degrees, and ω_p+ω_d=1；d_jIt is the number of degrees of design attributes, by the neighbour of design attributes Connect matrix calculus to obtain；ω_sIt is the penalty coefficient of long-chain connection, wherein ω_s>=1, ω_sIncreasing when part is propagated for change Add change propagation intensity.

Described step (4):

Analyze network model based on design alteration and change propagation strength model sets up change propagation path optimization target letter Number is as follows:

$\left.\begin{array}{cc}\arg \min \left(\max \underset{k}{\&Sigma;}{I}_{ij}^k\right)& k=1,2,...,N\\ s.t.({\mathrm{\&Delta;\&rho;}}_j^k-{\mathrm{\&rho;}}_j)<{10}^{-5}& j\&Element;F^k\end{array}\right\}---\left(4\right)$

Formula (4) defines initial designs attribute and changes total transmission intensity when other design attributes backward are propagated.When setting Meter nargin ρ_jAnd design variableDifference less than 10^-5Time, illustrate that change propagation convergence does not further affect other The change of design attributes produces.Every time after iteration, the pheromone that the ant colony in connection is left over updates according to following rule:

${\mathrm{\&tau;}}_{ij}=(1-\mathrm{\&rho;}){\mathrm{\&tau;}}_{ij}+\underset{l=1}{\overset{N_a}{\&Sigma;}}{\mathrm{\&Delta;\&tau;}}_{ij}^l---\left(5\right)$

Wherein, τ_ijBeing the pheromone value being delivered to the release of design attributes j from design attributes i, ρ is pheromone volatility coefficient, Wherein, 0 < ρ < 1；N_aIt is ant colony number,It it is the pheromone value of the l Formica fusca release.τ_ijExpression converts to other possible states Attraction degree；

Wherein, Q is constant, D_lIt it is the target function value of the l Formica fusca migration；

Change is tended to selection and is had maximumAdjacent design attributes propagate, condition conversion wish η_ijDirectly related InTherefore, condition conversion wish η_ijIt is defined as follows:

${\mathrm{\&eta;}}_{ij}=I_{ij}^k---\left(7\right)$

Every time after iteration, condition conversion wish η_ijAccording to currentValue is updated.The l Formica fusca is from design attributes i Climb to the probability of design attributes jFor

Wherein, α is for controlling τ_ijThe parameter of influence degree, α >=0, β is to control η_ijThe parameter of influence degree, β >=1；It is The l Formica fusca is directly connected to property set in design attributes i.

A kind of design alteration propagation prediction method based on planned network, also includes step (5): propagate optimum results and show Module, shows design alteration propagation path optimum results in the user interface.

Have in step (5) display: connection, change propagation intensity level and change between design attributes, design attributes pass Broadcast path.

Described product design change data base: for design Storage change record, described design alteration record includes: design Attribute sequence number, forward direction design attributes sequence number, attribute-name, be worth before changing, change after value and Date Of Change.

Described design alteration record is used for quantitative Analysis P_ij。

A kind of design alteration propagation forecast system based on planned network, including:

Network model is analyzed in design alteration: using design attributes as network node, by the connection between father and son's design attributes Relational representation is internodal connection；The annexation belonged between the design attributes of Same Part is defined as short chain, across part Annexation be defined as long-chain；Distinguishing annexation according to attribute can be divided into again parameter annexation and constraint to connect pass System；

Product design change data base: be responsible for storage product design change occur after each affected design attributes and Influence degree, for calculating the propagation effect probability between each design attributes；

Probability of spreading computing module: by calculating the change propagation probability between design attributes, as design alteration analysis The weight connected in network model；Set up objective design alteration and analyze network model；

Change propagation intensity analysis module: utilize the number of degrees of each design attributes, design margin, the weight of connection and long-chain The penalty coefficient connected sets up design alteration transmission intensity model；

Change propagation path optimization module: using accumulation change propagation intensity as object function, pass through ant colony optimization algorithm Obtain the change propagation optimal path of the largest cumulative change propagation intensity that correspondence minimizes；

Propagate optimum results display module: show in the user interface by design alteration propagation path optimum results, display Inside have: the connection between design attributes, design attributes, change propagation intensity level, change propagation path.

