CN109101698A - A kind of Feature Selection Algorithms based on injection molding model, device and storage medium - Google Patents
A kind of Feature Selection Algorithms based on injection molding model, device and storage medium Download PDFInfo
- Publication number
- CN109101698A CN109101698A CN201810796642.5A CN201810796642A CN109101698A CN 109101698 A CN109101698 A CN 109101698A CN 201810796642 A CN201810796642 A CN 201810796642A CN 109101698 A CN109101698 A CN 109101698A
- Authority
- CN
- China
- Prior art keywords
- variable
- injection molding
- characteristic
- molding model
- value
- 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
- 238000001746 injection moulding Methods 0.000 title claims abstract description 131
- 238000003860 storage Methods 0.000 title claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 title claims description 29
- 239000011159 matrix material Substances 0.000 claims description 45
- 238000004519 manufacturing process Methods 0.000 claims description 25
- 210000002569 neuron Anatomy 0.000 claims description 25
- 239000000243 solution Substances 0.000 claims description 19
- 238000001816 cooling Methods 0.000 claims description 18
- 238000002347 injection Methods 0.000 claims description 17
- 239000007924 injection Substances 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 11
- 239000000203 mixture Substances 0.000 claims description 6
- 238000000034 method Methods 0.000 abstract description 31
- 230000000694 effects Effects 0.000 abstract description 7
- 238000010187 selection method Methods 0.000 abstract 1
- 239000000047 product Substances 0.000 description 28
- 238000012549 training Methods 0.000 description 18
- 238000013461 design Methods 0.000 description 16
- 230000006870 function Effects 0.000 description 13
- 230000008569 process Effects 0.000 description 13
- 239000004033 plastic Substances 0.000 description 9
- 229920003023 plastic Polymers 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 238000004458 analytical method Methods 0.000 description 7
- 238000005034 decoration Methods 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 7
- 238000000465 moulding Methods 0.000 description 7
- 238000009826 distribution Methods 0.000 description 6
- 239000004615 ingredient Substances 0.000 description 5
- 239000000463 material Substances 0.000 description 5
- 238000005206 flow analysis Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 230000002068 genetic effect Effects 0.000 description 3
- 238000010008 shearing Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 241001269238 Data Species 0.000 description 2
- 239000000498 cooling water Substances 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 125000000174 L-prolyl group Chemical group [H]N1C([H])([H])C([H])([H])C([H])([H])[C@@]1([H])C(*)=O 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 239000006227 byproduct Substances 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000002826 coolant Substances 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 230000028161 membrane depolarization Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Genetics & Genomics (AREA)
- Physiology (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Injection Moulding Of Plastics Or The Like (AREA)
Abstract
本发明公开了一种基于注塑成型模型的特征选取方法，首先获取与注塑成型模型的各特征变量对应的样本数据，然后根据各样本数据计算与各特征变量对应的特征值，最后以预定义规则从各特征值中选取目标特征值，并将目标特征值对应的特征变量作为目标特征变量。该方法在确定注塑成型模型的特征变量时，先计算与各特征变量对应的特征值，最后从各特征值中选取目标特征值，并将目标特征值对应的特征变量作为目标特征变量，因此，最终的注塑成型模型的目标特征变量个数较少，简化了注塑成型模型的结构。本发明还公开了基于注塑成型模型的特征选取装置及存储介质，效果如上。
The invention discloses a feature selection method based on an injection molding model. Firstly, the sample data corresponding to each characteristic variable of the injection molding model is obtained, and then the characteristic value corresponding to each characteristic variable is calculated according to each sample data, and finally the pre-defined rule The target eigenvalue is selected from each eigenvalue, and the feature variable corresponding to the target eigenvalue is used as the target feature variable. When determining the characteristic variables of the injection molding model, this method first calculates the characteristic values corresponding to each characteristic variable, and finally selects the target characteristic value from each characteristic value, and uses the characteristic variable corresponding to the target characteristic value as the target characteristic variable. Therefore, The final injection molding model has fewer target feature variables, which simplifies the structure of the injection molding model. The invention also discloses a feature selection device and a storage medium based on the injection molding model, and the effect is as above.
Description
Technical field
The present invention relates to injection mould V-neck V domain, in particular to a kind of Feature Selection side based on injection molding model
Method, device and storage medium.
Background technique
The quality of injecting products is affected by various processing conditions, and process conditions include injected plastics material attribute, Shooting Technique
Parameter, injection mold, the filling of melt and cooling etc..Wherein, due to injected plastics material attribute not malleable, pass through change
Injected plastics material attribute is difficult to realize with the quality for adjusting injecting products；And pass through optimization molding proces s parameters, injection mold, solution
Filling and the factors such as cooling can be compared with malleable to the quality of production and production efficiency of injecting products.
It include many technological parameters in injection molding process, such as injection pressure, dwell pressure, melt temperature, maximum lock
Mould power, maximum wall shearing stress, cooling time, mold temperature, filling time, dwell time etc..Different technological parameters are to note
The quality of production of molding product and the influence degree of production efficiency are also different, and traditional method is by all conducts of numerous technological parameters
The design variable of injection molding model so that structure is complicated for injection molding model, and passes through since design variable is more
When the injection molding model solves the objective function of injecting products, cause its calculating process complicated；Wherein, objective function
It can be the quality of production and production efficiency of injecting products.
Therefore, how to reduce the design variable in injection molding model with the structure for simplifying injection molding model is this field
Technical staff's problem to be solved.
Summary of the invention
The purpose of the present invention is to provide a kind of Feature Selection Algorithms based on injection molding model, device and storages to be situated between
Matter reduces the design variable in injection molding model and simplifies the structure of injection molding model.
To achieve the above object, the embodiment of the invention provides following technical solutions:
Firstly, the embodiment of the invention provides a kind of Feature Selection Algorithms based on injection molding model, comprising:
Obtain sample data corresponding with each characteristic variable of injection molding model；
Characteristic value corresponding with each characteristic variable is calculated according to each sample data；
Object feature value is chosen from each characteristic value with predefined rule, and by the corresponding spy of the object feature value
Variable is levied as target signature variable.
Preferably, described characteristic value corresponding with each characteristic variable is calculated according to each sample data to include:
By each sample data composition sample matrix and covariance matrix is calculated according to the sample matrix；
The characteristic value corresponding with each characteristic variable is calculated using the covariance matrix.
Preferably, described by each sample data composition sample matrix and according to the calculating covariance of the sample matrix
Matrix includes:
Each sample data is formed into the sample matrix；
Become using in the sample matrix with each corresponding each sample data of characteristic variable and the feature
The number of amount calculates average characteristics numerical value；
Calculate the difference of each sample data Yu the average characteristics numerical value；
It is combined each difference to obtain the covariance matrix.
