CN109472105A - Semiconductor product yield Upper bound analysis method - Google Patents
Semiconductor product yield Upper bound analysis method Download PDFInfo
- Publication number
- CN109472105A CN109472105A CN201811396608.5A CN201811396608A CN109472105A CN 109472105 A CN109472105 A CN 109472105A CN 201811396608 A CN201811396608 A CN 201811396608A CN 109472105 A CN109472105 A CN 109472105A
- Authority
- CN
- China
- Prior art keywords
- product yield
- analysis method
- semiconductor product
- parameter
- upper bound
- 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]
- G06F30/30—Circuit design
- G06F30/39—Circuit design at the physical level
- G06F30/398—Design verification or optimisation, e.g. using design rule check [DRC], layout versus schematics [LVS] or finite element methods [FEM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
Abstract
The invention discloses a kind of semiconductor product yield Upper bound analysis methods, including selecting product yield analysis process parameter, each technological parameter is established to the functional relation of semiconductor product yield, is selected with the strongest independent variable parameter of semiconductor product yield correlation as X1, remaining independent variable parameter XNWith X1Corresponding relationship fN(X1) indicate: F (X1,X2,…,XN)=[F (X1)+F(X2)+…+F(XN)]/N=[F (f1(X1))+F(f2(X1))+…+F(fN(X1))]/N=F (X1), multiple regression is executed, particle swarm algorithm model is established, calculates and obtains optimal value of the parameter and highest yield.The present invention, which can calculate, obtains the semiconductor product yield upper limit.
Description
Technical field
The present invention relates to semiconductor fields, more particularly to a kind of semiconductor product yield Upper bound analysis method.
Background technique
Yield (CP/SP/FT) is the important indicator of semiconductor product quality, during product design and volume production, yield meeting
By various aspects, such as the influence of inline/WAT multi-parameter.In order to possibly increase the window of product yield unexpectedly, it is fixed to need
Optimal conditions.Its simple correlation to yield of certain parameter check can be selected, by experience usually to determine practical volume production
Window.For example, rule of thumb device speed and SRAM failure have correlation, engineer can be according to its correlation adjuster
The window of part speed.But actual product yield may be affected by factors simultaneously, same factor may also be to different
Yield index has different influences, there is the relationship of trade off between different indexs or factor.This Multiple factors that are related to are answered
Miscellaneous analysis system, there is also bigger difficulty for calculating assessment artificial at present, can not fully assess and measure.Meanwhile specified
When product yield target, it is also necessary to which knowing improves space by the achievable yield of adjustment parameter, but there are no sides up to now
Method can estimate its yield upper limit is how many.
Summary of the invention
The technical problem to be solved in the present invention is to provide a kind of analysis methods that can calculate the semiconductor product yield upper limit.
In order to solve the above technical problems, the present invention provides a kind of semiconductor product yield Upper bound analysis method, including following
Step:
1) product yield analysis process parameter is selected；
2) establishing each technological parameter is Y=F (X to the functional relation of semiconductor product yieldN), Y indicates product yield,
XNIndicate n-th process parameter value；
3) using the most strong technological parameter of correlation as independent variable parameter X1, remaining independent variable parameter XNWith X1Corresponding relationship
Use fN(X1) indicate:
F(X1,X2,…,XN)=[F (X1)+F(X2)+…+F(XN)]/N=[F (f1(X1))+F(f2(X1))+…+F(fN
(X1))]/N=F (X1)；
4) multiple regression is executed；
5) particle swarm algorithm model is established；
6) it calculates and obtains optimal value of the parameter and highest yield.
It is further improved the semiconductor product yield Upper bound analysis method, implementation steps 1) when, using good with certain product
The highest technological parameter of rate dependence is as product yield analysis process parameter.
It is further improved the semiconductor product yield Upper bound analysis method, implementation steps 1) when, it is calculated using random forest
Method is selected with certain highest technological parameter of product yield correlation as product yield analysis process parameter.
