CN109918713A - A kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design - Google Patents
A kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design Download PDFInfo
- Publication number
- CN109918713A CN109918713A CN201910065147.1A CN201910065147A CN109918713A CN 109918713 A CN109918713 A CN 109918713A CN 201910065147 A CN201910065147 A CN 201910065147A CN 109918713 A CN109918713 A CN 109918713A
- Authority
- CN
- China
- Prior art keywords
- product
- target product
- class
- gene
- functional unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design: one, using the function and behavioural information of first-order predicate logic method analysis target product, functional information being decomposed into Functional Unit using principle of D&R;Two, it is found out from product example library and forms sample set with target product product example of the same clan;Three, sample is gathered for K class, it is class belonging to target product apart from class where the smallest central point that the value of K, which calculates the Euclidean distance between target product and each class central point according to the type set of target product,;Four, attribute information relevant to the Functional Unit is extracted from the sample of the affiliated class of target product;Five, it is selected and the biggish m attribute of Functional Unit correlation using K folding cross validation method;Six, the m attribute and behavioural information selected are stored in the form of coding strand into product gene library, obtains the genetic knowledge of target product, the present invention can automatically obtain all critical learning influenced in conceptual design process.
Description
Technical field
The invention belongs to the technical fields that product genetic knowledge obtains, and in particular to a kind of gene of Product Conceptual Design
Automated Acquisition of Knowledge method.
Background technique
Conceptual design is the critical stage for determining product quality and products innovation, constraint of the stage to designer
It at least, is the stage that designer's experience, wisdom and creativeness are best embodied in the entire design process of product;The stage can determine
Fixed 80% or more product cost, and the phase, to need to pay decades of times even hundreds of after design for the incorrect decision that generates of the stage
Cost again makes up.
But there is conceptual phase finiteness, uncertainty, multidisciplinary property and a large amount of design of information to replace
For scheme.These features make the retrieval of design information and shared difficult, the design scheme that is obtained by the information of fuzzy finite
The problems such as feasibility is lower and the design cycle is long.Therefore, there is an urgent need to develop the pass in a kind of determining conceptual design process
Key information and in a kind of method that unified mode indicates these all critical learnings.
For this purpose, researcher introduces the concept of " product gene ", be defined as determining product key characteristic, it is right
Standardized letter that conceptual design process has great influence effect, being transmitted between parent product and filial generation product
Breath set.Design method (Genetics-Based Design, GBD) based on genetic idea is also proved to be a kind of effective
The method of development innovation scheme, this method have the blindness for reducing innovative design, increase design scheme space, improve
The advantages of concept feasible.Therefore, GBD in design innovative using more and more extensive.
The definition and acquisition of product gene are to study the premise and basis of GBD.Currently, in terms of product gene obtains
Research rest on the structuring of acquisition process mostly in terms of, lack effective implementation path or method to obtain product
Gene.
Summary of the invention
In view of this, the present invention provides a kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design, it can be automatic
Obtain all critical learning influenced in conceptual design process.
Realize that technical scheme is as follows:
A kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design, comprising the following steps:
Step 1: using the function and row of first-order predicate logic method analysis target product (Target Product, TP)
For information, then using principle of D&R (Principle of Decomposition and Reconstruction,
PDR functional information) is decomposed into Functional Unit;
Step 2: finding out the product example of the same clan with target product from product example library, sample set is formed;
Step 3: gathering sample for K class, the value of K calculates TP and each class according to the type set of target product
Euclidean distance between central point is class belonging to TP apart from the class where the smallest central point;
Step 4: extracting attribute information relevant to the Functional Unit from the sample of the affiliated class of TP;
Step 5: using K folding cross validation method select with the biggish m attribute of Functional Unit correlation, m is setting value;
Step 6: the m attribute that step 5 is selected and the behavioural information that step 1 obtains are stored in the form of coding strand
Into product gene library, that is, obtain the genetic knowledge of target product.
Further, it in step 3, is clustered using K-means clustering algorithm.
Further, in step 4, using based on maximum correlation and minimum redundancy (Max-Relevance and
Min-Redundancy, mRMR) standard best first increment select (Optimal First-Order Incremental
Selection, OFOIS) method carry out attribute information extraction.
The utility model has the advantages that
(1) the method for the present invention realizes the automatic acquisition of product gene, is the acquisition of all critical learning in conceptual design process
A kind of feasible, automation method is provided.
