CN109543239A - LTCC shrinkage pre-judging method based on nerual network technique - Google Patents
LTCC shrinkage pre-judging method based on nerual network technique Download PDFInfo
- Publication number
- CN109543239A CN109543239A CN201811270751.XA CN201811270751A CN109543239A CN 109543239 A CN109543239 A CN 109543239A CN 201811270751 A CN201811270751 A CN 201811270751A CN 109543239 A CN109543239 A CN 109543239A
- Authority
- CN
- China
- Prior art keywords
- shrinking percentage
- slurry
- ltcc
- coverage rate
- ceramic whiteware
- 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
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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N22/00—Investigating or analysing materials by the use of microwaves or radio waves, i.e. electromagnetic waves with a wavelength of one millimetre or more
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Electromagnetism (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Investigating Or Analyzing Materials By The Use Of Electric Means (AREA)
Abstract
The present invention relates to a kind of LTCC shrinkage pre-judging method based on nerual network technique determines each influence factor of shrinking percentage by orthogonal test；Using each influence factor as structural parameters, influence factor list is established to every layer of LTCC；Neural network is set, and according to production line historical data, training neural network obtains shrinking percentage prediction model；Utilize the shrinking percentage of shrinking percentage prediction model prediction LTCC product.The present invention is aiming at the problem that different LTCC products can not accurately know its shrinking percentage before operation, from LTCC product design and technique, it confirmed the factor of influence shrinking percentage, and the mathematical model between influence factor and shrinking percentage is established using nerual network technique, realize accurately estimating for LTCC contractibility.
Description
Technical field
The present invention relates to a kind of LTCC shrinkage pre-judging method based on nerual network technique, belongs to low-temperature co-fired ceramics skill
Art field.
Background technique
Based on the microwave circuit modules of LTCC Technology (LTCC), such as LTCC filter and power module,
Successful utilization, since LTCC can be realized the three dimensional design of microwave circuit, is recognized into the positive sample Satellite Product in China
To be microwave circuit miniaturization and then realizing one of the key technology of satellite miniaturization.
LTCC common process process is as shown in Figure 1, wherein step 9 is referred to as " low temperature co-fired " or is referred to as " altogether
Burn ", LTCC green can be shunk during cofiring, realize ceramics by soft to hard transformation, thus formed densification it is hard
Ceramic multilayer circuit, the degree that green is shunk during this are indicated by " shrinking percentage ".Therefore, the art circuits in process 1-8
Dimension of picture then needs to do enhanced processing according to the shrinking percentage, just can guarantee that final product is contracted in cofiring link in this way
Required full-size(d).
However, the shrinking percentage of LTCC is influenced by many factors, raw material supplier usually only provides a fixed ginseng
Examine shrinking percentage value, therefore the universal practice of industry is outside Current Domestic: technologist refers to the technical documentation of material supplier,
The amplification that the circuit diagram of designer is done to a proper proportion further according to experience, is then processed, finally in cofiring mistake
Green is retracted to size desired by designer in journey.But for some complex products, it is sometimes desirable to a product is first produced,
The case where measuring shrinking percentage, correcting amplification factor, then produce again.Since satellite uses LTCC product to use gold as raw material,
Secondary plate of throwing will cause the waste of time and resource, be unfavorable for controlling cost and progress.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of, and the LTCC based on nerual network technique is received
Shrinkage pre-judging method, realization accurate anticipation is made to the shrinking percentage of every money LTCC product before manufacture, thus improve yield rate and
Circuit performance shortens the process-cycle.
The object of the invention is achieved by following technical solution:
A kind of LTCC shrinkage pre-judging method based on nerual network technique is provided, is included the following steps:
(1) by orthogonal test, each influence factor of shrinking percentage is determined；
(2) using each influence factor as structural parameters, influence factor list is established to every layer of LTCC；
(3) neural network is set, and according to production line historical data, training neural network obtains shrinking percentage prediction model；
(4) shrinking percentage of shrinking percentage prediction model prediction LTCC product is utilized.
Preferably, steps are as follows for acquisition shrinking percentage prediction model:
3.1 building three-layer artificial neural networks, are classified as three parts: first layer is that input neuron is shunk for receiving
Each influence factor of rate, the second layer are the non-linear relation that hidden neuron is used to handle input with output, and third layer is output
Neuron is for exporting LTCC shrinkage；
3.2 utilize production line historical data, adjust the weighted value in neural network, make the output and history shrinking percentage of model
Data are consistent；
3.3, using the production line historical data being not used by, verify the accuracy of neural network model shrinking percentage；Work as model
Output and history shrinking percentage data between error when meeting the requirements, complete modeling, otherwise return to 3.1, adjust implicit nerve
First number.
Preferably, each influence factor of shrinking percentage includes: common metal slurry coverage rate, resistance slurry coverage rate, Kong Mi
Degree and cavity density.
Preferably, commonly using slurry and resistance slurry coverage rate includes that slurry C N30-080 coverage rate, slurry C N30-025 cover
Lid rate, slurry C N36-020 coverage rate, slurry DL10-088 coverage rate and slurry FX 87-101B coverage rate.
Preferably, orthogonal test includes cooperating corresponding silk screen to different slurries, and different coverage rates is arranged；By resistance
Cooperation is screen printed onto ceramic whiteware on piece；Different punch densities is realized in ceramic whiteware on piece, and uses bushing filling perforation；In ceramic whiteware on piece
Realize various sizes of cavity；Orthogonal test is carried out, obtains the value of shrinking percentage respectively, obtains production line historical data.
Preferably, for Ferro material, the step of orthogonal test, is as follows:
1.1 8 inches of ceramic whiteware pieces 24 of preparation are opened, and the bushing one of corresponding punching figure is opened, and print the silk screen L mono- of large-area graphs
, the silk screen M mono- of printing moderate area figure opens, and the silk screen S mono- of printing small area figure opens；
1.2 take 3 ceramic whiteware piece punchings, and carry out filling perforation using slurry C N30-078 and bushing, as group 1；
3 ceramic whiteware pieces are taken, using silk screen L printing slurry CN30-080, as group 2；
3 ceramic whiteware pieces are taken, using silk screen L printing slurry CN30-025, as group 3；
3 ceramic whiteware pieces are taken, using silk screen M printing slurry CN36-020, as group 4；
3 ceramic whiteware pieces are taken, using silk screen S printing slurry DL10-088, as group 5；
3 ceramic whiteware pieces are taken, using silk screen S printing slurry FX 87-101B, as group 6；
Each of 24 ceramic whiteware pieces are equally divided into four pieces by 1.3；
Be divided into respectively 12 1/4 piece of 1.4 groups 1~6 stacks respectively；
1.5 12 1/4 pieces for taking 3 ceramic whiteware pieces to be divided into stack, as group 7；
1.6 12 1/4 pieces for taking 3 ceramic whiteware pieces to be divided into stack, and with cavity is laser machined, as group 8；
1.7 by one point every piece in group 1-8 be four；
The product size of 1.8 measurement group 1-8, every group is fired a kind of product, measures shrinking percentage.
Preferably, the input signal of shrinking percentage prediction model is the variable for influencing shrinking percentage, is denoted as I=[I1,I2,
I3,……I147]T, the input signal of neural network model is that output data is X-direction shrinking percentage and the direction Y shrinking percentage, is denoted as O
=[X, Y]T；The non-linear formula of input and output are as follows:
Wherein p indicates the number of hidden layer neuron,Indicate m-th of input neuron and i-th hidden neuron it
Between weight parameter,It is to refer to the weight ginseng between i-th of neuron in the 1st output nerve network and hidden layer
Number,It is to refer to the weight parameter between i-th of neuron in the 2nd output nerve network and hidden layer,With
Respectively indicate the biasing weighted value of the 1st output neuron and the 2nd output neuron, sigmoid function formula be σ (γ)=
1/(1+e-γ)。
A kind of LTCC shrinkage pre-judging method based on nerual network technique is provided simultaneously, is included the following steps:
(1) by orthogonal test, common metal slurry coverage rate, resistance slurry coverage rate, hole density needed for producing are determined
Influence with cavity density to shrinking percentage；
(2) shrinking percentage mould is established according to every layer of each slurry coverage rate, resistance coverage rate, hole density and cavity density respectively
Type；
(3) using every layer of shrinking percentage model, the shrinking percentage of LTCC product is predicted.
Preferably, common metal slurry and resistance slurry needed for producing include slurry C N30-080, slurry C N30-025,
Slurry C N36-020 and slurry DL10-088.