Be further improved the semiconductor product yield Upper bound analysis method, by fitting algorithm in linear, multinomial, refer to
It selects to be fitted function of the optimal function as the technological parameter to semiconductor product yield in number, power function or exponential form
Relational expression.
It is further improved the semiconductor product yield Upper bound analysis method, implementation steps 5) when, establish particle swarm algorithm
Model uses following steps；
(a) population, including population size N, the position xi and speed vi of each particle and position limitation are initialized；
(b) the fitness value Fit [i] of each particle is calculated；
(c) to each particle, compared with its fitness value Fit [i] and individual extreme value Pi, if Fit [i] > Pi, is used
Fit [i] replaces Pi；
(d) to each particle, compared with its fitness value Fit [i] and global extremum Pg, if Fit [i] > Pg, is used
Fit [i] replaces Pg；
(e) according to the speed vi and position xi of following formula (1) (2) more new particle；
Wherein, W=0.75, C1=C2=1, r1/r2For the random number of each grey iterative generation；
If (f) meeting termination condition to exit, otherwise return step (b).
It is further improved the semiconductor product yield Upper bound analysis method, when executing step (f), termination condition is to reach
Maximum cycle.
It is further improved the semiconductor product yield Upper bound analysis method, using computer programming language by the analysis side
Method code.
It is further improved the semiconductor product yield Upper bound analysis method, chooses a certain product in certain period
Input of the WAT and CP data of wafer level as analysis method code, analysis method code export optimal WAT parameter
Value and highest yield.
It is further improved the semiconductor product yield Upper bound analysis method, using Python by the analysis method generation
Codeization.
Particle swarm algorithm (Particle Swarm Optimization, PSO) is one kind of swarm intelligence algorithm, basic to think
Think to be the predation for simulating flock of birds random search food, flock of birds is according to the exchange adjustment search road between experience and population
Diameter finds the most place of food by the cooperation and competition between individual in complex space.Compared to traditional algorithm, particle
Group's algorithm has the advantages that concept is concise, realizes convenient and fast convergence rate, non-linear, multiple peak problem is all had stronger
Ability of searching optimum.When being used for practical problem, the position of every bird is independent variable combination in particle swarm algorithm, is first generated just
Beginning population, i.e., the random initializtion a group particle in solution space, each particle are a feasible solution of optimization problem, and
One adaptive value (fitness value) is determined by objective function for it, Particles Moving direction and distance, particle are determined by speed
Current optimal particle will be pursued and moved, and through finally obtaining optimal solution by generation search.In every generation, particle will track two poles
Value, an optimal solution pbest found so far for particle itself, another is the optimal solution gbest that full population is found so far.
FAB is practical establishes particle swarm algorithm model by combining by the present invention, regard product yield CP/FT etc. as target letter
Number, the technological parameters such as inline/WAT are independent variable group, with random forests algorithm selection and certain product yield correlation highest
Technological parameter, each technological parameter is obtained to the functional relation of yield by fitting, particle swarm algorithm is recycled to acquire optimal work
Skill condition and yield upper limit value, it is particularly possible to optimal conditions solution be carried out to multifactor impact objective function, set for process conditions
Meter optimization and control provide reference, guarantee that product yield is stablized, while reducing human input.
Detailed description of the invention
Present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments:
Fig. 1 is the particle swarm algorithm model schematic diagram suitable for FAB practical problem.
Specific embodiment
The present invention provides one possible embodiments of particle swarm algorithm model established and be suitable for FAB practical problem, the population
Algorithm model is as follows:
Location/velocity: argument value/update step-length；
Position limitation: independent variable value range limits the space of particle search；
Inertia weight: for recording itself current speed, the memory of itself speed is then lost when being 0；
Studying factors: the power of weight coefficient (the understanding C1) and tracking total optimization value of Particle tracking itself history optimal value
Weight coefficient (social C2), generally takes [0~4]；
Using Product Process condition (inline/WAT parameter) and CP data as initial individuals/population, carried out using model
demo；
Export result: the CP (the yield upper limit) of optimal location and optimal location.
The present invention provides the feasible reality of semiconductor product yield Upper bound analysis method one for combining above-mentioned particle swarm algorithm model
Apply example, comprising the following steps:
1) it is selected with certain highest technological parameter of product yield correlation using random forests algorithm as product yield point
Analyse technological parameter；
2) each technological parameter is established to the functional relation of semiconductor product yield；
3) using the most strong technological parameter of correlation as independent variable parameter X1, remaining independent variable parameter XNWith X1Corresponding relationship
Use fN(X1) indicate: the optimal (R2 of fitting is selected in linear, multinomial, index, power function or exponential form by fitting algorithm
It is maximum) function as the technological parameter to the functional relation of semiconductor product yield.
F(X1,X2,…,XN)=[F (X1)+F(X2)+…+F(XN)]/N=[F (f1(X1))+F(f2(X1))+…+F(fN
(X1))]/N=F (X1)；
4) multiple regression is executed；
5) particle swarm algorithm model is established；
(a) population, including population size N, the position xi and speed vi of each particle and position limitation are initialized；
(b) the fitness value Fit [i] of each particle is calculated；
(c) to each particle, compared with its fitness value Fit [i] and individual extreme value Pi, if Fit [i] > Pi, is used
Fit [i] replaces Pi；
(d) to each particle, compared with its fitness value Fit [i] and global extremum Pg, if Fit [i] > Pg, is used
Fit [i] replaces Pg；
(e) according to the speed vi and position xi of following formula (1) (2) more new particle；
Wherein, W=0.75, C1=C2=1, r1/r2For the random number of each grey iterative generation；
If (f) reaching maximum cycle to exit, the maximum cycle range is 100 times -1000 times, this implementation
It is 500 times in example, otherwise return step (b).
6) it calculates and obtains optimal value of the parameter and highest yield；
7) utilize Python by above-mentioned steps code；
8) WAT the and CP data of a certain product wafer level in certain period are chosen as the defeated of the code
Enter, the code exports optimal WAT parameter value and highest yield.
Above by specific embodiment and embodiment, invention is explained in detail, but these are not composition pair
Limitation of the invention.Without departing from the principles of the present invention, those skilled in the art can also make many deformations and change
Into these also should be regarded as protection scope of the present invention.
Claims (9)
1. a kind of semiconductor product yield Upper bound analysis method, which comprises the following steps:
1) product yield analysis process parameter is selected；
2) establishing each technological parameter is Y=F (X to the functional relation of semiconductor product yieldN), Y indicates product yield, XNTable
Show n-th process parameter value；
3) using the most strong technological parameter of correlation as independent variable parameter X1, remaining independent variable parameter XNWith X1Corresponding relationship fN
(X1) indicate:
F(X1,X2,…,XN)=[F (X1)+F(X2)+…+F(XN)]/N=[F (f1(X1))+F(f2(X1))+…+F(fN(X1))]/
N=F (X1)；
4) multiple regression is executed；
5) particle swarm algorithm model is established；
6) it calculates and obtains optimal value of the parameter and highest yield.
2. semiconductor product yield Upper bound analysis method as described in claim 1, it is characterised in that: implementation steps 1) when, it uses
With certain highest technological parameter of product yield correlation as product yield analysis process parameter.
3. semiconductor product yield Upper bound analysis method as claimed in claim 2, it is characterised in that: implementation steps 1) when, it uses
Random forests algorithm is selected with certain highest technological parameter of product yield correlation as product yield analysis process parameter.
4. semiconductor product yield Upper bound analysis method as described in claim 1, it is characterised in that: online by fitting algorithm
Property, select to be fitted optimal function as the technological parameter to semiconductor product in multinomial, index, power function or exponential form
The functional relation of product yield.
5. semiconductor product yield Upper bound analysis method as described in claim 1, it is characterised in that: implementation steps 5) when, it establishes
Particle swarm algorithm model uses following steps；
(a) population, including population size N, the position xi and speed vi of each particle and position limitation are initialized；
(b) the fitness value Fit [i] of each particle is calculated；
(c) to each particle, compared with its fitness value Fit [i] and individual extreme value Pi, if Fit [i] > Pi, uses Fit
[i] replaces Pi；
(d) to each particle, compared with its fitness value Fit [i] and global extremum Pg, if Fit [i] > Pg, uses Fit
[i] replaces Pg；
(e) according to the speed vi and position xi of following formula (1) (2) more new particle；
Wherein, W=0.75, C1=C2=1, r1/r2For the random number of each grey iterative generation；
If (f) meeting termination condition to exit, otherwise return step (b).
6. semiconductor product yield Upper bound analysis method as claimed in claim 5, it is characterised in that: when executing step (f), terminate
Condition is to reach maximum cycle, and the maximum cycle range is 100 times -1000 times.
7. semiconductor product yield Upper bound analysis method as claimed in any one of claims 1 to 6, it is characterised in that: utilize calculating
Machine programming language is by the analysis method code.
8. semiconductor product yield Upper bound analysis method as claimed in claim 7, it is characterised in that: choose certain in certain period
Input of the WAT and CP data of one product wafer level as analysis method code, the output of analysis method code are optimal
WAT parameter value and highest yield.
9. semiconductor product yield Upper bound analysis method as claimed in claim 7, it is characterised in that: should using Python
Analysis method code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811396608.5A CN109472105B (en) | 2018-11-22 | 2018-11-22 | Semiconductor product yield upper limit analysis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811396608.5A CN109472105B (en) | 2018-11-22 | 2018-11-22 | Semiconductor product yield upper limit analysis method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109472105A true CN109472105A (en) | 2019-03-15 |
| CN109472105B CN109472105B (en) | 2023-09-19 |
Family
ID=65673059
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811396608.5A Active CN109472105B (en) | 2018-11-22 | 2018-11-22 | Semiconductor product yield upper limit analysis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109472105B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110516807A (en) * | 2019-08-29 | 2019-11-29 | 上海华力集成电路制造有限公司 | Semiconductor product yields extreme value calculation method and its extreme value computing system |
| CN110689067A (en) * | 2019-09-25 | 2020-01-14 | 上海华力集成电路制造有限公司 | Failure detection method, device and equipment for wafer and storage medium |
| CN112163799A (en) * | 2020-12-02 | 2021-01-01 | 晶芯成(北京)科技有限公司 | Yield analysis method and yield analysis system of semiconductor product |
| CN113168170A (en) * | 2019-05-22 | 2021-07-23 | 株式会社东芝 | Manufacturing condition output device, quality management system, and program |
| CN114456862A (en) * | 2020-11-04 | 2022-05-10 | 中国石油化工股份有限公司 | Optimization method for natural gas cryogenic denitrification pretreatment process parameters |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200638171A (en) * | 2005-04-26 | 2006-11-01 | Powerchip Semiconductor Corp | Real-time management systems and methods for manufacture management and yield rate analysis integration |
| CN104752259A (en) * | 2013-12-30 | 2015-07-01 | 中芯国际集成电路制造(上海)有限公司 | Sorting method for chips with inappropriate yield to technological level |
| US20160125759A1 (en) * | 2014-10-29 | 2016-05-05 | Transcontinental Industries, Inc. | Conditioned response toilet training systems and methods |
| CN106503459A (en) * | 2016-10-26 | 2017-03-15 | 南京信息工程大学 | A kind of improvement random forest temperature record method of quality control |
| CN108804850A (en) * | 2018-06-27 | 2018-11-13 | 大连理工大学 | A kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction |
-
2018
- 2018-11-22 CN CN201811396608.5A patent/CN109472105B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TW200638171A (en) * | 2005-04-26 | 2006-11-01 | Powerchip Semiconductor Corp | Real-time management systems and methods for manufacture management and yield rate analysis integration |
| CN104752259A (en) * | 2013-12-30 | 2015-07-01 | 中芯国际集成电路制造(上海)有限公司 | Sorting method for chips with inappropriate yield to technological level |