(2) the method for the present invention makes all critical learning in conceptual design process clear, and is provided for a kind of unification
Representation method provides the foundation for GBD, solves finiteness, ambiguity and uncertainty etc. because of conceptual design process information
Caused by innovative design blindly, the low problem of the feasibility of design scheme, also there is important meaning to product design efficiency is improved
Justice.
(3) the method for the present invention realizes the acquisition of key Design knowledge, sets to saving design knowledge retrieval time, improving
Meter efficiency is of great significance.In addition, the feasibility of the design scheme obtained based on these all critical learnings made clear is had
Effect improves, and the probability that these design knowledges are used for reference and reused in subsequent design increases, this is to the weight for improving design knowledge
It is of great significance with rate.
(4) it selects K-means clustering algorithm to carry out sample clustering in the method for the present invention, can quickly obtain cluster result,
And the algorithm is applicable not only to conventional data object, is also applied for big data, so that entire algorithm is with strong applicability.
(5) the present invention is based on the OFOIS algorithm of mRMR, classifying quality and the classification methods of maximum dependence standard
Effect is of equal value, and the classification error rate of the algorithm is lower, and can rapidly and accurately realize tagsort.
Detailed description of the invention
Fig. 1 is conceptual design process model of the present invention.
Fig. 2 is product of the present invention gene extraction process figure.
Fig. 3 is product of the present invention gene acquisition methods figure.
Fig. 4 is pesticide spraying of embodiment of the present invention aircraft casing organization plan figure.
Fig. 5 is the property parameters table of part of embodiment of the present invention sample.
Fig. 6 is two-dimensional visualization of embodiment of the present invention cluster result figure.
Fig. 7 is the cluster result figure of part of embodiment of the present invention sample.
Fig. 8 is the number of iterations of the embodiment of the present invention and cluster centre figure.
Fig. 9 is the corresponding part index number figure of Functional Unit of the embodiment of the present invention.
Specific embodiment
The present invention will now be described in detail with reference to the accompanying drawings and examples.
The present invention provides a kind of gene Automated Acquisition of Knowledge methods of Product Conceptual Design, pass through conceptual design first
Process model building analyzes the key element in conceptual design process.Then, analogy biological character expresses process, from conceptual design
Product gene is extracted in journey model, the coding method of product gene is studied later, propose a kind of based on Functional Unit
Product gene coding method proposes that the product gene of a kind of knowledge based and machine learning algorithm obtains automatically on this basis
Method.
Conceptual design process is one and takes out objective function from design requirement, and uses certain technological means will
Objective function is expressed as the process of organization plan.Therefore, conceptual design process can be represented simply as model shown in Fig. 1.
Function in conceptual design process model is analogized into biological character, organization plan analogizes to protein.Due to life
Physical property shape is to be eventually converted into protein by biological gene expression, therefore, in conceptual design, function to organization plan
Conversion process also can be regarded as realizing by the expression process of product gene.The present invention is in traditional function-behavior-
Conceptual design process is expressed as function-behavior-attribute-organization plan (FBAS) model on the basis of structure (FBS), wherein
Attribute refers to the physical attribute for influencing product intrinsic propesties and plays the geometric attribute of restriction effect to product integral layout.Therefore,
The product gene extracted from FBAS can be defined as using behavior and determinant attribute as base, using relationship between the two as hydrogen
The duplex structure of key, as shown in Fig. 2.Wherein, hydrogen bond only plays connection function.
In order to seek a kind of unified representation method of product gene, analogy biological gene coding mode of the present invention proposes one
Product gene coding method of the kind based on Functional Unit, as shown in table 1.
Table 1 is encoded based on the product gene of Functional Unit
Wherein, FiIt is i-th of Functional Unit, Start is promoter, and End is to terminate son, and N is gene address, ByIt is and FiIt is right
The behavior answered, AmReferring to influences FiThe determinant attribute information of expression mainly includes physical attribute and geometric attribute.
Below with reference to the embodiment of solid pesticide sprinkling aircraft casing (Aircraft Shell, AS) design to this hair
The acquisition methods of the product gene of bright proposition are illustrated.
AS is to form aircraft shape, contain solid pesticide and connect the device of aircraft other parts.The device can
It bears in flight course from inside and outside load, and the resistance of its shape and surface quality to being encountered in flight course
The size of power and frictional force has great influence.In addition, inner space and splendid attire solid pesticide of the shape of AS to aircraft
Measurer has decisive action.The organization plan of AS is as shown in Figure 4.