Preferably, described that object feature value is chosen from each characteristic value with predefined rule and the target is special
After the corresponding characteristic variable of value indicative is as target signature variable, further includes:
Using the target signature variable as input variable；
Using the quality of production of injecting products and production efficiency as output variable；
The injection molding model is constructed using the input variable and the output variable and to the injection mould
Type optimizes；
The injection molding model after optimization is solved to obtain the optimal solution of each target signature variable.
Preferably, described using the input variable and output variable building injection molding model and to the injection molding
Forming model, which optimizes, includes:
The injection molding model is constructed using the input variable and the output variable；
Set a variety of neuron numbers of the hidden layer of the injection molding model；
Using test sample data corresponding with the input variable and verifying sample data, from a variety of neurons
The optimal neuron number of the injection molding model is chosen in number；
It is obtained using the input variable, the output variable and the corresponding hidden layer of the optimal neuron number excellent
The injection molding model after change.
Preferably, described that object feature value is chosen from each characteristic value with predefined rule and the target is special
The corresponding characteristic variable of value indicative includes: as target signature variable
The characteristic value for being greater than targets threshold is chosen from each characteristic value, and will be greater than the spy of the targets threshold
Target signature variable of the corresponding characteristic variable of value indicative as the injection molding model.
Preferably, the target signature variable includes dwell pressure, melt temperature, cooling time, mold temperature and guarantor
Press the time.
Then, the embodiment of the invention discloses a kind of Feature Selection devices based on injection molding model, comprising:
Sample data obtains module, for obtaining sample data corresponding with each characteristic variable of injection molding model；
Characteristic value calculating module, for calculating feature corresponding with each characteristic variable according to each sample data
Value；
Target signature specification of variables module, for choosing object feature value from each characteristic value with predefined rule,
And using the corresponding characteristic variable of the object feature value as target signature variable.
Secondly, the Feature Selection device the embodiment of the invention discloses another kind based on injection molding model, comprising:
Memory, for storing computer program；
Processor realizes as above described in any item be based on for executing the computer program stored in the memory
The step of Feature Selection Algorithms of injection molding model.
Finally, being deposited on computer readable storage medium the embodiment of the invention discloses a kind of computer readable storage medium
Computer program is contained, is realized when computer program is executed by processor as above described in any item based on injection molding model
The step of Feature Selection Algorithms.
As it can be seen that a kind of Feature Selection Algorithms based on injection molding model disclosed by the invention, obtain first and are molded into
Then the corresponding sample data of each characteristic variable of pattern type calculates feature corresponding with each characteristic variable according to each sample data
Value finally chooses object feature value with predefined rule from each characteristic value, and the corresponding characteristic variable of object feature value is made
For target signature variable.Therefore, this method is when determining the characteristic variable of injection molding model, be not by it is all be molded into
The relevant characteristic variable of pattern type is all used as target signature variable, but first calculates characteristic value corresponding with each characteristic variable, most
Choose object feature value from each characteristic value afterwards, and using the corresponding characteristic variable of object feature value as target signature variable, because
This, the target signature variable number of final injection molding model is less, simplifies the structure of injection molding model.The present invention is also
Feature Selection device and storage medium based on injection molding model are disclosed, effect is as above.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of Feature Selection Algorithms flow diagram based on injection molding model disclosed by the embodiments of the present invention；
Fig. 2 is a kind of Feature Selection apparatus structure schematic diagram based on injection molding model disclosed by the embodiments of the present invention；
Fig. 3 is another Feature Selection apparatus structure signal based on injection molding model disclosed by the embodiments of the present invention
Figure；
Fig. 4 (a) is a kind of injection model front view of automobile interior decoration panel disclosed by the embodiments of the present invention；
Fig. 4 (b) is a kind of injection model sectional view of automobile interior decoration panel disclosed by the embodiments of the present invention；
Fig. 5 is a kind of accumulative total of variance curve graph of each characteristic variable disclosed by the embodiments of the present invention；
Fig. 6 is a kind of ELM extreme learning machine model schematic of automobile interior decoration panel disclosed by the embodiments of the present invention；
Fig. 7 (a) is the training result curve graph of the first ELM extreme learning machine model disclosed by the embodiments of the present invention；
Fig. 7 (b) is the training result curve graph of second of ELM extreme learning machine model disclosed by the embodiments of the present invention；
Fig. 7 (c) is the training result curve graph of the third ELM extreme learning machine model disclosed by the embodiments of the present invention；
Fig. 7 (d) is the training result curve graph of the 4th kind of ELM extreme learning machine model disclosed by the embodiments of the present invention；
Fig. 8 (a) is the predicted value of two target function values disclosed by the embodiments of the present invention and the scatter chart of experiment value；
Fig. 8 (b) is the output error curve graph of two objective functions disclosed by the embodiments of the present invention；
Fig. 9 is a kind of ELM learning machine-non-dominated sorted genetic algorithm flow diagram disclosed by the embodiments of the present invention；
Figure 10 is the initial population spatial distribution scatter plot disclosed by the embodiments of the present invention generated at random；
Figure 11 is the initial population evolution convergence curve graph disclosed by the embodiments of the present invention generated at random；
Figure 12 be Evolution of Population disclosed by the embodiments of the present invention to first front end in 100 generations Pareto optimal solution distribution
Scatter plot.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a kind of Feature Selection Algorithms based on injection molding model, device and storage medium,
Reduce the design variable in injection molding model and simplifies the structure of injection molding model.
Referring to Figure 1, Fig. 1 is a kind of Feature Selection Algorithms stream based on injection molding model disclosed by the embodiments of the present invention
Journey schematic diagram, this method comprises:
S101, acquisition sample data corresponding with each characteristic variable of injection molding model.
Specifically, injection molding model is using the technological parameter of injection molding as input variable, injection molding in the present embodiment
Product quality, production efficiency of product etc. are used as output variable, in conjunction with input variable, output variable, the constraint item of injection molding
The mathematical model of the foundation such as part.Wherein, the product quality of injecting products is properly termed as buckling deformation again, production efficiency is properly termed as
Forming period；Input variable is the arbitrary characteristic variable in each characteristic variable in the present embodiment, and each characteristic variable is main
There are melt temperature, mold temperature, injection pressure, injection time, dwell pressure, dwell time, cooling time etc..For example, g1
(X), g2(X) constraint condition as injection molding model, the quality of production (buckling deformation) W of injection molding modelapg(X)
With production efficiency (forming period) T of injection molding modelcyc(X) as the output variable (objective function) of injection molding model,
Using multiple characteristic variables as the input variable X of injection molding model, following injection molding model is established:
Wherein, objective function Wapg(X) and Tcyc(X) it is obtained by 2017 mold flow analysis of Moldflow Adviser, constrains item
Part g1(X) indicate that plastics have to fill up all models chamber, constraint condition g2(X) indicate that plastics must be within a preset time full of all
Type chamber, Nt_cavAnd Nf_cavThe number of type chamber for respectively indicating the number of whole type chambers and having filled up, tfAnd tf_preIt respectively indicates
Filling time and want seeking time.The foundation of the model can use Moldflow software carry out simulation injection molding experiment and to data into
Row processing, to establish the Shooting Technique Non-linear Optimal Model an of multiple objective function, multi input variable.Certainly, injection molding
The foundation of model can have different modeling patterns according to the difference of objective function, herein and be not construed as limiting.
Further, sample data is data value corresponding with each characteristic variable, for example, when corresponding to characteristic variable pressure maintaining
Between, sample data 20s, 21s, 22s, 21.5s.Certainly, according to actual injection molding environment, in the embodiment of the present invention
The size of the number of the sample data of characteristic variable and sample data herein and is not construed as limiting.Each sample data can be by just
It hands over and is obtained by 2017 streaming emulation of Mold flow Adviser.
S102, characteristic value corresponding with each characteristic variable is calculated according to each sample data.
Specifically, the size of the characteristic value of each characteristic variable represents this feature variable to injection mould in the present embodiment
The influence degree of type.I.e. characteristic value is bigger, the quality of production and production of the corresponding characteristic variable to the injecting products of injection molding
Efficiency influence degree is bigger.Its specific calculating process will describe in detail in next embodiment.
S103, object feature value is chosen from each characteristic value with predefined rule, and by the corresponding feature of object feature value
Variable is as target signature variable.
Specifically, predefined rule can have following methods in the present embodiment, first way can incite somebody to action first
Then each characteristic value is overlapped to obtain the sum of total characteristic value, then calculate each spy by each characteristic value by being ranked up from small greatly
Value indicative accounts for the accounting of the sum of entire total characteristic value, is then ranked up each accounting is descending, biggish accounting is folded
Add, if the stack result of larger accounting is already close to 100%, remaining characteristic value accounts for the accounting very little of the sum of entire characteristic value,
It then can be using the corresponding characteristic variable of the characteristic value of larger accounting as target signature variable.For example, if characteristic variable be 9,9
A characteristic variable corresponding eigenvalue is respectively 4.988,2.459,0.971,0.458,0.097,0.012,0.010,0.003,
0.000,0.001, then the accounting that the corresponding characteristic value of the first two characteristic variable accounts for the sum of all characteristic values adds up to be 82.743%,
At this point, the accounting of the accounting of third characteristic value and the 4th characteristic value and first characteristic value and second characteristic value is carried out
Being accumulated by accumulative accounting is 98.619%.At this point, first four characteristic value account for the accounting of the sum of total characteristic value already close to
100%, the accounting that then five characteristic values account for the sum of all characteristic values adds up to be 1.381%.Therefore, at this point it is possible to by preceding four
A characteristic value becomes the corresponding feature of this four characteristic values as object feature value, i.e., 4.988,2.459,0.971,0.458
Amount is used as target signature variable.The second way can be set a target accounting threshold value, each characteristic value accounted for all characteristic values
The sum of accounting be ranked up from big to small, and added up by sequence from big to small, when accumulated result is more than or equal to target
After accounting threshold value, using the accumulative corresponding characteristic variable of characteristic value as target signature variable, for example, with numerical value mentioned above
For, 98% is set by target accounting threshold value；The accounting that each characteristic value accounts for the sum of all characteristic values is calculated, then, by accounting
It is overlapped from big to small, at this point, first four characteristic value accounts for the accounting 98.619% of the sum of all characteristic values, is greater than target accounting
Threshold value 98%, therefore, using the corresponding characteristic variable of first four characteristic value as target signature variable.The third mode, it can
One targets threshold is set, will be greater than the characteristic value of targets threshold as object feature value, and the object feature value is corresponding
Target signature variable of the characteristic variable as injection molding model in the present embodiment, is implemented using the third mode as preferred
Example chooses the characteristic value for being greater than targets threshold, and the corresponding feature of characteristic value that will be greater than targets threshold that is, from each characteristic value
Target signature variable of the variable as injection molding model.Certainly, predefined rule is according to the actual environment of injection molding, can be with
The size of the target accounting threshold value, targets threshold and the characteristic value that have other regular also, mentioned above does not limit herein
It is fixed.
As it can be seen that a kind of Feature Selection Algorithms based on injection molding model disclosed by the invention, obtain first and are molded into
Then the corresponding sample data of each characteristic variable of pattern type calculates feature corresponding with each characteristic variable according to each sample data
Value finally chooses object feature value with predefined rule from each characteristic value, and the corresponding characteristic variable of object feature value is made
For target signature variable.Therefore, this method is when determining the characteristic variable of injection molding model, be not by it is all be molded into
The relevant characteristic variable of pattern type is all used as target signature variable, but first calculates characteristic value corresponding with each characteristic variable, most
Choose object feature value from each characteristic value afterwards, and using the corresponding characteristic variable of object feature value as target signature variable, because
This, the target signature variable number of final injection molding model is less, simplifies the structure of injection molding model.
Based on the above embodiment, in the present embodiment, feature corresponding with each characteristic variable is calculated according to each sample data
Value includes:
Each sample data is formed into sample matrix and covariance matrix is calculated according to the sample matrix；
Characteristic value corresponding with each characteristic variable is calculated using covariance matrix.
Wherein, as preferred embodiment, each sample data is formed into sample matrix and association side is calculated according to sample matrix
Poor matrix includes:
Each sample data is formed into sample matrix；
It is calculated using the number of each sample data corresponding with each characteristic variable in sample matrix and characteristic variable average
Character numerical value；
Calculate the difference of each sample data Yu average characteristics numerical value；
It is combined each difference to obtain covariance matrix.