| US20160125759A1 (en) * | 2014-10-29 | 2016-05-05 | Transcontinental Industries, Inc. | Conditioned response toilet training systems and methods |
| CN106503459A (en) * | 2016-10-26 | 2017-03-15 | 南京信息工程大学 | A kind of improvement random forest temperature record method of quality control |
| CN108804850A (en) * | 2018-06-27 | 2018-11-13 | 大连理工大学 | A kind of aero-engine transition state accelerator critical performance parameters prediction technique based on Space Reconstruction |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113168170A (en) * | 2019-05-22 | 2021-07-23 | 株式会社东芝 | Manufacturing condition output device, quality management system, and program |
| CN110516807A (en) * | 2019-08-29 | 2019-11-29 | 上海华力集成电路制造有限公司 | Semiconductor product yields extreme value calculation method and its extreme value computing system |
| CN110689067A (en) * | 2019-09-25 | 2020-01-14 | 上海华力集成电路制造有限公司 | Failure detection method, device and equipment for wafer and storage medium |
| CN114456862A (en) * | 2020-11-04 | 2022-05-10 | 中国石油化工股份有限公司 | Optimization method for natural gas cryogenic denitrification pretreatment process parameters |
| CN114456862B (en) * | 2020-11-04 | 2022-10-21 | 中国石油化工股份有限公司 | Optimization method for natural gas cryogenic denitrification pretreatment process parameters |
| CN112163799A (en) * | 2020-12-02 | 2021-01-01 | 晶芯成(北京)科技有限公司 | Yield analysis method and yield analysis system of semiconductor product |
| CN112163799B (en) * | 2020-12-02 | 2021-03-02 | 晶芯成(北京)科技有限公司 | Yield analysis method and yield analysis system of semiconductor product |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109472105B (en) | 2023-09-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109472105A (en) | Semiconductor product yield Upper bound analysis method | |
| CN108846517B (en) | Integration method for predicating quantile probabilistic short-term power load | |
| CN107798199B (en) | Hydroelectric generating set parameter closed-loop identification method | |
| CN111371607B (en) | Network flow prediction method for optimizing LSTM based on decision-making graying algorithm | |
| CN110308649B (en) | PID parameter optimization method based on PSO-SOA fusion algorithm and applied to industrial process control | |
| CN110619393A (en) | Traffic simulation software parameter calibration method, system and medium based on learning algorithm | |
| CN103489039B (en) | There is the freeway traffic flow amount fusion forecasting method of online self-tuning optimization ability | |
| CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| CN105427241B (en) | Distortion correction method for large-view-field display equipment | |
| CN109163911A (en) | A kind of fault of engine fuel system diagnostic method based on improved bat algorithm optimization ELM | |
| CN108171379A (en) | A kind of electro-load forecast method | |
| CN105429134A (en) | Grid voltage stability prediction method based on power big data | |
| CN110895772A (en) | Electricity sales amount prediction method based on combination of grey correlation analysis and SA-PSO-Elman algorithm | |
| CN103106331B (en) | Based on the lithographic line width Intelligent Forecasting of dimensionality reduction and increment type extreme learning machine | |
| CN108460462A (en) | A kind of Interval neural networks learning method based on interval parameter optimization | |
| CN111222286A (en) | Parameter optimization method based on power transmission line state estimation | |
| CN112000001A (en) | PID parameter setting optimization method based on improved Bayesian model | |
| CN111008790A (en) | Hydropower station group power generation electric scheduling rule extraction method | |
| CN106296434A (en) | A kind of Grain Crop Yield Prediction method based on PSO LSSVM algorithm | |
| CN110490422A (en) | A kind of target fighting efficiency method for situation assessment based on game cloud model | |
| CN116151102A (en) | Intelligent determination method for space target ultra-short arc initial orbit | |
| CN110321995A (en) | A kind of difference Particle Swarm Mixed Algorithm based on Stochastic inertia weight | |
| CN110197250A (en) | A kind of power battery on-line parameter identification method of multifactor impact | |
| CN110689108A (en) | Nonlinear system state estimation method | |
| CN114186709A (en) | Energy prediction method for optimizing key parameters of gray model based on emperor butterfly algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |