CN100545840C - A kind of punching part sample researching method - Google Patents

Abstract

A kind of punching part sample researching method for making full use of the successful experience of Sheet Metal Forming Technology, from the angle of resemble process, adopts nearest neighbor method to carry out the stamping parts case retrieval.Concrete, at first from the stamping parts three-dimensional model, extract between feature and feature and retrain, and according to stamping parts material information table, the material that obtains material is thick, intensity and hardness; From the angle of resemble process, calculate the similarity of new and old stamping parts example then; The similarity of the new and old stamping parts example of described calculating is to carry out in two stages of branch: the phase one is that benchmark calculates stamping parts topological relation similarity with the plane, filters out dissimilar old example, obtains similar old example, to dwindle the search space of follow-up phase; Subordinate phase by the comprehensive similarity of feature, constraint and material between new and old example, draws an old example the most similar from a collection of old example of phase one acquisition.This method has the realization of being easy to, resource cost is few, counting yield is high, more accurate advantage of reasonable.

Description

A kind of punching part sample researching method

Technical field

The invention belongs to the machine work field, relate to information retrieval and process for stamping, especially retrieve similar stamping parts, so that the method for better copying the Sheet Metal Forming Technology of the stamping parts that retrieves to carry out the technological design of new parts.

Background technology

The characteristics of stamping process design are weak theory, strong experience, and (Case based Reasoning CBR) comes the design technology scheme to be fit to adopt reasoning based on example in general.CBR utilizes existing similar example to solve problem, and therefore how retrieving similar example just becomes one of gordian technique of using CBR.At this problem of case retrieval, several different methods has been proposed both at home and abroad, mainly contain: nearest neighbor method, method of induction and based on method of knowledge etc.Yet the document that is applied to the stamping parts case retrieval of success is also few.Similar stamping parts has similar Sheet Metal Forming Technology.How from the angle of resemble process, rather than, retrieve similar stamping parts, so that better copy the Sheet Metal Forming Technology of the stamping parts that retrieves to carry out the technological design of new parts from the similar angle of shape.At present, also do not retrieve the patent of invention of these class methods.

Find that through literature search relevant punching part sample researching method mainly contains following several to prior art:

(1) Ke Xugui etc. is at " the similar case retrieval of blanking die CBR system " (forging technology 2005,4) in blanking part is divided into three types of blank, punching part and compound blanking spares, adopts the mode of coding to carry out geometric configuration, the grade of tolerance, material and matching ratio in batches;

(2) M.El-Mehalawi etc. is at " A database system of mechanical componentsbased on geometrics and topological similarity.Part II:indexing, retrieval, matching, and similarity assessment " in (COMPUTER-AIDED DESIGNVol.35 Issue 1), seek the face of coupling and the similarity that the mechanical component 3D solid is investigated on the limit;

(3) Chinese patent application number: 200410067152, patent name is: three-dimensional CAD model is based on the similarity estimating method of shape, and this patent readme is: " the STEP AP203 Part21 data of three-dimensional CAD model are converted to the attributed graph data; The dependency diagram data extracts the feature invariant of three-dimensional CAD model and constructs its feature invariant vector with this; Use self organizing neural network based on the feature invariant vector of structure cad model is carried out similarity assessment.”

Though method 1 can realize the similar case retrieval of blanking part, each attribute represents that with 1 and 0 coupling does not still match, and is too coarse simple.Though method 2 and 3 can be applicable to the stamping parts case retrieval, only consider that shape is similar, and do not consider the information that other are relevant with technology that possibly can't use for reference the process program of the stamping parts that retrieves, because shape is the most similar, technology is not necessarily the most similar.For example, the difference of material may cause the stamping parts of identical three-dimensional entity model, and the technology difference is very big.

Summary of the invention

The objective of the invention is at the deficiencies in the prior art, a kind of punching part sample researching method based on resemble process is proposed, make it solve the deficiency that exists in the background technology, realize the intelligent retrieval of similar stamping parts, thereby make full use of the successful experience of Sheet Metal Forming Technology, carry out the stamping process design of new parts.The present invention at first extracts from the stamping parts three-dimensional model between feature and feature and retrains, and according to stamping parts material information table, the material that obtains material is thick, intensity and hardness.From the angle of resemble process, divide two stages to calculate the similarity of new and old stamping parts example then.The number of phase one with feature and constraint on each plane of new and old example is that benchmark calculates stamping parts topological relation similarity, filters out some dissimilar old examples, obtains a collection of similar old example, dwindled the search space of follow-up phase; Subordinate phase by the comprehensive similarity of feature, constraint and material between new and old example, obtains an old example the most similar from a collection of old example of phase one acquisition.What the method for whole retrieval adopted is nearest neighbor method.

For achieving the above object, solution of the present invention is: a kind of punching part sample researching method based on resemble process, for making full use of the successful experience of Sheet Metal Forming Technology, from the angle of resemble process, adopt nearest neighbor method to carry out the stamping parts case retrieval.

Further, at first from the stamping parts three-dimensional model, extract between feature and feature and retrain, and according to stamping parts material information table, the material that obtains material is thick, intensity and hardness, from the angle of resemble process, divide two stages to calculate the similarity of new and old stamping parts example then.

Phase one is that benchmark calculates stamping parts topological relation similarity with the plane, filters out some dissimilar old examples, obtains a collection of similar old example, has dwindled the search space of follow-up phase; Subordinate phase by the comprehensive similarity of feature, constraint and material between new and old example, draws an old example the most similar from a collection of old example of phase one acquisition.

The described phase one is that benchmark calculates stamping parts topological relation similarity with the plane, according to the number of feature and constraint on each plane of new and old stamping parts, adopts the Euclidean distance in the nearest neighbor method to weigh each similarity to the plane, and search plane coupling is right.Afterwards, according to the right similarity value of plane coupling, draw new and old stamping parts topological relation similarity, and arrange from high to low, setting threshold obtains a collection of comparatively similar stamping parts.

Described subordinate phase is calculated new and old stamping parts example similarity, carries out in a collection of comparatively similar stamping parts of phase one acquisition.At first the plane of finding with the phase one is paired into the basis, and right, right according to the characteristic matching similarity value of feature that search is complementary on two planes that are complementary is calculated whole new and old stamping parts characteristic similarity; The second, retrain similarity between calculated characteristics, as long as the kind of two constraints is identical in the new and old stamping parts, just think that they are that coupling is right, and similarity is 1, calculate whole new and old stamping parts constraint similarity; The 3rd,, intensity thick according to the material of material and hardness adopt Euclidean distance to calculate the material similarity; Last weighted calculation feature, constraint and material comprehensive similarity

Specifically, the present invention is achieved by the following technical solutions: preliminary work: extract each feature by the stamping parts three-dimensional model diagram, and deposit with table 1.Of particular note, the plane is two planes that link to each other with it under the bending features.Feature is divided into main feature and auxilliary feature, and main feature comprises plane Flat, crooked Bend and pull and stretch Forming or the like, and auxilliary feature comprises circular hole Hole, (length) square hole Slot, profile hole Profile, hole flanging Extrusion hole, crimping Bead, projection Emboss, rim roll Flange, or the like.Except feature directly exerted an influence to Sheet Metal Forming Technology, the constraint between the feature also had very big influence to Sheet Metal Forming Technology, can be divided into accuracy constraint Precision-associated restriction, as two distance between borehole accuracy requirement was arranged; Near constraint Close-with restriction, near curved edges, pitch of holes is little as the hole.Extract to retrain between feature and deposit with table 2.

Table 1 mark sheet

Feature	Affiliated plane
		Flat1
Hole1	Flat1
		...	...

Table 2 restriction table

Constraint	Feature 1	Feature 2	Affiliated plane
				Precision-associated	Hole1	Hole2	Flat1
...	...	...

Need to formulate the new parts of Sheet Metal Forming Technology and each old part in the case library, all have separately mark sheet and restriction table.In addition, according to stamping parts material information table, the material that can obtain material is thick, intensity and hardness.After preliminary work is finished, just can carry out the stamping parts case retrieval.The retrieval of new old part similarity adopts nearest neighbor method to carry out, and specifically is divided into for two megastages:

Phase one: calculate new and old stamping parts topological relation similarity.Stamping parts of the present invention all refers to metal plate punching spare, its essential characteristic is exactly one or several planes, on these planes and the plane attached characteristics determined the topological relation in space, and topological relation is directly relevant with technology, so the phase one is investigated the number of feature and constraint on each plane of new and old example, promptly be benchmark computer memory topological relation similarity, filter old example with this with the plane.Phase one realizes by following 6 steps:

Step 1: establishing new example has a plane (Flat ₁ ^New, Flat ₂ ^New..., Flat _e ^New..., Flat _a ^New), old example has b plane (Flat ₁ ^Old, Flat ₂ ^Old..., Flat _f ^Old..., Flat _b ^Old).Form possible plane to (Flat ₁ ^New, Flat ₁ ^Old), (Flat ₁ ^New, Flat ₂ ^Old) ..., (Flat ₁ ^New, Flat _b ^Old), (Flat ₂ ^New, Flat ₁ ^Old), (Flat ₂ ^New, Flat ₂ ^Old) ..., (Flat ₂ ^New, Flat _b ^Old) ..., (Flat _a ^New, Flat ₁ ^Old), (Flat _a ^New, Flat ₂ ^Old) ..., (Flat _a ^New, Flat _b ^Old).A * b plane is right altogether.

Step 2:, adopt Euclidean distance to calculate each similarity) shown in as formula 1 to the plane according to the number of feature and constraint on each plane:

$S_{\mathrm{Old},f}^{\mathrm{New},e}=1-\sqrt{\frac{\underset{k=1}{\overset{p+r-m}{\mathrm{\Σ}}}{w}_k^{\mathrm{Fea}}\×{\left(\frac{{\mathrm{Flat}}_{e,k}^{\mathrm{New},\mathrm{Fea}}-{\mathrm{Flat}}_{f,k}^{\mathrm{Old},\mathrm{Fea}}}{{\mathrm{Flat}}_{e,k}^{\mathrm{New},\mathrm{Fea}}+{\mathrm{Flat}}_{f,k}^{\mathrm{Old},\mathrm{Fea}}}\right)}^2+\underset{t=1}{\overset{q+s-n}{\mathrm{\Σ}}}{w}_t^{\mathrm{Re}}\×{\left(\frac{{\mathrm{Flat}}_{e,t}^{\mathrm{New},\mathrm{Re}}-{\mathrm{Flat}}_{f,t}^{\mathrm{Old},\mathrm{Re}}}{{\mathrm{Flat}}_{e,t}^{\mathrm{New},\mathrm{Re}}+{\mathrm{Flat}}_{f,t}^{\mathrm{Old},\mathrm{Re}}}\right)}^2}{\underset{k=1}{\overset{p+r-m}{\mathrm{\Σ}}}{w}_k^{\mathrm{Fea}}+\underset{t=1}{\overset{q+s-n}{\mathrm{\Σ}}}{w}_t^{\mathrm{Re}}}}---1)$

In the formula, p is plane Flat _e ^NewFeature kind sum; R is plane Flat _f ^OldFeature kind sum; M is plane Flat _e ^NewWith plane Flat _f ^OldIdentical feature species number; w _k ^FeaBe k kind feature weight (k=1,2 ..., p+r-m); Flat _{E, k} ^{New, Fea}Be plane Flat _e ^NewThe number of k kind feature; Flat _{F, k} ^{Old, Fea}Be plane Flat _f ^OldThe number of k kind feature; Q is plane Flat _eW constraint kind sum; S is plane Flat _f ^OldConstraint kind sum; N is plane Flat _e ^NewWith plane Flat _f ^OldIdentical constraint species number; w _t ^ReBe t kind constraint weight (t=1,2 ..., q+s-n); Flat _{E, t} ^{New, Re}Be plane Flat _e ^NewThe number of t kind constraint; Flat _{F, t} ^{Old, Re}Be plane Flat _f ^OldThe number of t kind constraint.

Step 3: with a * b plane to by its similarity value S _{Old, f} ^{New, e}Arrange from high to low, form data list 3.

Table 3 coupling is right

Feature 1	Feature 2	Similarity
			Flat e New	Flat f Old	S Old，f New，e
...	...	...

Step 4: get first row in the table 3, promptly find the pair of planar coupling to (Flat _e ^New, Flat _f ^Old), be placed in the new table 4 (structure is with table 3); All comprise Flat in the delete list 3 then _e ^NewAnd Flat _f ^OldRow.

Step 5: repeating step 4 is empty up to table 3.(a, b) the individual plane that is complementary is right, and is recorded in the table 4 so just to have found min.

Step 6: with all the similarity value S in the table 4 _{Old, f} ^{New, e}Addition multiply by 2 again divided by a+b, obtains topological relation similarity value.

That is: $S_{\mathrm{Topo}\log y}=\frac{\left(\underset{1}{\overset{\min (a,b)}{\mathrm{\Σ}}}{S}_{\mathrm{Old},f}^{\mathrm{New},e}\right)\×2}{a+b}---2)$

With the S that obtains _TopologyAll values is arranged from high to low, and setting threshold η, prepares the subordinate phase similarity relatively.

Subordinate phase: calculate new and old stamping parts comprehensive similarity.The similarity of constraint and material between integrated survey feature, feature, realize by following 4 steps:

Step 1: calculated characteristics similarity.The plane of finding with the phase one is paired into the basis, and the feature that search is complementary on two planes that are complementary is right.Flat for example _e ^NewAnd Flat _f ^OldBe that the plane that is complementary is right, Flat _e ^NewOn 4 circular hole Hole are arranged, 1 (length) square hole Slot, 1 hole flanging Extrusion hole and 1 crooked Bend; Flat _f ^OldOn 2 circular hole Hole are arranged, 1 (length) square hole Slot, 1 hole flanging Extrusionhole and 1 crooked Bend; So just thinking has a pair of (length) square hole Slot, a pair of hole flanging Extrusionhole, and pair of curved Bend and 2 couples of circular hole Hole are complementary.A pair of (length) square hole Slot, a pair of hole flanging Extrusion hole and pair of curved Bend directly can find, because have only an individual features on the plane separately.How from Flat _e ^NewOn 4 circular hole Hole and Flat _f ^OldOn 2 circular hole Hole find 2 couples of circular hole Hole to be complementary, the method for employing is with all combinations by formula 3) calculate and look for the highest 2 couple of similarity.At last will each mate the plane on all matching characteristics of searching to all putting into table 4.Each coupling in the his-and-hers watches 4 is to (comprise plane to) by formula 3) calculate its characteristic similarity.

$S_i^{\mathrm{Feature}}=1-\sqrt{\frac{\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{w}_j^{\mathrm{parameter}}\×{\left(\frac{{\mathrm{Parameter}}_{i,j}^{\mathrm{New}}-{\mathrm{Parameter}}_{i,j}^{\mathrm{Old}}}{{\mathrm{Parameter}}_{i,j}^{\mathrm{New}}+{\mathrm{Parameter}}_{i,j}^{\mathrm{Old}}}\right)}^2}{\underset{j=1}{\overset{l}{\mathrm{\Σ}}}{w}_j^{\mathrm{parameter}}}}---3)$

In the formula, l is that characteristic matching is to i characteristic parameter sum.For the plane, characteristic parameter refers to its area; Refer to its sweep length and angle of bend for bending; Refer to its diameter for circular hole; Refer to its area or the like for profile hole, in addition, characteristic parameter also comprises form tolerance, surfaceness.w _j ^ParameterBe j characteristic parameter weight (j=1,2 ..., l); Parameter _{I, j} ^NewFor characteristic matching to j the characteristic ginseng value of i in new parts; Parameter _{I, j} ^OldFor characteristic matching to j the characteristic ginseng value of i in old part.

Calculate whole new and old part feature similarity, as formula 4) shown in:

$S_{\mathrm{Feature}}=\frac{(\underset{i=1}{\overset{h}{\mathrm{\Σ}}}{w}_i^{\mathrm{Fea}}\×S_i^{\mathrm{Feature}})\×2}{\underset{i=1}{\overset{c}{\mathrm{\Σ}}}{w}_i^{\mathrm{Fea}}+\underset{i=1}{\overset{d}{\mathrm{\Σ}}}{w}_i^{\mathrm{Fea}}}---4)$

In the formula, h be the right sum of characteristic matching (i=1,2 ..., h); C be the new parts feature sum (i=1,2 ..., h, h+1 ..., c); D be the old part feature sum (i=1,2 ..., h, h+1 ..., d); w _i ^FeaBe the weight of feature, value is with formula 1) middle w _k ^FeaS _i ^FeatureBe formula 3) shown in.

Step 2: retrain similarity between calculated characteristics.As long as the kind that retrains in the new old part is identical, just think that they mate, and similarity is 1.For example new parts has two accuracy constraints, and old part has an accuracy constraint, and one near constraint, so just think that to find an Accuracy Matching right, and similarity is 1.

Calculate whole new old part constraint similarity, as formula 5) shown in:

$S_{\mathrm{Restriction}}=\frac{\left(\underset{i^{\′}=1}{\overset{h^{\′}}{\mathrm{\Σ}}}{w}_{i^{\′}}^{\mathrm{Re}}\right)\×2}{\underset{i^{\′}=1}{\overset{c^{\′}}{\mathrm{\Σ}}}{w}_{i^{\′}}^{\mathrm{Re}}+\underset{i^{\′}=1}{\overset{d^{\′}}{\mathrm{\Σ}}}{w}_{i^{\′}}^{\mathrm{Re}}}---5)$

In the formula, h ' for the right sum of constraint coupling (i '=1,2 ..., h '); The sum that c ' retrains for new parts (i '=1,2 ..., h ', h '+1 ..., c '); The sum that d ' retrains for old part (i '=1,2 ..., h ', h '+1 ..., d '); w _{I '} ^ReBe the weight of constraint, value is with formula 1) middle w _t ^Re

Step 3: calculate the material similarity.Investigate the thickness of new and old part material, the similarity between intensity and hardness, by formula 6) calculate.

$S_{\mathrm{Material}}=1-\sqrt{w_{\mathrm{th}}\×{\left(\frac{M_{\mathrm{th}}^{\mathrm{New}}-M_{\mathrm{th}}^{\mathrm{Old}}}{M_{\mathrm{th}}^{\mathrm{New}}+M_{\mathrm{th}}^{\mathrm{Old}}}\right)}^2+w_{\mathrm{in}}\×{\left(\frac{M_{\mathrm{st}}^{\mathrm{New}}-M_{\mathrm{st}}^{\mathrm{Old}}}{M_{\mathrm{st}}^{\mathrm{New}}+M_{\mathrm{st}}^{\mathrm{Old}}}\right)}^2+w_{\mathrm{ha}}\×{\left(\frac{M_{\mathrm{ha}}^{\mathrm{New}}-M_{\mathrm{ha}}^{\mathrm{Old}}}{M_{\mathrm{ha}}^{\mathrm{New}}+M_{\mathrm{ha}}^{\mathrm{Old}}}\right)}^2}---6)$

In the formula, W _ThWeight for material thickness; w _StWeight for the strength of materials; w _HaBe the weight of material hardness, w _Th+ w _In+ w _Ha=1; M _Th ^NewBe the new material one-tenth-value thickness 1/10; M _Th ^OldBe old material thickness value; M _St ^NewBe the new material intensity level; M _St ^OldBe old strength of materials value; M _Ha ^NewBe the new material hardness number; M _Ha ^OldBe old material hardness value.

More than 3 steps do not have sequencing, as long as all realize after, then can carry out following step 4.

Step 4: calculate comprehensive similarity.By formula 7) calculate.

S＝w _Feature×S _Feature+w _Restriction×S _Restriction+w _Material×S _Material 7)

In the formula, w _FeatureBe characteristic similarity S _FeatureWeight; w _RestrictionBe constraint similarity S _RestrictionWeight; w _MaterialBe material similarity S _MaterialWeight, w _Feature+ w _Restriction+ w _Material=1; S _FeotureBe characteristic similarity, as formula 4) shown in; S _RestrictionFor the constraint similarity, as formula 5) shown in; S _MoterialBe the material similarity, as formula 6) shown in.

Resulting S value is arranged from high to low, obtain a most similar old example of stamping parts.So far finish the retrieval of the similar example of stamping parts.

The present invention has substantive distinguishing features, has solved the problem that exists in the background technology, provides a kind of angle from resemble process to investigate the method for the similar case retrieval of stamping parts.1. the principle of this method is a nearest neighbor method, and algorithm is easy, be easy to realize, resource cost is few, counting yield is high; 2. the constraint and the similarity of material between the new and old stamping parts spatial topotaxy of integrated survey, feature, feature, than the similarity of only considering shape, more help with reference to existing process program, and the similarity value adopts Euclidean distance to calculate, than only judging that with 1 and 0 coupling does not match, more accurate reasonable; 3. this method adopts two stages to retrieve, and the calculating by the phase one just can filter out many dissimilar stamping parts, thereby has dwindled the search space of subordinate phase, and the raising retrieval rate is more than 50%.

Description of drawings

Fig. 1 is the phase one process flow diagram of a kind of embodiment of punching part sample researching method of the present invention;

Fig. 2 is the subordinate phase process flow diagram of a kind of embodiment of punching part sample researching method of the present invention;

Fig. 3 is an embodiment of the present invention part drawing.

Embodiment

Below in conjunction with embodiment content of the present invention is illustrated.

As shown in Figure 3, the stamping parts of a pending stamping process design is arranged, i.e. new parts among Fig. 3.Up to a hundred the old parts of successfully having formulated Sheet Metal Forming Technology are arranged in the case library, existing 11 stamping parts that only list wherein, promptly 001 old part among Fig. 3 is to 011 old part.Adopt the method among the present invention to retrieve, new parts and which old part are the highest from the angle similarity of technology.Detailed process is summarized as follows:

At first extract between the feature of new parts and feature and retrain.New parts has 4 faces, 3 bendings, and 2 hole flangings, 8 circular holes, 2 oblong aperture and 1 projection are deposited these features and affiliated plane with table.Constraint comprises 2 accuracy constraints between feature, deposits with table.Carry out the topological similarity of phase one then and calculate, with as follows behind the resulting sort result.

No. 008 old part S _Topology=0.82794 No. 005 old part S _Topology=0.78451

No. 004 old part S _Topology=0.76656 No. 003 old part S _Topology=0.66597

No. 010 old part S _Topology=0.59410 No. 007 old part S _Topology=0.58264

No. 002 old part S _Topology=0.57067 No. 006 old part S _Topology=0.43125

No. 009 old part S _Topology=0.32190 No. 001 old part S _Topology=0.30219

No. 011 old part S _Topology=0.09565

In above-mentioned computation process, the weight of all main features all gets 0.7, and the weight of all auxilliary features all gets 0.3, and the weight of all constraints all gets 0.6.Weight is according to feature the influence degree of technology to be determined, and is to allow to revise.

If threshold value η=0.7 then has only No. 008, No. 005 and No. 004 old part enters the calculating of subordinate phase similarity, dwindle range of search 72%.The result of subordinate phase retrieval is as follows by arranging from high to low.

No. 008 old part S=0.80375 of No. 005 old part S=0.81655

No. 004 old part S=0.74163

Therefore definite No. 005 old part and new parts, from the angle of technology, similarity is the highest.And then can formulate the Sheet Metal Forming Technology of new parts with reference to the Sheet Metal Forming Technology scheme of No. 005 old part.

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 should be within protection scope of the present invention for improvement and modification that the present invention makes according to announcement of the present invention.

Claims (1)

1, a kind of punching part sample researching method, it is characterized in that: for making full use of the successful experience of Sheet Metal Forming Technology, angle from resemble process, adopt nearest neighbor method to carry out the stamping parts case retrieval: at first from the stamping parts three-dimensional model, to retrain between extraction feature and feature, and according to stamping parts material information table, the material that obtains material is thick, intensity and hardness; From the angle of resemble process, calculate the similarity of new and old stamping parts example then;

The similarity of the new and old stamping parts example of described calculating is to carry out in two stages of branch: the phase one is that benchmark calculates stamping parts topological relation similarity with the plane, filters out dissimilar old example, obtains similar old example, to dwindle the search space of follow-up phase; Subordinate phase by the comprehensive similarity of feature, constraint and material between new and old example, draws an old example the most similar from a collection of old example of phase one acquisition;

The described phase one is that benchmark calculates stamping parts topological relation similarity with the plane, is the number according to feature and constraint on each plane of new and old stamping parts, adopts the Euclidean distance in the nearest neighbor method to weigh each similarity to the plane, and search plane coupling is right; Afterwards, according to the right similarity value of plane coupling, draw new and old stamping parts topological relation similarity, and arrange from high to low, setting threshold obtains similar stamping parts;

Described subordinate phase is calculated new and old stamping parts example similarity, is to carry out in a collection of comparatively similar stamping parts of phase one acquisition; At first the plane of finding with the phase one is paired into the basis, and right, right according to the characteristic matching similarity value of feature that search is complementary on two planes that are complementary is calculated whole new and old stamping parts characteristic similarity; The second, retrain similarity between calculated characteristics, as long as the kind of two constraints is identical in the new and old stamping parts, just think that they are that coupling is right, and similarity is 1, calculate whole new and old stamping parts constraint similarity; The 3rd,, intensity thick according to the material of material and hardness adopt Euclidean distance to calculate the material similarity; Last weighted calculation feature, constraint and material comprehensive similarity.

Images