Specifically, sample matrix is the corresponding sample data composition of each characteristic variable, covariance square in the present embodiment
Battle array is the difference between the average value by the corresponding sample data of each characteristic variable sample data corresponding with all characteristic variables
Value combination obtains.For example, if characteristic variable is 9, respectively x1, x2, x3, x4, x5, x6, x7, x8, x9；Each characteristic variable pair
There should be 49 sample datas, then the sample matrix constituted is then X=[x1,x2,...,x9], wherein xi=[xi1,xi2,...,
xi49], i=1,2 ..., 9.Then, it is calculated according to the number of the corresponding sample data of all characteristic variables and characteristic variable flat
Equal character numerical valueAverage characteristics numerical valueCalculation formula it is as follows:
Then the sample data and average characteristics numerical value of each characteristic variable are calculatedBetween difference, difference calculate
Formula is as follows:
In above-mentioned two formula, i=1,2 ..., 9, each characteristic variable xiIn include 49 sample datas.Therefore, it assists
Variance matrix C can be indicated are as follows:
Obtain the characteristic value that each characteristic variable is calculated after covariance matrix and the corresponding feature of each characteristic variable to
Amount.Other than directly characterizing each characteristic variable to the influence degree of the injecting products of injection molding with the size of characteristic value,
The influence degree to the injecting products of injection molding can be characterized by the variance of each feature vector.Each characteristic variable is one
A principal component.For example, being λ by the characteristic value that above formula calculates each characteristic variableiAnd feature vector ηi, then it is each it is main at
Dividing can indicate are as follows:
yi=ηi TX
For each principal component yiIts variance can be calculate by the following formula:
Var(yi)=Var (ηi TX)=ηi TVar(X)ηi=λiηi Tηi
Wherein, i=0,1 ..., 9；As it can be seen that the variance of each principal component characteristic value corresponding with its is directly proportional, that is,
It says, the size of the characteristic value of each characteristic variable determines the variance size of each characteristic variable.Therefore, the characteristic value of characteristic variable
Or variance can represent this feature variable to the quality of production of injecting products and the influence degree size of production cycle.
It should be noted that in the present embodiment, of the number of characteristic variable, the corresponding sample data of each characteristic variable
Number is not construed as limiting herein.
Based on the above embodiment, in the present embodiment, object feature value is chosen from each characteristic value with predefined rule, and
After using the corresponding characteristic variable of object feature value as target signature variable, further includes:
Using target signature variable as input variable；
Using the quality of production of injecting products and production efficiency as output variable；
Using input variable and output variable building injection molding model and injection molding model is optimized；
Injection molding model after optimization is solved to obtain the optimal solution of each target signature variable.
Wherein, as preferred embodiment, using input variable and output variable building injection molding model and to injection molding
Forming model, which optimizes, includes:
Injection molding model is constructed using input variable and output variable；
Set a variety of neuron numbers of the hidden layer of injection molding model；
Using test sample data corresponding with input variable and verifying sample data, chosen from a variety of neuron numbers
The optimal neuron number of injection molding model；
Being molded into after being optimized using input variable, output variable and the corresponding hidden layer of optimal neuron number
Pattern type.
Specifically, injection molding model can be the ELM model established by ELM extreme learning machine in the present embodiment,
Injection molding model mainly has three-decker.That is input layer, intermediate hidden layer and output layer.This is illustrated with specific example below
The establishment process and solution procedure of injection molding model.For example, if the target signature variable finally chosen is respectively mold temperature
Degree, melt temperature, dwell pressure, dwell time, cooling time.Then using five characteristic variables as the defeated of injection molding model
Enter variable, using product quality and production cycle as the output variable of injection molding model.Then the mind of intermediate hidden layer is determined again
Through first number, the injection molding after combining input variable, output variable to be optimized after the number of final neuron is obtained
Model.Wherein, the optimization process of injection molding model is as follows: setting 5,6,7,8 for the number of hidden neuron respectively, then
Respectively obtaining input variable is 5, output variable 2, the injection molding model that hidden neuron number is 5,6,7,8.Then, sharp
Four injection molding models are tested with test sample data, verifying sample data and training sample data, are then obtained each
The training error of a injection molding model output therefrom chooses the corresponding hidden neuron number of minimum training error as optimal
Neuron number, in the present embodiment, when hidden neuron number is 6, corresponding output error is minimum, then by 6 minds
Through injection molding model of first number as optimal neuron number and after establishing optimization.
A kind of Feature Selection device based on injection molding model disclosed by the embodiments of the present invention is introduced below, is asked
Referring to fig. 2, Fig. 2 is a kind of Feature Selection apparatus structure schematic diagram based on injection molding model disclosed by the embodiments of the present invention,
The device includes:
Sample data obtains module 201, for obtaining sample data corresponding with each characteristic variable of injection molding model；
Characteristic value calculating module 202, for calculating spy corresponding with each characteristic variable according to each sample data
Value indicative；
Target signature specification of variables module 203, for choosing target signature from each characteristic value with predefined rule
Value, and using the corresponding characteristic variable of the object feature value as the target signature variable.
As it can be seen that a kind of Feature Selection device based on injection molding model disclosed by the embodiments of the present invention, first with sample
Notebook data obtains module and obtains sample data corresponding with each characteristic variable of injection molding model, then characteristic value calculating module
Characteristic value corresponding with each characteristic variable is calculated according to each sample data, ideal characteristic variable setting module is with predefined rule
Object feature value is then chosen from each characteristic value, and using the corresponding characteristic variable of object feature value as target signature variable.Cause
This, this method is not by all features relevant to injection molding model when determining the characteristic variable of injection molding model
Variable is all used as target signature variable, but first calculates characteristic value corresponding with each characteristic variable, finally selects from each characteristic value
Object feature value is taken, and using the corresponding characteristic variable of object feature value as target signature variable, therefore, final injection molding
The target signature variable number of model is less, simplifies the structure of injection molding model.
Feature Selection device the embodiment of the invention also discloses another kind based on injection molding model, refers to Fig. 3, figure
3 be another Feature Selection apparatus structure schematic diagram based on injection molding model disclosed by the embodiments of the present invention, the device packet
It includes:
Memory 301, for storing computer program；
Processor 302, for executing the computer program stored in the memory to realize as above any one implementation
The step of Feature Selection Algorithms based on injection molding model that example is mentioned.
Feature Selection device of the another kind based on injection molding model disclosed by the embodiments of the present invention, effect are as previously mentioned
A kind of Feature Selection Algorithms based on injection molding model effect, in this not go into detail.
This programme in order to better understand, a kind of computer readable storage medium provided in an embodiment of the present invention, computer
It is stored with computer program on readable storage medium storing program for executing, realizes that any embodiment as above is mentioned when computer program is executed by processor
The Feature Selection Algorithms based on injection molding model the step of.
A kind of computer readable storage medium disclosed by the embodiments of the present invention, effect are as mentioned above a kind of based on injection molding
The effect of the Feature Selection Algorithms of forming model, in this not go into detail.
Technical solution provided by the invention is described in detail by taking the inner veneer of automobile as an example below.Refer to Fig. 4
(a) and Fig. 4 (b), Fig. 4 (a) are a kind of injection model front view of automobile interior decoration panel disclosed by the embodiments of the present invention, Fig. 4 (b)
For a kind of injection model sectional view of automobile interior decoration panel disclosed by the embodiments of the present invention.The present embodiment is with automobile interior decoration panel
Study subject designs moulding model by Pro/E Creo Parametric2.0 Three-dimensional Design Software, from overall structure
On from the point of view of, have an irregular elliptical openings in the middle part of plastic, multiple raised bolts are arranged at bottom, between multiple accessories
Connection.Its total outer dimension are as follows: 240.00mm × 138.651mm × 3.495mm, in addition to convex portion, whole thickness distribution
It is relatively uniform.
By 3 d part model import CAE software Moldflow Adviser 2017, design shaping dies, flow passage system,
Cooling system, and it is filled, pressure maintaining, cooling and warping Analysis.We are set as cavity running gate system herein.Separately
Outside, the running gate system of injection mold, the forming quality for choosing whether appropriately to will have a direct impact on plastic of gate location, improperly
Gate location may cause short a series of bad phenomenons and the defect such as penetrate, spray, being detained, being recessed.In order to obtain optimal pour
Mouth position, the present embodiment obtain the Best gate location of the plastic part, cast gate using gate location analytic function in Moldflow
Position coordinates are X:249.982mm Y:10.695mm Z:10.000mm.
According to the specific size of the part, die size is defined as A plate thickness 125mm, B plate thickness 100mm, entire die size
For 800mm × 900mm × 225mm.The material that mold is selected is Tool Steel P-20:Generic.Runner and cast gate are related
Parameter is shown in Table 1-1.
Table 1-1 flow passage system parameter
In the humidity control system of entire mold, cooling system is extremely important.In general, in the entire injection molding of a plastic
In forming period, more than half of entire forming period can take up cooling time, so the quality of design of Cooling System can be straight
Connect the production efficiency for influencing product.The design of cooling system mainly adjusts the temperature of entire mold by cooling water, this is
System is mainly concerned with the positions and dimensions and distribution form of cooling water, further includes each seed ginseng such as temperature, flow velocity of cooling medium
Number.8 cooling circuits are devised in entire mold, have 8 pipelines and 4 circuits (respectively in type chamber and type core in each part
In).
Three-dimensional injection molding model is initially set up, after establishing threedimensional model, needs to obtain the sample data of injection model, sample
The acquisition process of notebook data is as follows: setting optimization aim as quality (buckling deformation) Wapg(X) and efficiency (forming period) tcyc
(X), 9 characteristic variables include mold temperature Tmold, melt temperature Tmelt, dwell pressure Pp, dwell time tp, t cooling timec、
Maximum clamp force Fclp, maximum wall shearing stress Pw, filling time tfilWith injection pressure PinjDeng.
The characteristics of according to Moldflow 2017 mold flow analysis of Adviser, setting section characteristic variable is only needed, other are corresponding
Parameter can be obtained by emulation, choose mold temperature T hereinmold, melt temperature TmeltWith dwell time tpThree variables divide and take 6
A level carries out orthogonal test.In view of the attribute of PP material itself, the range for setting mold temperature is 50 DEG C -75 DEG C, between taking
Every 5 DEG C of 6 level points；The range of melt temperature is 230 DEG C -255 DEG C, takes 6 level points at 5 DEG C of interval；Dwell time takes
7s-12s takes 6 level points of interval 1s, devises orthogonal arrage, then emulated by Moldflow Adviser 2017, so that it may
Obtain the corresponding experimental data of other characteristic variables and target variable.Table 1-2 is partial orthogonality test sample tables of data.
Table 1-2 partial orthogonality test sample tables of data
After obtaining above-mentioned sample data, need to choose target signature variable from each characteristic variable, specific choice method is adopted
With Principal Component Analysis, it is as follows to be specifically chosen process:
It is related to 9 decision variables altogether herein, each variable there are 49 empirical values, that is, the sample matrix constituted is X=
[x1,x2,...,x9], wherein xi=[xi1,xi2,...,xi49]T, i=1,2 ..., 9.It needs before principal component analysis to sample
Carry out centralization, each sample dataI.e.The covariance matrix of sample matrix is
Its eigenvalue λ is acquired by covariance matrix againiAnd corresponding feature vector ηi, after can be obtained by mapping in this way
Principal component yi=ηi TX, for each ingredient yiIts variance is calculated as follows,
Var(yi)=Var (ηi TX)=ηi TVar(X)ηi=ηi TC49×49ηi=λiηi Tηi
By formula it can be seen that the variance of each ingredient it is corresponding feature vector it is directly proportional, i.e. the size of characteristic value can
To reflect the variance size that there emerged a principal component, the weight sequencing of each principal component can be obtained by according to the sequence of characteristic value size, then
It is descending by its weight according to the criterion of very big accumulative variance contribution ratio, add up to choose p principal component.This paper PCA is analyzed
To population variance explanation be shown in Table 1-3, accumulative total of variance figure is shown in that Fig. 5, Fig. 5 are a kind of each characteristic variable disclosed by the embodiments of the present invention
Accumulative total of variance curve graph.
The population variance that table 1-3PCA is analyzed is explained
Pass through formula Var (yi)=Var (ηi TX)=ηi TVar(X)ηi=ηi TC49×49ηi=λiηi Tηi, it is noted that standardization
The eigenvalue λ of covariance matrix afterwardsiIt is numerically equal to the variance Var (y of its corresponding feature vectori), become in initial characteristics
We take out all 9 characteristic values and corresponding feature vector in quantity space.As can be seen from table 1-3 that first and second is main
The characteristic value of ingredient is both greater than 1, and the variance (characteristic value) of the two principal components is respectively 4.988 and 2.459, and accumulative accounting is
82.743%.In order to make principal component matrix cover the information of original data space as far as possible, need the 3rd 4th principal component
Take into account, their population variance accounting is respectively 10.792% and 5.084%, at this point, accumulative variance accounting reaches
98.619%. and the 5th to the ninety percentth point of characteristic value is respectively less than 0.1, add up variance accounting is overall 1.381%.Therefore only
Preceding 4 principal components are taken, and the characteristic vector space for being approximately considered preceding 4 principal components composition can express the complete of original variable space
Portion's information.
In addition, can more be clear that the corresponding characteristic value of each feature vector of covariance matrix from Fig. 5, i.e.,
The variance of each principal component.Particularly, to the aggregate-value of the 4th principal component (accumulative variance accounting) already close to total value, this is more
Add and intuitively illustrates that preceding 4 principal components replace population sample information there are enough reliabilities.
More specifically, table 1-4 lists the component matrix for each principal component being made of the linear combination of 9 variables,
4 principal components comprising enough original variable data informations have been obtained from previous analysis, added the 5th to the in the table again
As a comparison, specific component matrix and vector coefficient are shown in Table 1-4 to 9 ingredients.
Table 1-4 component matrix and vector coefficient
Table 1-4 gives principal component loading matrix, and each charge values of listing all show each variable and related principal component
Related coefficient, in the case of the 1st column, 0.990 is actually the related coefficient of dwell pressure and the 1st principal component.In addition, for standard
Dwell pressure column vector data after change carry out it with the 1st principal component to return its coefficient of determination R that is easy to get2=0.98, then phase
Its load R=0.99 answered, i.e. load of the dwell pressure in the 1st principal component.
Based on the above analysis, it can be noted that in preceding 4 principal components, dwell pressure, melt temperature, cooling time, mold
The load of this 5 variables of temperature and dwell time is all considerable degree of to be greater than other 4 variables, especially in the 1st principal component,
Its related coefficient shows these the 1st principal components of variable with stronger correlativity all close to 1, and other 4 variables
Load it is smaller, there are two types of explain: first, it is smaller to overall influence, be similar to noise factor；Second, in dimensionality reduction mistake
The information lost in journey is more, at this point, 4 obtained above principal component can almost express whole letters in original sample space
Breath, therefore the information of 4 variables loss here is larger relative to its own, but for totality, can be neglected.
Both possible conclusions are consistent: i.e. this 4 injection pressure, maximum clamp force, maximum wall shearing stress, filling time changes
Amount is smaller to general impacts, for another angle, we have observed that mold temperature Tmold, melt temperature Tmelt, dwell pressure
Pp, dwell time tpWith t cooling timecThis 5 variables are the main affecting factors (ingredient) in original sample space.Therefore below
In analysis, we will be analyzed this 5 variables in all original sample spaces, and 5 variables can compare
It is complete to indicate whole information content.
After obtaining 5 target signature variables, using target signature variable as input variable, by product quality and production life cycle
Injection molding model is established as output variable, initial setting hidden neuron number is 6, that is, establishes three layers of single hidden layer
ELM extreme learning machine, as shown in fig. 6, Fig. 6 is a kind of ELM model signal of automobile interior decoration panel disclosed by the embodiments of the present invention
Figure.
Input variable is that the previously mentioned characteristic variable that carried out is extracted and 5 designs in the orthogonal test after dimensionality reduction
Variable, i.e. mold temperature Tmold, melt temperature Tmelt, dwell pressure Pp, dwell time tpWith t cooling timec, output variable contains
Two objective functions: quality (buckling deformation) and efficiency (forming period).
In order to more accurately determine the number of hidden neuron, and 5,6,7,8 are set by hidden neuron respectively, and
Training analysis is carried out, as a result such as Fig. 7 (a), it is 5,6,7,8 that 7 (b), 7 (c), 7 (d), 4 figures, which respectively correspond hidden neuron,
The case where, Fig. 7 (a) is the training result curve graph of the first ELM model；Fig. 7 (b) is the training result of second of ELM model
Curve graph；Fig. 7 (c) is the training result curve graph of the third ELM model；Fig. 7 (d) is the training result of the 4th kind of ELM model
Curve graph；Train is training sample data in figure, and Test is verifying sample data, and Validation is test sample data, Y
Axis is output error.By such as Fig. 7 (a), the comparison of 7 (b), 7 (c), 7 (d), 4 figures, it can be found that 7 (a), 7 (b), 7 (c)
Occur Premature Convergence, especially 7 (a) figures in figure, can not be trained, and 7 (c) and 7 (d) training error is respectively
4.3138 and 2.5654, error is larger.And 7 (c) and 7 (d) have over-fitting.When hidden neuron number is 6, i.e., 7
(b) figure error drops to 0.32977, this error is in tolerance interval, so finally determining that hidden neuron number is 6.
In the application, the main function of ELM extreme learning machine is by the predicted value and experiment value after neural metwork training
Between error come for later NSGA II (non-dominated sorted genetic algorithm) carry out fitness assignment, Fig. 8 (a) be two targets
The predicted value of functional value and the scatter chart of experiment value, wherein Tc is forming period experiment value, and Tcp is forming period prediction
Value, Warpage are buckling deformation experiment value, and Warpagep is buckling deformation predicted value, and Fig. 8 (b) is the defeated of two objective functions
Error curve diagram out, for forming period, whole prediction average error is 9.6%, removes the average mistake of extreme point
Difference only 0.016；It is whole to predict that error is 8.7% for buckling deformation amount, as can be seen that depolarization from Fig. 8 (b)
Outside the other biggish extreme point of error, most predicted values and experiment value have good compatibility.
Fig. 9 is referred to, Fig. 9 is a kind of ELM learning machine-non-dominated sorted genetic algorithm process disclosed by the embodiments of the present invention
Schematic diagram；As shown in figure 9, carrying out mold flow analysis according to orthogonal arrage first in Moldflow Adviser 2017, obtain initial
Training data；Then, training data is imported into instruction prediction target function value in ELM extreme learning machine, according to ELM extreme learning machine
Training error result carries out PARITO non-dominated ranking to sample, calculates fitness value, and initialization population carries out NSGAII heredity
Operation stops evolving until reaching evolutionary generation requirement.Figure 10 is the initial population spatial distribution scatter plot generated at random, figure
11 be the initial population evolution convergence curve graph generated at random, and after evolving to for 100 generations, the error of Evolution of Population iterative process is suitable
Angle value is answered to start to tend towards stability, in other words, the Pareto optimal solution set of this when also gradually tends towards stability, and Figure 12 is population
Evolve to the distribution scatter plot of the Pareto optimal solution of first front end in 100 generations.
On the basis of Figure 11, designer can be very easily from each point is found in Pareto optimal solution set in figure
Corresponding design variable, i.e. forming parameters substitute into Moldflow Adviser2017 and carry out mold flow analysis, later to examine
The superiority for testing its design scheme, particularity and requirement further according to deisgn product itself select relatively optimal technological design
Parameter.
Table 1-5 is the Moldflow simulation analysis of corresponding 4 Pareto optimal solutions in Figure 12 as a result, can from Figure 12
Seen with apparent when guaranteeing quality (buckling deformation), can unavoidably extend the forming period of product, so,
Under the premise of having no special requirements, we select the scheme of compromise as far as possible.
Table 1-5Pareto optimal solution (part)
We choose the 4th group of solution in table 1-5 in Pareto optimal solution set, as the molding proces s parameters after optimization, and
It is compared with the forming results for the technological parameter being not optimised, focuses on quality (buckling deformation) defect and production effect of product
In the difference of rate (forming period), on the one hand, be not optimised handicraft product that is, from the point of view of the quality problems of product from buckling deformation amount
The warpage variable for being more than maximum nominal deviation is 21.2%, and the buckling deformation amount after optimization more than maximum nominal deviation is
4.95%, low buckling deformation region accounting is increased to 74.4% from 57.3%；On the other hand, come from the molding cycle of entire product
It says, molding cycle when being not optimised is 40.55s, and after Optimizing Process Parameters, the molding cycle of product is 32.41s, entire to form
Period reduces about 20.07%, and under the premise of guaranteeing product quality, production efficiency is also improved.In addition to the two are main
Outside aspect, for the simulation result of optimization front and back product there are also more differences, specific parameter is shown in Table 1-6.
The comparison of table 1-6 optimum results
The analysis that the above several optimal solutions only concentrated to Pareto solution carry out compares, and demonstrates the suitable of this paper algorithm model
With property, when coping with specific design problem, product designer can have stressing property to go to choose according to specific design requirement
Pareto optimal solution, to obtain optimal design scheme.
Above to a kind of Feature Selection Algorithms based on injection molding model, device and storage medium provided herein
It is described in detail.Specific examples are used herein to illustrate the principle and implementation manner of the present application, the above reality
The explanation for applying example is merely used to help understand the present processes and its core concept.It should be pointed out that for the art
For those of ordinary skill, under the premise of not departing from the application principle, can also to the application, some improvement and modification can also be carried out,
These improvement and modification are also fallen into the protection scope of the claim of this application.
Each embodiment is described in a progressive manner in specification, the highlights of each of the examples are with other realities
The difference of example is applied, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment
Speech, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is referring to method part illustration
?.
It should also be noted that, in the present specification, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment for including a series of elements not only includes that
A little elements, but also including other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Claims (10)
1. a kind of Feature Selection Algorithms based on injection molding model characterized by comprising
Obtain sample data corresponding with each characteristic variable of injection molding model；
Characteristic value corresponding with each characteristic variable is calculated according to each sample data；
Object feature value is chosen from each characteristic value with predefined rule, and the corresponding feature of the object feature value is become
Amount is used as target signature variable.
2. the Feature Selection Algorithms according to claim 1 based on injection molding model, which is characterized in that the basis is each
The sample data calculates characteristic value corresponding with each characteristic variable
By each sample data composition sample matrix and covariance matrix is calculated according to the sample matrix；
The characteristic value corresponding with each characteristic variable is calculated using the covariance matrix.
3. the Feature Selection Algorithms according to claim 2 based on injection molding model, which is characterized in that described by each institute
It states sample data composition sample matrix and covariance matrix is calculated according to the sample matrix and include:
Each sample data is formed into the sample matrix；
Using in the sample matrix with each corresponding each sample data of characteristic variable and the characteristic variable
Number calculates average characteristics numerical value；
Calculate the difference of each sample data Yu the average characteristics numerical value；
It is combined each difference to obtain the covariance matrix.
4. the Feature Selection Algorithms according to claim 1 based on injection molding model, which is characterized in that described with predetermined
Adopted rule chooses object feature value from each characteristic value, and using the corresponding characteristic variable of the object feature value as target
After characteristic variable, further includes:
Using the target signature variable as input variable；
Using the quality of production of injecting products and production efficiency as output variable；
Using the input variable and the output variable construct the injection molding model and to the injection molding model into
Row optimization；
The injection molding model after optimization is solved to obtain the optimal solution of each target signature variable.
5. the Feature Selection Algorithms according to claim 4 based on injection molding model, which is characterized in that described to utilize institute
It states input variable and the output variable constructs the injection molding model and optimizes to the injection molding model and includes:
The injection molding model is constructed using the input variable and the output variable；
Set a variety of neuron numbers of the hidden layer of the injection molding model；
Using test sample data corresponding with the input variable and verifying sample data, from a variety of neuron numbers
Choose the optimal neuron number of the injection molding model；
After obtaining optimization using the input variable, the output variable and the corresponding hidden layer of the optimal neuron number
The injection molding model.
6. the Feature Selection Algorithms according to claim 1 based on injection molding model, which is characterized in that described with predetermined
Adopted rule chooses object feature value from each characteristic value, and using the corresponding characteristic variable of the object feature value as target
Characteristic variable includes:
The characteristic value for being greater than targets threshold is chosen from each characteristic value, and will be greater than the characteristic value of the targets threshold
Target signature variable of the corresponding characteristic variable as the injection molding model.
7. the Feature Selection Algorithms based on injection molding model described in -6 any one according to claim 1, which is characterized in that
The target signature variable includes dwell pressure, melt temperature, cooling time, mold temperature and dwell time.
8. a kind of Feature Selection device based on injection molding model characterized by comprising
Sample data obtains module, for obtaining sample data corresponding with each characteristic variable of injection molding model；
Characteristic value calculating module, for calculating characteristic value corresponding with each characteristic variable according to each sample data；
Target signature specification of variables module, for choosing object feature value from each characteristic value with predefined rule, and will
The corresponding characteristic variable of the object feature value is as target signature variable.
9. a kind of Feature Selection device based on injection molding model characterized by comprising
Memory, for storing computer program；
Processor, for executing the computer program stored in the memory to realize as described in any one of claim 1 to 7
The Feature Selection Algorithms based on injection molding model the step of.
10. a kind of computer readable storage medium, computer program, feature are stored on the computer readable storage medium
It is, the computer program is executed by processor as described in any one of claim 1 to 7 based on injection mould to realize
The step of Feature Selection Algorithms of type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810796642.5A CN109101698B (en) | 2018-07-19 | 2018-07-19 | Feature selection method and device based on injection molding model and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810796642.5A CN109101698B (en) | 2018-07-19 | 2018-07-19 | Feature selection method and device based on injection molding model and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109101698A true CN109101698A (en) | 2018-12-28 |
| CN109101698B CN109101698B (en) | 2023-04-18 |
Family
ID=64846710
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810796642.5A Active CN109101698B (en) | 2018-07-19 | 2018-07-19 | Feature selection method and device based on injection molding model and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109101698B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112115579A (en) * | 2020-08-12 | 2020-12-22 | 江苏师范大学 | Multi-target optimization method for injection molding process parameters of glass fiber reinforced plastics |
| CN113103535A (en) * | 2021-03-17 | 2021-07-13 | 贵州大学 | GA-ELM-GA-based injection molding part mold parameter optimization method |
| US11403535B2 (en) * | 2018-12-21 | 2022-08-02 | Industrial Technology Research Institute | Model-based machine learning system |
| CN115619050A (en) * | 2022-12-19 | 2023-01-17 | 绵阳丰瑞尚雅木业有限公司 | Method and system for predicting processing quality of fireproof aluminum-wood veneer and storage medium |
| CN117532885A (en) * | 2024-01-10 | 2024-02-09 | 成都航空职业技术学院 | Intelligent auxiliary system, method and storage medium for 3D printing |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2008072047A (en) * | 2006-09-15 | 2008-03-27 | Omron Corp | Model preparing device, analyzing device for process abnormality, and those method and program |
| US20110066391A1 (en) * | 2009-09-11 | 2011-03-17 | University Of Cincinnati | Methods and systems for energy prognosis |
| US20120123583A1 (en) * | 2010-11-16 | 2012-05-17 | Mks Instruments, Inc. | Controlling a Discrete-Type Manufacturing Process with a Multivariate Model |
| CN105550744A (en) * | 2015-12-06 | 2016-05-04 | 北京工业大学 | Nerve network clustering method based on iteration |
| CN106597260A (en) * | 2016-12-29 | 2017-04-26 | 合肥工业大学 | Simulation circuit fault diagnosis method based on continuous wavelet analysis and ELM network |
-
2018
- 2018-07-19 CN CN201810796642.5A patent/CN109101698B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2008072047A (en) * | 2006-09-15 | 2008-03-27 | Omron Corp | Model preparing device, analyzing device for process abnormality, and those method and program |
| US20110066391A1 (en) * | 2009-09-11 | 2011-03-17 | University Of Cincinnati | Methods and systems for energy prognosis |
| US20120123583A1 (en) * | 2010-11-16 | 2012-05-17 | Mks Instruments, Inc. | Controlling a Discrete-Type Manufacturing Process with a Multivariate Model |
| CN105550744A (en) * | 2015-12-06 | 2016-05-04 | 北京工业大学 | Nerve network clustering method based on iteration |
| CN106597260A (en) * | 2016-12-29 | 2017-04-26 | 合肥工业大学 | Simulation circuit fault diagnosis method based on continuous wavelet analysis and ELM network |
Non-Patent Citations (1)
| Title |
|---|
| 王光华: ""PCA和GA-ELM在立体图像质量客观评价中的应用"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11403535B2 (en) * | 2018-12-21 | 2022-08-02 | Industrial Technology Research Institute | Model-based machine learning system |
| CN112115579A (en) * | 2020-08-12 | 2020-12-22 | 江苏师范大学 | Multi-target optimization method for injection molding process parameters of glass fiber reinforced plastics |
| CN113103535A (en) * | 2021-03-17 | 2021-07-13 | 贵州大学 | GA-ELM-GA-based injection molding part mold parameter optimization method |
| CN115619050A (en) * | 2022-12-19 | 2023-01-17 | 绵阳丰瑞尚雅木业有限公司 | Method and system for predicting processing quality of fireproof aluminum-wood veneer and storage medium |
| CN115619050B (en) * | 2022-12-19 | 2023-03-28 | 绵阳丰瑞尚雅木业有限公司 | Method and system for predicting processing quality of fireproof aluminum-wood veneer and storage medium |
| CN117532885A (en) * | 2024-01-10 | 2024-02-09 | 成都航空职业技术学院 | Intelligent auxiliary system, method and storage medium for 3D printing |
| CN117532885B (en) * | 2024-01-10 | 2024-05-07 | 成都航空职业技术学院 | Intelligent auxiliary system, method and storage medium for 3D printing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109101698B (en) | 2023-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109101698A (en) | A kind of Feature Selection Algorithms based on injection molding model, device and storage medium | |
| CN110640982A (en) | A multi-objective optimization method of injection molding process parameters for thin-walled injection molded parts | |
| CN112115579A (en) | Multi-target optimization method for injection molding process parameters of glass fiber reinforced plastics | |
| CN113722992B (en) | Multi-objective optimization method for injection molding process parameters of injection molding piece with insert | |
| Meiabadi et al. | Optimization of plastic injection molding process by combination of artificial neural network and genetic algorithm | |
| CN109800461A (en) | Crucial gabarit parameter optimization method and device for tire construction light-weight design | |
| CN110674598A (en) | An optimization method of injection molding process based on support vector machine and particle swarm optimization | |
| CN112069610B (en) | Injection molding process parameter optimization method for transparent complex multi-cavity plastic part | |
| AlKaabneh et al. | A combined analytical hierarchical process (AHP) and Taguchi experimental design (TED) for plastic injection molding process settings | |
| CN107972243A (en) | Shooting Technique optimization method and Shooting Technique based on SOM neutral nets | |
| Yarlagadda et al. | Development of a hybrid neural network system for prediction of process parameters in injection moulding | |
| CN113103535A (en) | GA-ELM-GA-based injection molding part mold parameter optimization method | |
| US20070179917A1 (en) | Intelligent design optimization method and system | |
| Sedighi et al. | Optimisation of gate location based on weld line in plastic injection moulding using computer-aided engineering, artificial neural network, and genetic algorithm | |
| Panda et al. | Integrated AHP and fuzzy TOPSIS approach for the selection of a rapid prototyping process under multi-criteria perspective | |
| Tayalati et al. | Application of supervised machine learning methods in injection molding process for initial parameters setting: prediction of the cooling time parameter | |
| Yang et al. | Research on Optimization of Injection Molding Process Parameters of Automobile Plastic Front‐End Frame | |
| Chen et al. | Self-adaptive weights based on balanced residual decay rate for physics-informed neural networks and deep operator networks | |
| Márquez et al. | An efficient adaptive fuzzy inference system for complex and high dimensional regression problems in linguistic fuzzy modelling | |
| Ali et al. | Design quality and robustness with neural networks | |
| EP3850593B1 (en) | Method for improving prediction relating to the production of a polymeric product | |
| CN118349656A (en) | Heat supply customer service building method and system based on generation type large model | |
| CN115167150B (en) | Batch process two-dimensional off-orbit strategy staggered Q learning optimal tracking control method with unknown system dynamics | |
| CN116822360A (en) | Power system frequency trajectory prediction method, device, medium and equipment | |
| CN113393023B (en) | Mold quality evaluation method, apparatus, device and storage medium |
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 |