Beneficial effects of the present invention:

Set up fine granularity design alteration based on design attributes and relation thereof and analyze network model, from product design change number According to storehouse is excavated the weighted value that change propagation probability connects as this model, analyze net based on this relatively objective design alteration Network model ensure that the accuracy of propagation forecast；Set up change propagation intensity mathematical modulo by Complex Networks Analysis technology simultaneously Type, and consider the nargin of design parameter, it is possible to obtain optimum change propagation.

Some conflict phenomenons after design alteration often do not observe under the large scale rank such as system, parts, but are put into Just become abnormal obvious under the small size rank such as attribute, parameter.Carrying out change propagation analysis in dependence granularity can be objective, fixed Calculate to amount fault tolerance and the change propagation influence degree of assembly, make change on purpose along the path that fault tolerance is strong Propagate, thus make design alteration communication process restrain as early as possible.

Accompanying drawing explanation

Fig. 1 is the frame construction drawing of design alteration propagation prediction method based on design attributes net；

Network model's exemplary plot is analyzed in Fig. 2 design alteration；

Fig. 3 change propagation path optimizing exemplary plot.

Detailed description of the invention

The invention will be further described with embodiment below in conjunction with the accompanying drawings.

A kind of design alteration propagation forecast system based on planned network, including:

Network model is analyzed in design alteration: this model is using design attributes as network node, between father and son's design attributes Annexation be expressed as internodal connection；The annexation belonged between the design attributes of Same Part is defined as short chain, Annexation across part is defined as long-chain；Distinguish annexation according to attribute and can be divided into again parameter annexation and constraint even Connect relation；This network model is easy to be designed by Complex Networks Analysis technology the fast quantification of change propagation coverage and divides Analysis；

Product design change data base: be responsible for storage product design change occur after each affected design attributes and Influence degree, for evaluating the propagation effect probability between each design attributes；

Probability of spreading evaluation module: this module, by the record in Analysis on Data Mining design alteration data base, is commented Change propagation probability between valency design attributes, analyzes the weight connected in network model in this, as design alteration；As Network model is analyzed in the relatively objective design alteration of Foundation；

Change propagation intensity analysis module: consider the number of degrees of each design attributes, design margin, the weight of connection and long-chain The penalty coefficient connected sets up design alteration transmission intensity model, propagates, for design alteration, the evaluation number that optimizing provides quantitative；

Change propagation path optimization module: using accumulation change propagation intensity as object function, calculated by heuristic optimization Method obtains the change propagation optimal path of the largest cumulative change propagation intensity that correspondence minimizes；

Propagate optimum results display module: show in the user interface by design alteration propagation path optimum results, display Inside have: the connection between design attributes, design attributes, change propagation intensity level, change propagation path；

In above description, primary claim module is: network model, probability of spreading evaluation module, change are analyzed in design alteration More transmission intensity analyzes module and change propagation path optimization module.

The present invention mainly utilize Java language, MySQL database,Ontology editor realizes.As it is shown in figure 1, this Invention specifically includes that network model, product design change data base, probability of spreading evaluation module, change biography are analyzed in design alteration Broadcast intensity analysis module and change propagation path optimization module.

One, network model is analyzed in design alteration is according to design theory, by design attributes from top under gradually refinement, End form becomes a big Attribute Association network.This model is using design attributes as network node, between father and son's design attributes Annexation is expressed as internodal connection, such as Fig. 2.With set V={v₁,v₂,...,v_nCorresponding design attributes collection, and define E Collection is connected for internodal.Network model is analyzed in this design alteration can be denoted as G={V, E}.Belong to the design attributes of Same Part Between annexation be defined as short chain, the annexation across part is defined as long-chain.Distinguish annexation according to attribute again may be used The fast quantification being designed change propagation coverage by Complex Networks Analysis technology to be divided into this network model to be easy to divides Analysis.

In Fig. 2, design attributes is expressed as circle, and annexation is expressed as the line of band arrow, and design attributes can be according to part Ownership is clustered, and every class is distinguished by dotted line.Wherein annexation includes parameter annexation and constraint annexation two class.Ginseng Number annexation is set up according to the physical law in design theory, and such connection can be collectively expressed as functional form y=f (x₁, x₂,...,x_n), wherein y is father's design attributes, x_iRepresent the sub-design attributes of i-th.In network model's figure is analyzed in design alteration Arrow is pointed to father's design attributes by sub-design attributes.One parameter annexation is to be designed by father's design attributes and many height Attribute forms, and the value of father's design attributes is determined by the value of separate sub-design attributes, and i.e. father's design attributes value is according to physics Law changes along with the change of sub-design attributes value,In network model is analyzed in design alteration, due to Separate between sub-design attributes, the change of sub-design attributes does not directly affect its fraternal attribute, but passes through shadow Ring other brother's attributes of his father's design attributes remote-effects.Change father's design attributes can not directly occur, and needs to pass this change Passing sub-design attributes, the size of amount of change needs the balance of designer.Therefore, in parameter annexation, father designs genus Property changes along with the change of arbitrary sub-design attributes；When father's design attributes changes, its unique sub-design attributes necessarily becomes More, and when it has many sub-design attributes, the change degree difference of every sub-design attributes.

Constraint annexation is the design rule artificially formulated due to assembling, function matching, performance guarantee, uses function f (x₁,x₂,...,x_n) >=0 represents.Each design attributes in constraint annexation has identical dimension, divides in design alteration Analysis network model figure uses four-headed arrow to represent.Constraint annexation is likely difficult to meet when the change of certain constrained attributes, Therefore other constrained attributes need to make to change and could continue to ensure that connector closes restriction relation.In constraint annexation, appoint The change of one attribute all can cause the change of other connection attributes, change degree to need artificial appointment.

Two, probability of spreading evaluation module is by the record in Analysis on Data Mining design alteration data base, and evaluation sets Change propagation probability between meter attribute, analyzes the weight connected in network model in this, as design alteration；Based on this Set up relatively objective design alteration and analyze network model.Design alteration is analyzed in network model, and each connection has one Weight, i.e. probability of spreading P_ij。P_ijIt it is the probability that follow-up attribute occurs change when forerunner attribute i occurs change.This probability can To be obtained by the change record in data mining design alteration data base.P_ijIt is at a record in design alteration data base Middle design attributes i occurs on the premise of the conditional probability of attribute j occurs, i.e.

$P_{ij}=P\left(v_j\right|v_i)=\frac{P(v_i\&cap;v_j)}{P\left(v_i\right)}=P\left(v_i\right|v_j)\frac{P\left(v_j\right)}{P\left(v_i\right)}=P_{ji}\frac{P\left(v_j\right)}{P\left(v_i\right)}---\left(1\right)$

In formula (4), due to P (v_i) and P (v_j) the most unequal, P_ijAnd P_jiUnequal.If do not had between two design attributes Connect, P_ijEqual to 0.And design attributes i and its all adjacent attributes probability of spreading sum be 1.

$\underset{j\&Element;F}{\&Sigma;}{P}_{ij}=1---\left(2\right)$

Three, change propagation intensity analysis module considers the number of degrees of each design attributes, design margin, the weight of connection and length The penalty coefficient that chain connects sets up design alteration transmission intensity model, propagates, for design alteration, the evaluation that optimizing algorithm provides quantitative Index.

Due to the coupled relation between design attributes, the value of the most a certain design attributes changes, and this change can progressively It is broadcast to other design attributes.In network model is analyzed in design alteration, the transmission intensity between design attributes and probability of spreading Directly related.During change propagation, change can preferably have the connection of bigger probability of spreading and propagate.Accumulative propagation is general Rate is persistently carried out and exponential decrease along with propagate.Analyze network model due to design alteration and often there is small-world network genus Property, it is all the important factor in order for change propagation except probability of spreading, the number of degrees of node and long-chain connect.The number of degrees of node The biggest, having the most adjacent attribute is affected by change propagation.For avoiding changing steady spread, select in change propagation path The design attributes with relatively lordotic should be avoided during selecting as far as possible.Simultaneously, it is considered to the topological attribute of small-world network, difference sets There is a small amount of long-chain between meter part to connect.Analyze in network model from design alteration and account for most short chains and connect different It is that long-chain connects the change propagation affected dramatically between design parts, needs to give bigger propagation penalty coefficient. It addition, nargin can absorb all or part of amount of change present in existing design parameter, absorption is played for change propagation Or the effect of buffering.Design margin is the limit allowable value difference with currency.Distinguishing from the angle of design margin, design attributes can To be divided into absorption node, buffer joint and transmission node.Therefore, introduce the impact of change propagation quantification of intensities change propagation, Connect and design margin including probability of spreading, the node number of degrees, long-chain.