The major parameter of AS is as shown in table 2.
2 shell attribute information table of table
As shown in figure 3, the acquisition process of product gene of the embodiment of the present invention is as follows:
Step 1: extracting its function according to the action principle of AS using first-order predicate logic SV (P) O method and behavior being dynamic
Word, S indicate that subject, V indicate that actional verb, O indicate that effective object, P indicate the complement of V or O, and using decomposed and reconstituted
Function Decomposition is Functional Unit by principle (PDR) technology.
The major function and actional verb of 3 AS of table
As shown in Table 3, the major function of aircraft is " constituting aircraft shape ", and main actional verb includes " holding
By ", " influence ", " decision ", " connection ".
Function Decomposition is carried out using major function of the PDR technology to AS, the Functional Unit for obtaining AS is as shown in table 4.
The Functional Unit table of 4 AS of table
Step 2: selecting 99 from product example library forms sample set with example AS of the same clan.The attribute of these samples
Type is identical as AS, but corresponding parameter value and AS are not exactly the same.If sample set is D={ x0,x1,…,x98, each
Sample can be expressed as xi=(xi0,xi1,…,xij…,xim), wherein xijFor the attribute information for including in sample, m=15.
The partial parameters value of sample set is as shown in Figure 5.
Step 3: clustering using K-means clustering algorithm to the sample set in step (2), and judge target product
Class where TP.
In cluster, need to determine the number of cluster first, i.e., the value of k in algorithm.Generally, the shape of aircraft casing
Shape can be divided into three kinds: 1) cone-column-skirt shape (cone angle 1 ≠ 0, cone angle 2=0, cone angle 3 ≠ 0);2) double columnar form (cone angle 1 ≠ 0, cone
Angle 2=cone angle 3 ≠ 0);3) triconic (cone angle 1 ≠ 0, cone angle 2 ≠ 0, cone angle 3 ≠ 0).Therefore, in the present embodiment, variable k
Value be set as 3.Clustering algorithm is realized using Python3.6, the result after cluster is as shown in Fig. 6, Fig. 7 and Fig. 8.It can by figure
Know, algorithm reaches termination condition when iterating to the 16th time, stops iteration, export sample that every one kind is included and
The central point of class.Contained sample number is respectively 27,48 and 24 in class 1-3 after iteration.
After obtaining cluster sample and cluster centre result, need to judge target product AS institute using Euclidean distance formula
Class.
Wherein, ρ (X, U) indicates the Euclidean distance between X and U, xiIndicate the ith attribute column of target product TP,Table
Show in each class central point with xiCorresponding column vector.
By calculating, learn that AS belongs to the second class: double columnar form.
Step 4: using 48 samples in TP and class 2 as research object, using the best first increment based on mRMR
Attribute relevant to each Functional Unit is successively selected in selection (OFOIS) from the property parameters of sample set.
MRMR (formula 4) refers to that maximum correlation and minimum redundancy standard, basic thought are first according to maximum phase
Close property standard (formula 2) select with the biggish feature of class correlation, then using minimum redundancy standard (formula 3) to maximum
The feature selected in correlation criterion is calculated, and the feature of redundancy is removed.
Wherein, D (S, fλ) indicate property set S and Functional Unit fλBetween correlation, xiFor the element in property set, S is
Element number in set S, I (xi;fλ) indicate xiWith fλBetween mutual information.
Wherein, R (S) indicates the redundancy in property set S between element, xi,xjFor the element in property set S, | S | for collection
Close the element number in S, I (xi;xj) indicate xiWith xjBetween mutual information.
Max φ (D-R)=D-R (4)
Wherein, φ (D-R) indicates the difference of maximum correlation D and minimum redundancy R, is the function representation of mRMR standard
Formula.
OFOIS refers to based on mRMR standard, only a feature is selected to be added in feature set from all features every time.
OFOIS is a kind of algorithm with maximum dependence (Max-Dependency) equivalence, which can be rapidly and accurately to feature
Classify, and the error rate classified is very small.Assuming that total feature set is X, and we have m-1 and a certain Functional Unit
fλRelevant feature, be from { X-Sm-1M features relevant to the Functional Unit are selected in feature set, then it should meet
Wherein, the collection for the m-1 feature composition selected is combined into Sm-1, xiFor set Sm-1In a certain element, xjFor collection
Close { X-Sm-1In a certain element, I (x;Y) mutual information of x and y is indicated.