Preferably, orthogonal test includes cooperating corresponding silk screen to different slurries, and different coverage rates is arranged；By resistance
Cooperation is screen printed onto ceramic whiteware on piece；Different punch densities is realized in ceramic whiteware on piece, and uses bushing filling perforation；In ceramic whiteware on piece
Realize various sizes of cavity；Orthogonal test is carried out, obtains the value of shrinking percentage respectively, obtains each slurry coverage rate, resistance covering
The data of the influence of rate, hole density and cavity density to shrinking percentage.
The invention has the following advantages over the prior art:
(1) present invention produces aiming at the problem that different LTCC products can not accurately know its shrinking percentage before operation from LTCC
Product design and processes set out, it is thus identified that influence the factor of shrinking percentage, and establish influence factor and receipts using nerual network technique
Mathematical model between shrinkage realizes accurately estimating for LTCC contractibility.
(2) present invention obtains mature neural network model, realizes quickly using historical data training neural network
Accurately prediction shrinking percentage, it is out of control to avoid shrinking percentage.
(3) present invention guarantees the success rate once gone into operation for be pilot, the product of full gold system.
Detailed description of the invention
Fig. 1 is LTCC conventional flowsheet schematic diagram of the present invention；
Fig. 2 influences contrast schematic diagram to shrinking percentage in X-direction for different factors；
Fig. 3 is Ferro base product data statistics exemplary diagram；
Fig. 4 is neural network confirmatory experiment training process schematic diagram；
Fig. 5 is neural network confirmatory experiment training result precision schematic diagram；
Fig. 6 is the structure chart of neural network model.
Specific embodiment
Thinking of the invention are as follows: 1. confirmations influence the factor of shrinking percentage；2. being established according to the data that production line history generates
Mathematical model corresponding relationship between influence factor and shrinking percentage.The present invention will be with the Ferro material in LTCC material system
For, illustrate concrete implementation approach.
1. confirmation influences the factor of shrinking percentage, grout material, hole density, cavity density are generally included.
As shown in Figure 1, be related to being all located at the operation of green step 9 it is low temperature co-fired before, including green off-chip, green
Punching, through-hole filling, conductor printing, it is secondary begin to speak, green stack, green pressing and green cutting.The technique ginseng of these operations
Number be it is cured, primary difference is that every money product type of stock used, the density beginning to speak and punch.Therefore, it carries out first
The single experiment of these factors, the slurry used in view of Ferro system is numerous, so the present invention only considers that production line is common
Slurry.For the scene of closing to reality application, single experiment has used three kinds of different silk screens, and top is dispersed with 10mm* respectively
The square of 10mm, 5mm*5mm, 1mm*1mm, and it is respectively designated as L, M, S type silk screen.
Cooperate different slurries different silk screens, resistance cooperation is screen printed onto ceramic whiteware on piece, realizes in ceramic whiteware on piece
Different punch densities, and filled a vacancy using bushing；Various sizes of cavity is realized in ceramic whiteware on piece, carries out orthogonal test respectively,
The value of shrinking percentage is obtained respectively.
The project specifically to be examined is as shown in table 1.
Table 1
Due to being also difference cofiring after cutting product in LTCC actual production process, the feelings of full page cofiring are seldom used
Condition, so this experiment is printed using full page, one point is four, then the mode of lamination, the detailed process of orthogonal test are as follows:
Table 2
After cofiring, the present invention counts mini-test result, as shown in table 3:
Table 3
By taking X-direction as an example, the data of the mini-test of disparity items have been counted, and have listed histogram (Fig. 2) with explanation
Influence of the disparity items to shrinking percentage.
The shrinking percentage nominal value that Ferro producer provides are as follows: 84%, the data that dialogue tile of the present invention is tested are
84.68%, nominal value is substantially conformed to, i.e. the shrinking percentage reference line of ceramic whiteware piece shown in dotted line in Fig. 2, it can be seen that " strong
It is maximum that group " slurry C N30-080, CN36-020 and resistance FX 87-101B deviate reference line, and is all to hinder to shrink, this also with
Production line being experienced as in use is the same, experience have shown that CN30-080, CN36-020 and resistance slurry all cannot be big
Area uses, and is very easy to warpage.Meanwhile for " tenderness send " slurry C N30-025, CN30-078 for, it is basic to melt with white tiles
That closes is fine.Welding resistance DL10-088 does not substantially influence tile, and cavity is the factor that only one is conducive to tile contraction.
Accordingly, the present invention combs and confirmed to influence the principal element of shrinking percentage: slurry C N30-080 coverage rate, slurry
CN30-025 coverage rate, slurry C N36-020 coverage rate, welding resistance slurry (DL10-088) coverage rate, resistance slurry coverage rate, hole
Density and cavity density, totally 7 factors.Based on this, every money LTCC product has its distinctive technological parameter attribute.
2. the mathematical model between influence factor and shrinking percentage is established, according to production line historical data, training neural network.
In recent years, nerual network technique is one of the hot spot studied both at home and abroad.Neural network has self-learning function, association
Storage function and high speed find the advantages of optimization solution, be widely used in image recognition and processing, semiconductor devices modeling and
Future anticipation etc..Nerual network technique is used in view of its excellent data-handling capacity and modeling flexibility, the present invention
Establish the mathematical model between influence factor and shrinking percentage.
Since LTCC is multilayered structure, every layer of circuit is relatively independent, basic for each layer of ceramic chips process experienced
All it is type metal slurry, welding resistance, resistance, punches, begin to speak, and begins to speak so that each layer, which is likely to, becomes surface layer, institute
Be designed with identical input data format to each layer with the present invention: slurry C N30-080 coverage rate, slurry C N30-025 are covered
Rate, slurry C N36-020 coverage rate, welding resistance slurry coverage rate, resistance slurry coverage rate, hole density and cavity density, totally 7 changes
Amount.
LTCC conventional products are both less than equal to 21 layers at present, in order to facilitate technologist's statistical data, it is assumed that in data volume
21 layers are limited to, i.e. total input data number of data model is 21*7=147 data, and product of the number of plies less than 21 layers is in blank
Layer is corresponding to mend 0, and output data is X-direction shrinking percentage and the direction Y shrinking percentage, totally 2 data.Input data statistical form such as Fig. 3
It is shown.
The input signal of neural network model is the variable for influencing shrinking percentage, is denoted as I=[I1,I2,I3,……I147]T, mind
Input signal through network model is that output data is X-direction shrinking percentage and Y-direction shrinking percentage, is denoted as O=[X, Y]T.
The structure chart of neural network model such as Fig. 6, the non-linear formula of input and output are as follows:
Wherein p indicates the number of hidden layer neuron.Indicate m-th of input neuron and i-th hidden neuron it
Between weight parameter,It is to refer to the weight parameter between i-th of neuron in the 1st output nerve network and hidden layer.It is to refer to the weight parameter between i-th of neuron in the 2nd output nerve network and hidden layer.WithTable respectively
Show the biasing weighted value of the 1st output neuron and the 2nd output neuron.Sigmoid function formula is σ (γ)=1/ (1+
e-γ)。
In Fig. 3, a kind of each product of behavior (corresponding technique number), each product has 21 layers, each layer of packet
The data contained have: CN30-080 coverage rate, CN30-025 coverage rate, CN36-020 coverage rate, welding resistance coverage rate, resistance covering
Rate, hole density and cavity density.After this data structure has been determined, the present invention establishes neural network model.Neural network
Model is all fine with the compatibility of each platform by written in Java, in order to facilitate technologist's use, by secondary encapsulation at
Matlab program, to facilitate technologist's modification wherein parameter.
Program is very succinct, in total includes 3 main files:
Data data: the training data and test data that neural network uses are contained.
Main function: neural network parameter, including data structure, neuronal quantity, training data wherein can be set
Amount, amount of test data etc..
Traing file: for running neural metwork training and test program.
By the study to production line historical data, neural network can be used to predict that the LTCC of new operation product is received
Shrinkage, the present invention test neural network model using technique number L17036 product.The actual measurement of L17036 product is shunk
Rate is 84.08% and 84.13%.The data of L17036 never occur in the training process as test data, also never
It is entered in trained sample database in any form, to guarantee the correctness of verifying.
Neural network is set first:
Input data columns I=147；
Output data number O=3；
Hidden neuron number H=15；
Set T=31；
Test data number E=1；
Training；
Training data and test data (L17036) are placed in data file, totally 32 row data, wherein preceding 31 behavior
Training data, last line are test data L17036, namely the T=31 and E=1 of setting the inside.
Program is run, neural network has reached required precision by 10 iteration, neural metwork training and anticipation are completed,
Training precision reaches 10-6, as shown in Figure 4,5:
For test data, the shrinking percentage of estimating provided is 84.102% and 84.1822%, and the present invention is to output result
It is compared with measured result:
| X | Y | |
| Neural network anticipation | 84.1020 | 84.1822 |
| Actual measured value | 84.08 | 84.13 |
| Error | 0.022% | 0.1522% |
As can be seen that neural network, in X, Y-direction has been presented for very good anticipation, it is better than ± 2% error water
It is flat, reach given requirements, realizes the function of accurately estimating its shrinking percentage according to every money product.It is gone into operation using neural network to new
LTCC product carries out shrinking percentage and estimates, and estimates the effect that efficiency and precision are all satisfied with.
In another embodiment, the number between influence factor and shrinking percentage can be obtained by the way of fitting of a polynomial
Model is learned, estimates shrinking percentage using mathematics.
The present invention fundamentally reverses technologist's situation out of control to shrinking percentage anticipation result, so that every kind of product has
Relative to accurate LTCC shrinkage data, have substantive progress to LTCC product quality.
The above, optimal specific embodiment only of the invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.
The content that description in the present invention is not described in detail belongs to the well-known technique of professional and technical personnel in the field.
Claims (10)
1. a kind of LTCC shrinkage pre-judging method based on nerual network technique, which comprises the steps of:
(1) by orthogonal test, each influence factor of shrinking percentage is determined；
(2) using each influence factor as structural parameters, influence factor list is established to every layer of LTCC；
(3) neural network is set, and according to production line historical data, training neural network obtains shrinking percentage prediction model；
(4) shrinking percentage of shrinking percentage prediction model prediction LTCC product is utilized.
2. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 1, which is characterized in that received
Steps are as follows for shrinkage prediction model:
3.1 building three-layer artificial neural networks, are classified as three parts: first layer is input neuron for receiving shrinking percentage
Each influence factor, the second layer are the non-linear relation that hidden neuron is used to handle input with output, and third layer is output nerve
Member is for exporting LTCC shrinkage；
3.2 utilize production line historical data, adjust the weighted value in neural network, make the output and history shrinking percentage data of model
Unanimously；
3.3, using the production line historical data being not used by, verify the accuracy of neural network model shrinking percentage；It is defeated when model
When the error between history shrinking percentage data is met the requirements out, modeling is completed, otherwise returns to 3.1, adjustment hidden neuron
Number.
3. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 1, which is characterized in that shrinking percentage
Each influence factor include: common metal slurry coverage rate, resistance slurry coverage rate, hole density and cavity density.
4. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 3, which is characterized in that common slurry
Material and resistance slurry coverage rate include slurry C N30-080 coverage rate, slurry C N30-025 coverage rate, slurry C N36-020 covering
Rate, slurry DL10-088 coverage rate and slurry FX87-101B coverage rate.
5. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 1, which is characterized in that orthogonal examination
It tests including cooperating corresponding silk screen to different slurries, different coverage rates is set；Resistance cooperation is screen printed onto ceramic whiteware piece
On；Different punch densities is realized in ceramic whiteware on piece, and uses bushing filling perforation；Various sizes of cavity is realized in ceramic whiteware on piece；
Orthogonal test is carried out, obtains the value of shrinking percentage respectively, obtains production line historical data.
6. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 1, which is characterized in that for
The step of Ferro material, orthogonal test, is as follows:
1.1 8 inches of ceramic whiteware pieces 24 of preparation are opened, and the bushing one of corresponding punching figure is opened, and the silk screen L mono- for printing large-area graphs opens, print
The silk screen M mono- of brush moderate area figure opens, and the silk screen S mono- of printing small area figure opens；
1.2 take 3 ceramic whiteware piece punchings, and carry out filling perforation using slurry C N30-078 and bushing, as group 1；
3 ceramic whiteware pieces are taken, using silk screen L printing slurry CN30-080, as group 2；
3 ceramic whiteware pieces are taken, using silk screen L printing slurry CN30-025, as group 3；
3 ceramic whiteware pieces are taken, using silk screen M printing slurry CN36-020, as group 4；
3 ceramic whiteware pieces are taken, using silk screen S printing slurry DL10-088, as group 5；
3 ceramic whiteware pieces are taken, using silk screen S printing slurry FX 87-101B, as group 6；
Each of 24 ceramic whiteware pieces are equally divided into four pieces by 1.3；
Be divided into respectively 12 1/4 piece of 1.4 groups 1~6 stacks respectively；
1.5 12 1/4 pieces for taking 3 ceramic whiteware pieces to be divided into stack, as group 7；
1.6 12 1/4 pieces for taking 3 ceramic whiteware pieces to be divided into stack, and with cavity is laser machined, as group 8；
1.7 by one point every piece in group 1-8 be four；
The product size of 1.8 measurement group 1-8, every group is fired a kind of product, measures shrinking percentage.
7. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 1, which is characterized in that shrinking percentage
The input signal of prediction model is the variable for influencing shrinking percentage, is denoted as
I=[I1,I2,I3,.....I147]T, the input signal of neural network model is that output data is X-direction shrinking percentage and the side Y
To shrinking percentage, it is denoted as O=[X, Y]T；The non-linear formula of input and output are as follows:
Wherein p indicates the number of hidden layer neuron,It indicates between m-th of input neuron and i-th of hidden neuron
Weight parameter,It is to refer to the weight parameter between i-th of neuron in the 1st output nerve network and hidden layer,
It is to refer to the weight parameter between i-th of neuron in the 2nd output nerve network and hidden layer,WithTable respectively
Show that the biasing weighted value of the 1st output neuron and the 2nd output neuron, sigmoid function formula are σ (γ)=1/ (1+
e-γ)。
8. a kind of LTCC shrinkage pre-judging method based on nerual network technique, which comprises the steps of:
(1) by orthogonal test, common metal slurry coverage rate, resistance slurry coverage rate, hole density and chamber needed for producing are determined
Influence of the volume density to shrinking percentage；
(2) shrinking percentage model is established according to every layer of each slurry coverage rate, resistance coverage rate, hole density and cavity density respectively；
(3) using every layer of shrinking percentage model, the shrinking percentage of LTCC product is predicted.
9. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 8, which is characterized in that production institute
It needs common metal slurry and resistance slurry includes slurry C N30-080, slurry C N30-025, slurry C N36-020 and slurry DL10-
088。
10. the LTCC shrinkage pre-judging method based on nerual network technique according to claim 8, which is characterized in that orthogonal
Test includes cooperating corresponding silk screen to different slurries, and different coverage rates is arranged；Resistance cooperation is screen printed onto ceramic whiteware
On piece；Different punch densities is realized in ceramic whiteware on piece, and uses bushing filling perforation；Various sizes of chamber is realized in ceramic whiteware on piece
Body；Orthogonal test is carried out, the value of shrinking percentage is obtained respectively, it is close to obtain each slurry coverage rate, resistance coverage rate, hole density and cavity
Spend the data of the influence to shrinking percentage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811270751.XA CN109543239B (en) | 2018-10-29 | 2018-10-29 | LTCC shrinkage pre-judging method based on neural network technology |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811270751.XA CN109543239B (en) | 2018-10-29 | 2018-10-29 | LTCC shrinkage pre-judging method based on neural network technology |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109543239A true CN109543239A (en) | 2019-03-29 |
| CN109543239B CN109543239B (en) | 2023-06-09 |
Family
ID=65845461
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811270751.XA Active CN109543239B (en) | 2018-10-29 | 2018-10-29 | LTCC shrinkage pre-judging method based on neural network technology |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109543239B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111251474A (en) * | 2020-02-21 | 2020-06-09 | 山东理工大学 | Ceramic laser turning composite plastic processing method based on acoustic emission signal characteristic identification and automatic matching of processing parameters |
| CN111967211A (en) * | 2020-07-15 | 2020-11-20 | 中国电子科技集团公司第二十九研究所 | LTCC manufacturing process flow generation system and method |
| CN112140413A (en) * | 2020-09-02 | 2020-12-29 | 金发科技股份有限公司 | Method and system for predicting die sinking shrinkage rate of plastic part |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH09223871A (en) * | 1996-02-16 | 1997-08-26 | Mitsubishi Electric Corp | Manufacture of low-temperature burned multilayered ceramic substrate |
| US20080222067A1 (en) * | 2007-03-07 | 2008-09-11 | National Taiwan University Of Science And Technology | Prediction method of near field photolithography line fabrication using by the combination of taguchi method and neural network |
| CN101908083A (en) * | 2010-05-13 | 2010-12-08 | 上海市特种设备监督检验技术研究院 | Artificial neural network-based Q345 welding joint mechanical property prediction method |
| CN105631558A (en) * | 2016-03-22 | 2016-06-01 | 国家电网公司 | BP neural network photovoltaic power generation system power prediction method based on similar day |