When kth time is propagated, change propagation strength definition is:

$I_{ij}^k=\left\{\begin{array}{cc}0& {\mathrm{\&Delta;\&rho;}}_i^k\&le;{\mathrm{\&rho;}}_i\\ {\mathrm{\&omega;}}_s\&lsqb;{\mathrm{\&omega;}}_p(1-P_{ij})+{\mathrm{\&omega;}}_d\frac{d_j}{\underset{j\&Element;F^k}{\&Sigma;}{d}_j}\&rsqb;(1-\frac{{\mathrm{\&rho;}}_i}{{\mathrm{\&Delta;\&rho;}}_i^k})& {\mathrm{\&Delta;\&rho;}}_i^k>{\mathrm{\&rho;}}_i\end{array}\right.,i\&Element;F^{k-1}---\left(3\right)$

Wherein ρ_iIt is the change nargin when kth time is propagated of design attributes i, Δ ρ^k _iFor required change value；F^kRepresent The design attributes set affected by kth time change propagation；ω_pAnd ω_dBe respectively probability of spreading and the weighted value of the node number of degrees and ω_p+ω_d=1；d_jIt is that the node number of degrees can be calculated by the adjacency matrix of design attributes；ω_s(ω_s>=1) it is that long-chain connects Penalty coefficient, for change when part is propagated increase change propagation intensity.

Change propagation intensity is the biggest, and the connected amount of change needed for Subsequent attributes is the biggest.In other words, strong along propagating Spend big connection to carry out change propagation and can affect Subsequent attributes to a greater degree.

Four, change propagation path optimization module is using accumulation change propagation intensity as object function, passes through heuristic optimization Algorithm obtains the change propagation optimal path of the largest cumulative change propagation intensity that correspondence minimizes.The change of design attributes is permissible Propagated by different paths, even if the attribute of some design attributes different variation is joined directly together also can be affected by. In order to prevent the variation of key Design attribute or produce the change propagation of avalanche type, need to make rational planning for change before implementing change Propagation path more.Owing to complexity and the mobility of change propagation strength model of network model is analyzed in design alteration, need Come for change propagation optimum path search by heuristic value.In the present invention, ant colony optimization algorithm is used to obtain change propagation Path optimizing.This propagation path has the largest cumulative change propagation intensity minimized.Net is analyzed based on above design alteration It is as follows that network model and change propagation strength model set up change propagation path optimization object function:

$\left.\begin{array}{cc}\arg \min \left(\max \underset{k}{\&Sigma;}{I}_{ij}^k\right)& k=1,2,...,N\\ s.t.({\mathrm{\&Delta;\&rho;}}_j^k-{\mathrm{\&rho;}}_j)<{10}^{-5}& j\&Element;F^k\end{array}\right\}---\left(4\right)$

Above-mentioned formula defines initial designs attribute and changes total transmission intensity when other design attributes backward are propagated.When setting The difference of meter nargin and design variable is less than 10^-5Time, change propagation convergence does not further affect other design attributes Change produces.Every time after iteration, the pheromone in connection updates according to following rule:

τ_ij←(1-ρ)τ_ij+Δτ_ij (5)

${\mathrm{\&Delta;\&tau;}}_{ij}=\underset{l=1}{\overset{N_a}{\&Sigma;}}{\mathrm{\&Delta;\&tau;}}_{ij}^l---\left(6\right)$

Wherein, τ_ijIt it is the pheromone value being delivered to j release from i；ρ (0 < ρ < 1) is pheromone volatility coefficient；N_aIt it is ant colony Number, Δ τ^l _ijIt it is the pheromone value of the l Formica fusca release.

Wherein Q is constant, D_lIt it is the target function value of the l Formica fusca migration.

Change is tended to selection and is had maximumAdjacent design attributes propagate, condition conversion wish η_ijDirectly related InTherefore, η_ijIt is defined as follows:

${\mathrm{\&eta;}}_{ij}=I_{ij}^k---\left(8\right)$

Every time after iteration, η_ijAccording to currentValue is updated.The l Formica fusca climbs to design attributes j from design attributes i Probability be

Wherein, α (0≤α) is for controlling τ_ijThe parameter of influence degree, β (β >=1) is to control η_ijThe parameter of influence degree.N^l _i It it is the l Formica fusca adjacent segments point set allowable in design attributes i.τ_ijAnd η_ijRepresent the attraction journey converted to other possible states Degree and departure degree.

Five, product design change data base: according to designer, product is modified record, on the one hand by setting up number According to storehouse by change record according to design attributes sequence number, forward direction design attributes sequence number, attribute-name, be worth before changing, change after be worth, change Dates etc. organize, it is simple to organize design alteration and evaluate；On the one hand, setting up change affects forward index, the most right Answer the occurrence number of other design attributes of each design attributes, it is simple to be quickly calculated probability of spreading.

Six, optimum results display module is propagated: show in the user interface by design alteration propagation path optimum results, with Certain design attributes as initially changing node, its propagation path optimum results example such as Fig. 3 obtained according to the present invention.In display Have: the connection between design attributes, design attributes, change propagation intensity level, change propagation path.

Although the detailed description of the invention of the present invention is described by the above-mentioned accompanying drawing that combines, but not the present invention is protected model The restriction enclosed, one of ordinary skill in the art should be understood that on the basis of technical scheme, and those skilled in the art are not Need to pay various amendments or deformation that creative work can make still within protection scope of the present invention.

Claims (10)

1. a design alteration propagation prediction method based on planned network, is characterized in that, comprise the steps:

Step (1): set up design alteration and analyze network model；

Step (2): calculate the change propagation probability between the design attributes of storage in product design change data base；Design is belonged to Change propagation probability between property analyzes the weight connected in network model as design alteration；Thus set up and more objectively set Meter analysis on altered project network model；

Step (3): utilize punishing of probability of spreading, the number of degrees of design attributes, design attributes change nargin, weight and long-chain connection Penalty factor sets up design alteration transmission intensity model, calculates design attributes change propagation intensity；

Step (4): carry out change propagation path optimization: using change propagation intensity as object function, pass through ant colony optimization algorithm Obtain design alteration and analyze the change propagation path of the corresponding largest cumulative change propagation intensity minimized, described change on network Propagation path is optimum propagation path.

A kind of design alteration propagation prediction method based on planned network, is characterized in that, also include Step (5): propagate optimum results display module, design alteration propagation path optimum results is shown in the user interface.

A kind of design alteration propagation prediction method based on planned network, is characterized in that, described step Suddenly the step of (1) is:

It is expressed as internodal connection as network node, the annexation between father and son's design attributes using design attributes, according to Design attributes is top-down gradually to be refined, and ultimately forms a design alteration and analyzes network model；Belong to the design of Same Part Annexation between attribute is defined as short chain, and the annexation between the design attributes of part is defined as long-chain.

A kind of design alteration propagation prediction method based on planned network, is characterized in that,

Described step (1):

With set V={v₁,v₂,...,v_nCorresponding design attributes collection, and to define E be internodal to connect collection, design alteration analysis Network model is denoted as G={V, E}；

Design attributes is expressed as circle, and annexation is expressed as the line of band arrow, and design attributes is clustered according to affiliated part, every class Distinguished by dotted line；Annexation includes parameter annexation and constraint annexation two class.

A kind of design alteration propagation prediction method based on planned network, is characterized in that,

Described parameter annexation is expressed as functional form y=f (x₁,x₂,...,x_n), wherein y is father's design attributes, x_iRepresent the I sub-design attributes；In network model is analyzed in design alteration, arrow is pointed to father's design attributes by sub-design attributes；One parameter Annexation is made up of father's design attributes and many sub-design attributes, and the value of father's design attributes is set by separate son The value of meter attribute determines,

$dy=(\&part;f/\&part;x_i)dx;$

Analyzing in network model in design alteration, separate due between sub-design attributes, the change of sub-design attributes is also The most directly change the fraternal attribute of sub-design attributes, but indirectly changed the brother of sub-design attributes by change father's design attributes Attribute；The change of father's design attributes realizes by changing its sub-design attributes, and the size of amount of change is by being manually set；Therefore, In parameter annexation, father's design attributes changes along with the change of arbitrary sub-design attributes；When father's design attributes changes, If father's design attributes has a sub-design attributes, sub-design attributes changes, if father's design attributes has the design of many height Attribute, changing according to different setting change weights of every sub-design attributes.

A kind of design alteration propagation prediction method based on planned network, is characterized in that,

Described constraint annexation is that the assembly constraint in order to meet between parts, function coordinate and ensure that performance is artificially made Fixed design rule, with function f (x₁,x₂,...,x_n) >=0 represents；

Each design attributes in constraint annexation has identical dimension, uses double in network model is analyzed in design alteration Represent to arrow；

Constraint annexation is when the change of arbitrary constrained attributes, and other constrained attributes are also made change and could be continued to ensure to connect Meet restriction relation；

In constraint annexation, the change of any attribute all can cause the change of other connection attributes, and change weight is artificial Set.

A kind of design alteration propagation prediction method based on planned network, is characterized in that,

Described step (2):

Design alteration is analyzed in network model, and each connection has a weight, i.e. probability of spreading P_ij；P_ijIt is when forerunner's attribute There is the probability of follow-up attribute generation change during change in i；P_ijOccur in a record equal in design alteration data base The conditional probability of attribute j occurs, i.e. on the premise of design attributes i

$P_{ij}=P\left(v_j\right|v_i)=\frac{P(v_i\&cap;v_j)}{P\left(v_i\right)}=P\left(v_i\right|v_j)\frac{P\left(v_j\right)}{P\left(v_i\right)}=P_{ji}\frac{P\left(v_j\right)}{P\left(v_i\right)}---\left(1\right)$

In formula (1), due to P (v_i) and P (v_j) unequal, P_ijAnd P_jiUnequal；If do not connected between two design attributes, P_ij Equal to 0；And the probability of spreading sum of design attributes i adjacent attribute all with design attributes i is 1；

$\underset{j\&Element;F}{\&Sigma;}{P}_{ij}=1---\left(2\right)$

P(v_j|v_i) meaning of parameters be that design alteration data base occurs in a record occur on the premise of design attributes i The conditional probability of attribute j；P(v_i∩v_j) meaning of parameters be design attributes i and j simultaneously appears in design alteration data base one Probability in record；P(v_i) meaning of parameters be that design alteration data base occurs the probability of design attributes i；P(v_j) parameter It is meant that the probability that design attributes j occurs in design alteration data base；P(v_i|v_j) meaning of parameters be in design alteration data base The conditional probability of attribute i occurs on the premise of design attributes j occurring in a record；P_jiMeaning of parameters be when forerunner belong to Property j occur change time follow-up attribute i occur change probability.

A kind of design alteration propagation prediction method based on planned network, is characterized in that,

Described step (3):

When kth time is propagated, design attributes change propagation intensityIt is defined as:

$I_{ij}^k=\left\{\begin{array}{cc}0& {\mathrm{\&Delta;\&rho;}}_i^k\&le;{\mathrm{\&rho;}}_i\\ {\mathrm{\&omega;}}_s\&lsqb;{\mathrm{\&omega;}}_p(1-P_{ij})+{\mathrm{\&omega;}}_d\frac{d_j}{\underset{j\&Element;F^k}{\&Sigma;}{d}_j}\&rsqb;(1-\frac{{\mathrm{\&rho;}}_i}{{\mathrm{\&Delta;\&rho;}}_i^k})& {\mathrm{\&Delta;\&rho;}}_i^k\&le;{\mathrm{\&rho;}}_i\end{array}\right.,i\&Element;F^{k-1}---\left(3\right)$

Wherein, ρ_iIt is the change nargin when kth time is propagated of design attributes i,For design attributes i when kth time is propagated Required change value；F^kRepresent the design attributes set affected by kth time change propagation；ω_pIt is the weighted value of probability of spreading, ω_d It is the weighted value of the design attributes number of degrees, and ω_p+ω_d=1；d_jIt is the number of degrees of design attributes, by the adjacency matrix of design attributes It is calculated；ω_sIt is the penalty coefficient of long-chain connection, wherein ω_s>=1, ω_sWhen part is propagated, change is being increased for change Transmission intensity.

A kind of design alteration propagation prediction method based on planned network, is characterized in that,

Described step (4):

Analyze network model based on design alteration and change propagation strength model sets up change propagation path optimization object function such as Under:

$\left.\begin{array}{cc}\arg \min \left(\max \underset{k}{\&Sigma;}{I}_{ij}^k\right)& k=1,2,\mathrm{...},N\\ s.t.({\mathrm{\&Delta;\&rho;}}_j^k-{\mathrm{\&rho;}}_j)<{10}^{-5}& j\&Element;F^k\end{array}\right\}---\left(4\right)$

Formula (4) defines initial designs attribute and changes total transmission intensity when other design attributes backward are propagated；When design is abundant Degree ρ_jAnd design variableDifference less than 10^-5Time, illustrate that change propagation convergence the most further affects other designs and belongs to Property change produce；Every time after iteration, the pheromone that the ant colony in connection is left over updates according to following rule:

${\mathrm{\&tau;}}_{ij}=(1-\mathrm{\&rho;}){\mathrm{\&tau;}}_{ij}+\underset{l=1}{\overset{N_a}{\&Sigma;}}{\mathrm{\&Delta;\&tau;}}_{ij}^l---\left(5\right)$

Wherein, τ_ijBeing the pheromone value being delivered to the release of design attributes j from design attributes i, ρ is pheromone volatility coefficient, wherein, 0<ρ<1；N_aIt is ant colony number,It it is the pheromone value of the l Formica fusca release；τ_ijRepresent the attraction converted to other possible states Degree；

Wherein, Q is constant, D_lIt it is the target function value of the l Formica fusca migration；

Change is tended to selection and is had maximumAdjacent design attributes propagate, condition conversion wish η_ijIt is directly related to Therefore, condition conversion wish η_ijIt is defined as follows:

${\mathrm{\&eta;}}_{ij}=I_{ij}^k---\left(7\right)$

Every time after iteration, condition conversion wish η_ijAccording to currentValue is updated；The l Formica fusca climbs to from design attributes i The probability of design attributes jFor

Wherein, α is for controlling τ_ijThe parameter of influence degree, α >=0, β is to control η_ijThe parameter of influence degree, β >=1；It it is l Formica fusca is directly connected to property set in design attributes i.

10. a design alteration propagation forecast system based on planned network, is characterized in that, including:

Network model is analyzed in design alteration: using design attributes as network node, by the annexation between father and son's design attributes It is expressed as internodal connection；The annexation belonged between the design attributes of Same Part is defined as short chain, across the company of part Connecing contextual definition is long-chain；Distinguish annexation according to attribute and can be divided into again parameter annexation and constraint annexation；

Product design change data base: be responsible for each affected design attributes and impact after storage product design change occurs Degree, for calculating the propagation effect probability between each design attributes；

Probability of spreading computing module: by calculating the change propagation probability between design attributes, analyze network as design alteration The weight connected in model；Set up objective design alteration and analyze network model；

Change propagation intensity analysis module: utilize the number of degrees of each design attributes, design margin, the weight of connection and long-chain to connect Penalty coefficient set up design alteration transmission intensity model；

Change propagation path optimization module: using accumulation change propagation intensity as object function, obtained by ant colony optimization algorithm The change propagation optimal path of the largest cumulative change propagation intensity that correspondence minimizes；

Propagate optimum results display module: show in the user interface by design alteration propagation path optimum results, show content Have: the connection between design attributes, design attributes, change propagation intensity level, change propagation path.