In this example, need to select determinant attribute relevant to each Functional Unit, the corresponding index value of Functional Unit is as schemed
Shown in 9.In the present invention, according to feature and function member between relationship, feature is divided into two classes: it is related to Functional Unit and with
Functional Unit is uncorrelated.Therefore, it can use OFOIS and carry out feature selecting.The algorithm is realized using Python3.6 programming language,
Obtain that the results are shown in Table 5, attribute therein only lists corresponding subscript.
5 OFOIS arithmetic result table of table
Step 5: being selected from the sequence of attribute obtained in step (4) using eight folding cross-validation methods related to Functional Unit
The biggish attribute of property;
As shown in table 5, a sequence with each Functional Unit correlation size has been obtained using OFOIS algorithm, still,
For each Functional Unit, all features are all listed, and do not play the purpose for selecting the feature set of correlation maximum.For
This, the present embodiment using 8 folding cross-validation methods to attribute obtained in table 5 sequence further selected, thus obtain and
The relevant some determinant attributes of Functional Unit, as shown in table 6.
6 determinant attribute information table of table
Step 6: by the determinant attribute information that step (5) obtains and the verb information that step (1) obtains with product gene
The form of coding strand is stored into product gene library.
These attributes form product gene together with the actional verb that step (1) obtains, and with product gene coding strand
Form is stored in product gene library, as shown in table 7.
The product gene library table of 7 AS of table
In conclusion the above is merely preferred embodiments of the present invention, being not intended to limit protection model of the invention
It encloses.All within the spirits and principles of the present invention, any modification, equivalent replacement, improvement and so on should be included in this hair
Within bright protection scope.
Claims (3)
1. a kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design, which comprises the following steps:
Step 1: using the function and behavioural information of first-order predicate logic method analysis target product, then using decomposed and reconstituted
Functional information is decomposed into Functional Unit by principle;
Step 2: finding out the product example of the same clan with target product from product example library, sample set is formed;
Step 3: gathering sample for K class, the value of K calculates target product and each class according to the type set of target product
Euclidean distance between central point is class belonging to target product apart from class where the smallest central point;
Step 4: extracting attribute information relevant to the Functional Unit from the sample of the affiliated class of target product;
Step 5: using K folding cross validation method select with the biggish m attribute of Functional Unit correlation, m is setting value;
Step 6: being stored the m attribute that step 5 is selected and the behavioural information that step 1 obtains in the form of coding strand to production
In product gene pool, that is, obtain the genetic knowledge of target product.
2. a kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design as described in claim 1, which is characterized in that step
In three, clustered using K-means clustering algorithm.
3. a kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design as described in claim 1, which is characterized in that step
In four, mentioning for attribute information is carried out using the best first increment selection method based on maximum correlation and minimum redundancy standard
It takes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910065147.1A CN109918713B (en) | 2019-01-23 | 2019-01-23 | Method for automatically acquiring gene knowledge of product concept design |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910065147.1A CN109918713B (en) | 2019-01-23 | 2019-01-23 | Method for automatically acquiring gene knowledge of product concept design |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109918713A true CN109918713A (en) | 2019-06-21 |
CN109918713B CN109918713B (en) | 2021-06-01 |
Family
ID=66960689
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910065147.1A Active CN109918713B (en) | 2019-01-23 | 2019-01-23 | Method for automatically acquiring gene knowledge of product concept design |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109918713B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110534155A (en) * | 2019-08-29 | 2019-12-03 | 合肥工业大学 | Fast packing formula Gene Selection Method based on maximal correlation minimal redundancy |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH09160987A (en) * | 1995-12-13 | 1997-06-20 | Hitachi Tohoku Software Kk | Method for searching for optimum solution |