-
2018
- 2018-10-29 CN CN201811270751.XA patent/CN109543239B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH09223871A (en) * | 1996-02-16 | 1997-08-26 | Mitsubishi Electric Corp | Manufacture of low-temperature burned multilayered ceramic substrate |
| US20080222067A1 (en) * | 2007-03-07 | 2008-09-11 | National Taiwan University Of Science And Technology | Prediction method of near field photolithography line fabrication using by the combination of taguchi method and neural network |
| CN101908083A (en) * | 2010-05-13 | 2010-12-08 | 上海市特种设备监督检验技术研究院 | Artificial neural network-based Q345 welding joint mechanical property prediction method |
| CN105631558A (en) * | 2016-03-22 | 2016-06-01 | 国家电网公司 | BP neural network photovoltaic power generation system power prediction method based on similar day |
Non-Patent Citations (1)
| Title |
|---|
| 陆广华等: "基于BP神经网络的注塑成型收缩率预测", 《塑料》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111251474A (en) * | 2020-02-21 | 2020-06-09 | 山东理工大学 | Ceramic laser turning composite plastic processing method based on acoustic emission signal characteristic identification and automatic matching of processing parameters |
| CN111967211A (en) * | 2020-07-15 | 2020-11-20 | 中国电子科技集团公司第二十九研究所 | LTCC manufacturing process flow generation system and method |
| CN111967211B (en) * | 2020-07-15 | 2023-03-14 | 中国电子科技集团公司第二十九研究所 | LTCC manufacturing process flow generation system and method |
| CN112140413A (en) * | 2020-09-02 | 2020-12-29 | 金发科技股份有限公司 | Method and system for predicting die sinking shrinkage rate of plastic part |
| CN112140413B (en) * | 2020-09-02 | 2022-04-08 | 金发科技股份有限公司 | Method and system for predicting die sinking shrinkage rate of plastic part |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109543239B (en) | 2023-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109543239A (en) | LTCC shrinkage pre-judging method based on nerual network technique | |
| CN103747639B (en) | A kind of high-rise board fabrication method | |
| CN106572600B (en) | A kind of linear harmomegathus control method of PCB | |
| CN104217063B (en) | Emulation design method for steel rail rolling pass | |
| CN107803987A (en) | Adaptive layered processing method, system and increasing material manufacturing equipment for increasing material manufacturing | |
| CN110245380A (en) | Soft instrument training and sample compensation process | |
| CN105246258A (en) | Circuit board expanding-shrinking proportion control method and system | |
| CN108415393A (en) | A kind of GaAs product quality consistency control method and system | |
| CN101478069A (en) | Microwave filter assistant debugging method based on nuclear machine learning | |
| CN108304963B (en) | Expansion and shrinkage prediction method for multilayer circuit board | |
| CN109586301B (en) | Intelligent distribution network local voltage control method based on synchronous phasor measurement equivalence | |
| CN108446616A (en) | Method for extracting roads based on full convolutional neural networks integrated study | |
| CN108541141A (en) | A kind of subregion alignment method of conducting aperture layer and circuitous pattern layer | |
| CN104581748A (en) | A method and device for identifying a scene in a wireless communication network | |
| CN113778513A (en) | Automatic generation method, device and equipment of PCB drilling program and storage medium | |
| CN104536396A (en) | Soft measurement modeling method used in cement raw material decomposing process in decomposing furnace | |
| CN108509719A (en) | A kind of ironcasting based on BP neural network breaks core prediction technique and system | |
| CN117351005B (en) | Method and system for detecting coating defects of carbon-coated foil | |
| CN112541634B (en) | Method and device for predicting top-layer oil temperature and discriminating false alarm and storage medium | |
| CN109388861A (en) | A kind of exponential unit service life estimation of distribution parameters method | |
| CN107622476A (en) | Image Super-resolution processing method based on generative probabilistic model | |
| CN107110661B (en) | The method for manufacturing the components that there is the size similar to tolerance stack to require | |
| CN111106054A (en) | Method for prejudging wafer calibration value and storage medium | |
| CN108563565A (en) | Flight landing guides system reliability Quantitative Analysis Model method for building up | |
| CN113133226B (en) | Circuit board high-precision back drilling method and circuit board |
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 |