US20070181743A1 (en) * | 2006-02-08 | 2007-08-09 | Lockheed Martin Corporation | Method for streamline traced external compression inlet |
CN101364244A (en) * | 2008-08-07 | 2009-02-11 | 同济大学 | Product gene modelling approach based on behavior semantic network identity model |
CN101777087A (en) * | 2010-01-26 | 2010-07-14 | 同济大学 | Product function gene modeling system and method |
CN104572590A (en) * | 2015-01-27 | 2015-04-29 | 大连大学 | Product structure design structure matrix partitioning method |
CN105956301A (en) * | 2016-05-10 | 2016-09-21 | 北京理工大学 | Function-concept-decision model-based reconfigurable machine tool configuration design method |
CN107944184A (en) * | 2017-09-04 | 2018-04-20 | 北京理工大学 | With reference to life system theory and the Method of Conceptual Design of FBS |
CN108846259A (en) * | 2018-04-26 | 2018-11-20 | 河南师范大学 | A kind of gene sorting method and system based on cluster and random forests algorithm |
-
2019
- 2019-01-23 CN CN201910065147.1A patent/CN109918713B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH09160987A (en) * | 1995-12-13 | 1997-06-20 | Hitachi Tohoku Software Kk | Method for searching for optimum solution |
US20070181743A1 (en) * | 2006-02-08 | 2007-08-09 | Lockheed Martin Corporation | Method for streamline traced external compression inlet |
CN101364244A (en) * | 2008-08-07 | 2009-02-11 | 同济大学 | Product gene modelling approach based on behavior semantic network identity model |
CN101777087A (en) * | 2010-01-26 | 2010-07-14 | 同济大学 | Product function gene modeling system and method |
CN104572590A (en) * | 2015-01-27 | 2015-04-29 | 大连大学 | Product structure design structure matrix partitioning method |
CN105956301A (en) * | 2016-05-10 | 2016-09-21 | 北京理工大学 | Function-concept-decision model-based reconfigurable machine tool configuration design method |
CN107944184A (en) * | 2017-09-04 | 2018-04-20 | 北京理工大学 | With reference to life system theory and the Method of Conceptual Design of FBS |
CN108846259A (en) * | 2018-04-26 | 2018-11-20 | 河南师范大学 | A kind of gene sorting method and system based on cluster and random forests algorithm |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110534155A (en) * | 2019-08-29 | 2019-12-03 | 合肥工业大学 | Fast packing formula Gene Selection Method based on maximal correlation minimal redundancy |
Also Published As
Publication number | Publication date |
---|---|
CN109918713B (en) | 2021-06-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106096727B (en) | A kind of network model building method and device based on machine learning | |
CN108170765B (en) | Poverty-stricken and living fund assisting recommendation method based on multidimensional analysis of on-school behavior data | |
Purshouse | On the evolutionary optimisation of many objectives | |
CN111726349B (en) | GRU parallel network flow abnormity detection method based on GA optimization | |
CN109685653A (en) | A method of fusion deepness belief network and the monitoring of the credit risk of isolated forest algorithm | |
CN103164631B (en) | A kind of intelligent coordinate expression gene analyser | |
CN109326328A (en) | A kind of extinct plants and animal pedigree evolution analysis method based on pedigree cluster | |
Dan et al. | A synthesized data mining algorithm based on clustering and decision tree | |
CN117743719B (en) | Page element identification method | |
CN109918713A (en) | A kind of gene Automated Acquisition of Knowledge method of Product Conceptual Design | |
Kumar et al. | Application of genetic algorithms for the prioritization of association rules | |
CN112163069A (en) | Text classification method based on graph neural network node feature propagation optimization | |
Zhang et al. | A projection pursuit dynamic cluster model based on a memetic algorithm | |
Ortiz-Servin et al. | Nuclear fuel lattice performance analysis by data mining techniques | |
CN115795657A (en) | Optimization method based on design space reduction | |
Augugliaro et al. | ℓ1‐Penalized Methods in High‐Dimensional Gaussian Markov Random Fields | |
Li et al. | [Retracted] Text Knowledge Acquisition Method of Collaborative Product Design Based on Genetic Algorithm | |
CN117313576A (en) | Bayesian optimization method for analyzing importance degree of airfoil physical quantity | |
Czech | Novel Methods for Analyzing and Visualizing Phylogenetic Placements | |
CN114282112A (en) | Automatic recommendation method for signal reconstruction scheme | |
Verbruggen | and Luc De Raedt1, 2 | |
Diao et al. | Heuristic search for fuzzy-rough bireducts and its use in classifier ensembles | |
CN115630729A (en) | Energy consumption demand side energy consumption prediction method based on DBN deep confidence network | |
Le | Novel Strategies to Accelerate Search Algorithms in Data Reduction | |
Suhaimi et al. | Optimizing cluster of questions by using dynamic mutation in genetic algorithm |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |