CN101320258B - Two-stage virtual measurement method - Google Patents

Abstract

The present invention discloses a virtual measurement method by using dual-phase virtual measurement system by which a dual-phase virtual measurement value can be produced to take immediacy and accuracy into consideration. Wherein, as soon as the process parameter data of a workpiece are collected, the estimation step of the first phase immediately calculates the first-phase virtual measurement value (VMI) of the workpiece in order to satisfy the requirement on immediacy; the estimation step of the second phase recalculates the second-phase virtual measurement value (VMII) of all the workpieces in a cassette belonging to a workpiece sampled for measurement in order to satisfy the requirement on accuracy only after the actual measurement value (used for retraining or calibration) of the workpiece sampled for measurement is obtained. Moreover, the method can also produce the confidence indexes and the integral similarity indexes of the first-phase virtual measurement value and the second-phase virtual measurement value.

Description

Utilize the virtual measurement method of two-stage virtual measuring system

Technical field

The present invention relates to a kind of virtual measurement method, relate in particular to a kind of two-stage virtual measurement method of taking into account immediacy and accuracy.

Background technology

Batch to batch (Run-to-Run; R2R) advanced technologies control (Advanced ProcessControl; APC) be widely used in semiconductor and the TFT-LCD factory to improve the production capacity of technology.Define as SEMI E133 specification, the control of R2R is a kind of technology of revising formulation parameter; Or in batch between select controlled variable, to improve treatment efficiency.Wherein one batch (Run) can be a batch (Batch), parcel (Lot) or other workpiece (Workpiece), and this workpiece can be the wafer of semiconductor industry or the glass substrate of TFT-LCD industry.Using unit of cargo to unit of cargo (Lot-to-Lot; During L2L) control, only need to measure the single workpiece in the whole unit of cargo, with purpose as feedback and feedforward control.Yet, when size of components is further dwindled, just need to use more strict process control.In the case, the control of L2L may be accurate inadequately, and must adopt workpiece to workpiece (Workpiece-to-Workpiece; W2W) control.In the control of W2W, each workpiece in the unit of cargo all needs measured.Therefore, for measuring each workpiece in the unit of cargo, time production cycle that the user need use a large amount of survey instruments and significantly increase.In addition, when carrying out the actual measurement of workpiece, can cause inevitably to measure and incur loss through delay, this measures to incur loss through delay and then can cause complicated control problem.

Therefore, need provide a kind of method of virtual measurement, just can provide (virtual) measured value of each workpiece so as to each workpiece that does not need actual measurement, to carry out the control of W2W.For example: under the metering system that still uses L2L control (only taking a sample test a single workpiece in the whole unit of cargo), carry out the control of W2W.Yet, generally speaking, use virtual measurement method to provide virtual measurement value to the control of W2W need take into account immediacy and accuracy, otherwise can't satisfy the demand of above-mentioned W2W control.

Summary of the invention

Therefore, press for very much a kind of two-stage virtual measurement method of development, to satisfy the demand of W2W control.

Summary of the invention

Technical matters to be solved by this invention is to provide a kind of virtual measurement method that utilizes the two-stage virtual measuring system, so as to the virtual measurement value in two kinds of stages is provided, and takes into account immediacy and accuracy simultaneously.

According to above-mentioned purpose of the present invention, provide a kind of virtual measurement method that utilizes the two-stage virtual measuring system.According to preferred embodiment of the present invention, this two-stage virtual measuring system comprises at least: a technological parameter data pre-processing module and a measurement data pre-processing module, in this virtual measurement method, at first this two-stage virtual measuring system is obtained the many groups of historical technological parameter data of producing board, each group technological parameter data comprises several technological parameters and respective value thereof, and wherein a plurality of historical workpiece are according to should the historical technological parameter data of many groups producing; This two-stage virtual measuring system also obtains a plurality of historical measurements from measuring board, and wherein these a little historical measurements are respectively workpiece according to these historical technological parameter mades () measured value for example: wafer or glass substrate.Then, use these historical technological parameter data and these a little historical measurements to set up first through this two-stage virtual measuring system and estimate pattern, wherein first foundation that estimates pattern is to estimate algorithm according to one, and this estimates algorithm and for example can be: multiple regression (Multi-Regression) algorithm, neural network (Neural Network; NN) algorithm or other prediction algorithm.Again, in this two-stage virtual measurement method, more seeing through this two-stage virtual measuring system uses aforesaid historical technological parameter data and historical measurement data to set up first reference model, wherein the foundation of this first reference model is basis and the aforesaid different reference algorithm of algorithm that estimates, and for example can be: multiple regression algorithm, neural network algorithm or other prediction algorithm.In this two-stage virtual measurement method, more see through this two-stage virtual measuring system and use aforesaid historical technological parameter, and set up the first statistical distance pattern according to a statistics distance algorithm again.This statistical distance algorithm for example can be: mahalanobis distance (Mahalanobis Distance) algorithm.

Then, this two-stage virtual measuring system is waited for the technological parameter data of each workpiece that collection production board is sent.After the complete technological parameter data aggregation of some workpiece is finished, see through this two-stage virtual measuring system and carry out a phase one immediately and estimate step.Estimate in the step in this phase one, the technological parameter data to the first of importing this workpiece estimate pattern, and calculate the phase one virtual measurement value (VM of this workpiece _I), to satisfy the demand of immediacy.Estimate in the step in this phase one, more import technological parameter data to the first reference model of this workpiece, and calculate the first reference prediction value.Then, see through that this confidence index module is calculated the overlapping area between the distribution of the distribution (Distribution) of phase one virtual measurement value of this workpiece and the first reference prediction value respectively and confidence desired value (the Reliance Index that produces the phase one virtual measurement value of this workpiece; RI), wherein work as overlapping area more greatly, then the confidence desired value is higher, represents the confidence level of the phase one virtual measurement value that corresponds to heal high.Estimate in the step in this phase one, also import technological parameter data to the first statistical distance pattern of this workpiece, and calculate overall similarity desired value (the GlobalSimilarity Index of the pairing technological parameter data of phase one virtual measurement value of this workpiece; GSI).

Generally speaking, production system can be extracted a certain workpiece out (promptly by the workpiece taken a sample test in each blocks casket; As taking a sample test 1 in the card casket that 25 wafer is housed at), and deliver to and measure board and measure.When measuring board and obtain the actual measured value of this workpiece of being taken a sample test, carry out a subordinate phase and estimate step.Estimate in the step in this subordinate phase, this technological parameter data and actual measured value by the workpiece taken a sample test is added historical technological parameter data and historical measurements, train aforesaid first to estimate pattern and first reference model and become one second and estimate pattern and one second reference model again; Or come adjustment aforesaid first to estimate pattern and first reference model by the technological parameter data of the workpiece taken a sample test and actual measured value and become one second and estimate pattern and one second reference model with this.Then, the technological parameter data to the second of all workpiece in the card casket under the workpiece that input is taken a sample test estimate the pattern and second reference model, and recalculate the subordinate phase virtual measurement value (VM of each workpiece in this card casket _II) and the second reference prediction value.Then, calculate respectively each workpiece of this card in casket subordinate phase virtual measurement value the distribution of distribution and the second reference prediction value between overlapping area and produce the confidence desired value of the subordinate phase virtual measurement value of each interior workpiece of this card casket, it is bigger wherein to work as overlapping area, then the confidence desired value is higher, represents the confidence level of the second virtual measurement value that corresponds to higher.The subordinate phase virtual measurement value (VM that this place is estimated again _II) will be than the phase one virtual measurement value (VM of previous gained _I) accurately, to satisfy the demand of accuracy.Simultaneously, estimate pattern with second and replace first and estimate pattern and become new first and estimate pattern, so that be used for estimating the phase one virtual measurement value (VM of the workpiece of newly coming in _I) and its confidence desired value.

Estimate in the step in this subordinate phase, also the technological parameter data of the aforesaid workpiece of being taken a sample test can be added historical technological parameter data, train the aforesaid first statistical distance pattern again and become one second statistical distance pattern; Or come the aforesaid first statistical distance pattern of adjustment and become one second statistical distance pattern with these technological parameter data by the workpiece taken a sample test.Then, technological parameter data to the second statistical distance pattern of all workpiece in the card casket under the workpiece that input is taken a sample test, and recalculate the overall similarity desired value (GSI) that this blocks the pairing technological parameter data of subordinate phase virtual measurement value of each workpiece in casket.Simultaneously, replace the first statistical distance pattern and become the first new statistical distance pattern with the second statistical distance pattern, so that be used for estimating the phase one virtual measurement value (VM of the workpiece of newly coming in _I) the overall similarity desired value (GSI) of pairing technological parameter data.

Again, this two-stage virtual measurement method also comprises at least: carry out the pre-treatment of technological parameter data, with the technological parameter data of suppressing exception, and screen out more unessential parameter in the technological parameter data.

Again, this two-stage virtual measurement method also comprises at least: carry out the measurement data pre-treatment, to screen out the exceptional value in the actual measured value.

According to another preferred embodiment of the present invention, carrying out subordinate phase when estimating step, for must train again or only must adjustment the mechanism of (estimate, with reference to and statistical distance pattern) be, when the production board has left unused above a schedule time or artificially manual the indication has been arranged, just training again, otherwise only must adjustment get final product.

Therefore, use the present invention, the relative confidence index of virtual measurement value and the overall similarity index in two kinds of stages can be provided, thereby can take into account immediacy and accuracy simultaneously, satisfy the demand of W2W control.

Description of drawings

For more complete understanding the present invention and advantage thereof, please refer to above-mentioned narration and cooperate following graphic, wherein:

Fig. 1 is the configuration diagram that illustrates according to the two-stage virtual measuring system of preferred embodiment of the present invention;

Fig. 2 is the synoptic diagram that illustrates the confidence desired value of explanation preferred embodiment of the present invention;

Fig. 3 is the synoptic diagram that illustrates the confidence metrics-thresholds of explanation preferred embodiment of the present invention;

Fig. 4 is the schematic flow sheet that illustrates according to the two-stage virtual Measurement Algorithm of preferred embodiment of the present invention;

Fig. 5 illustrates to use phase one virtual measurement value that preferred embodiment of the present invention obtained and subordinate phase virtual measurement value to batch synoptic diagram to batch (R2R) control system;

Fig. 6 illustrates the result schematic diagram of using phase one virtual measurement value, subordinate phase virtual measurement value and actual measured value that preferred embodiment of the present invention obtained.

Wherein, Reference numeral:

10: technological parameter data pre-processing module

12: the measurement data pre-processing module

20,20a, 20b: produce board

30: measure board

40: confidence index module

50: similarity index module

60: estimate pattern

62: two stage calculation mechanism

80: the card casket

82: workpiece

90,90a, 90b: two-stage virtual measuring system

94a, 94b:R2R control system

100: the phase one estimates step

102: the technological parameter data of collecting a certain workpiece

110: whether the technological parameter data of inspecting this workpiece are collected and are finished

120: the VM that calculates this workpiece _IRI that follows with it and GSI

200: subordinate phase estimates step

202: the actual measurement data of collecting a certain workpiece

210: whether the actual measurement data of inspecting this workpiece is collected and is finished

220: inspect the actual measurement data of this workpiece and the relevance of technological parameter data

230: whether belong to same workpiece with decision actual measurement data and technological parameter data

240: judge whether produce board leaves unused above one period schedule time

250: inspect whether manual indication is arranged

260: training estimates pattern, reference model and statistical distance pattern again

270: adjustment estimates pattern, reference model and statistical distance pattern

280: upgrade pattern, reference model and the statistical distance pattern of estimating

290: the VM that calculates all workpiece in the affiliated card casket of this workpiece _IIRI that follows with it and GSI

A: phase one virtual measurement point

Embodiment

Please refer to Fig. 1, it is the configuration diagram according to the two-stage virtual measuring system of preferred embodiment of the present invention.The two-stage virtual measuring system 90 of present embodiment comprises at least: technological parameter data pre-processing module 10, measurement data pre-processing module 12, estimate pattern 60, confidence index module 40 and similarity index module 50.Technological parameter data pre-processing module 10 is at putting in order and standardization from the original process supplemental characteristic of producing board 20, the suppressing exception data also filter out important parameter, inessential parameter is got rid of, avoiding producing interference effect, and the impact prediction precision.Measurement data pre-processing module 12 is at screening from the measurement data of measuring board 30, to remove exceptional value wherein.Estimating pattern 60 is to utilize two stage calculation mechanism 62 and estimate phase one and the subordinate phase virtual measurement value (VM that algorithm estimates a plurality of workpiece (not indicating) in the card casket 80 _IAnd VM _II).The algorithm that estimates that may select for use has: various prediction algorithms such as multiple regression algorithm, neural network algorithm.Confidence index module 40 is the confidence levels that are used for assessing the virtual measurement value, and produces confidence index (RI).Similarity index module 50 is technological parameter data of being used for assessing present input and the similarity degree that estimates all supplemental characteristics that are used for training modeling in the pattern 60, and the similarity index (GSI) of generation technological parameter, this similarity index is to judge the confidence degree of virtual measurement system in order to auxiliary confidence index.

Before estimating pattern 60 runnings, must be from producing technological parameter data (historical technological parameter data) that board 20 obtained and being sent to technological parameter data pre-processing module 10 and measurement data pre-processing module 12 respectively, to carry out the data pre-treatment from measuring the obtained mass measurement data (historical measurements) of board 30.The input data that these technological parameter data after pre-treatment and standardization and mass measurement data are the pattern of estimating 60.Then, adopt historical technological parameter data and corresponding historical mass measurement data train (foundation) for example neural network (NN) estimate pattern.Estimate pattern 60 and have two stage calculation mechanism 62, in order to difference phase one virtual measurement value (VM _I) and subordinate phase virtual measurement value (VM _II) confidence desired value (RI) and the overall similarity desired value (GSI) corresponding with it.The running of this pair stage calculation mechanism 62 will be described after explanation confidence desired value (RI) theoretical foundation relevant with overall similarity desired value (GSI).

Below, the confidence desired value theoretical foundation relevant with the overall similarity desired value is described earlier.

Whether confidence index and technological parameter similarity index can be trusted in order to understand the virtual measurement value in real time.Whether confidence index module 40 is the technological parameter data of producing board by analyzing, and calculates one between confidence value (confidence desired value) zero and between one, can be trusted with the result who judges virtual measurement.Similarity index module 50 is in order to calculate the overall similarity desired value of technological parameter.Technological parameter data that the overall similarity desired value is defined as present input and the similarity degree that estimates all supplemental characteristics that are used for training modeling in the pattern 60.

The confidence index

As shown in table 1, suppose to collect the data that the n group is measured at present, comprise technological parameter data (X _i, i=1,2 ..., n) and corresponding actual measurement Value Data (y _i, i=1,2 ..., n), wherein every group of process data includes p parameter (autoregressive parameter 1 is to parameter p), i.e. X _i=[x _{I, 1}, x _{I, 2}..., x _{I, p}] ^TIn addition, technological parameter data when also collecting (m-n) actual production, but remove y _N+1, there is no the actual measurement Value Data outward, promptly in the workpiece of (m-n) actual production, for example only take a sample test that the first stroke workpiece carries out actual measurement, again with its actual measurement y _N+1Infer the quality of other (m-n-1) workpiece.

Table 1 raw data example

In table 1, y ₁, y ₂..., y _nBe historical measurements, y _N+1Actual measured value for the unit one in the workpiece unit of cargo just aborning.Usually, one group of actual measured value (y _i, i=1,2 ..., n) for having average μ, the normality of standard deviation sigma is distributed, i.e. y _i～N (μ, σ ²).

At sample group (y _i, i=1,2 ..., average n) and standard deviation with all actual measured value data normalizations after, can obtain

(being also referred to as z mark (z Scores)), wherein the average of each z mark is 0, standard deviation is 1, promptly

For actual measurement data, if

More near 0, represent that then measurement data is more near the specification central value.Its standardized formula is as follows:

$Z_{y_i}=\frac{y_i-\stackrel{\‾}{y}}{{\mathrm{\σ}}_y},i=\mathrm{1,2},...,n---\left(1\right)$

$\stackrel{\&OverBar;}{y}=\frac{1}{n}(y_1+{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="S071B0908820070620E000021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2+...+y_n)---\left(2\right)$

${\mathrm{\&sigma;}}_y=\sqrt{\frac{1}{n-1}[{(y_1-\stackrel{\&OverBar;}{y})}^2+{(y_2-\stackrel{\&OverBar;}{y})}^2+...+{(y_n-\stackrel{\&OverBar;}{y})}^2]}---\left(3\right)$

Wherein

y _iIt is i group actual measurement Value Data;

Be the actual measurement Value Data behind i group data normalization;

Average for all actual measurement Value Datas;

σ _yStandard deviation for all actual measurement Value Datas;

Explanation herein is that the algorithm that estimates of application class neural network (NN) algorithm is set up the pattern that estimates of carrying out virtual measurement, and the checking of setting up with the reference algorithm of for example multiple regression algorithm this estimate the reference model of pattern.Yet the present invention also can use other algorithm for estimating algorithm or with reference to algorithm, as long as be to be different to estimate algorithm with reference to algorithm, so the present invention is also not subject to the limits.

When application class neural network algorithm and multiple regression algorithm, be error sum of squares (Sum of Square Error as its condition of convergence; SSE) under the Zui Xiao condition, during and n → ∞, this two-mode actual measured value after the standardization separately is defined as With , then its all should with the actual measured value after the real standardization

Identical.In other words, when n → ∞,

All represent the actual measured value after the standardization, but change its title for purpose in response to different mode.Therefore And

Expression

Be same allocated, but because different estimation models, make that the mean value of these two kinds of prediction algorithms is different with the estimated value of standard deviation.Also be that the average estimator that NN estimates after the mode standardization will be different with the standard deviation estimator with the average estimator after the multiple regression mode standardization with the standard deviation estimator.

The confidence desired value is the Reliability that is designed to judge the virtual measurement value, so the confidence desired value should be taken into account that the statistics of virtual measurement value is distributed and the statistics of actual measured value is distributed

Similarity degree between the two.Yet, when applying virtual is measured, there is no the Reliability (significantly, just not needed virtual measurement) that actual measured value can be used for assessing the virtual measurement value if obtain actual measured value.So the present invention adopts the statistics of being estimated by reference algorithm (for example multiple regression algorithm) to distribute

Replace

Statistics distribute.Of the present inventionly also can be other relevant prediction algorithm, so the present invention is also not subject to the limits with reference to algorithm.

Please refer to Fig. 2, it is the synoptic diagram of the confidence desired value of explanation preferred embodiment of the present invention.Confidence desired value of the present invention is defined as the distribution of calculating the prediction (virtual measurement value) that estimates pattern (for example adopting neural network (NN) algorithm)

The distribution of the prediction (reference measurement values) of (shown in curve 1) and reference model (for example adopting the multiple regression algorithm)

(shown in curve 2) common factor area coverage values (overlapping area A) between the two.Therefore, the formula of confidence desired value is as follows:

$\mathrm{RI}=2{\&Integral;}_{\frac{Z_{{\hat{y}}_{\mathrm{Ni}}}+Z_{{\hat{y}}_{\mathrm{<figure-callout\; id="2"\; label="ri"\; filenames="S071B0908820070620E000021.png"\; state="\{\{state\}\}">ri</figure-callout>}}}}{2}}^{\&infin;}\frac{1}{\sqrt{2\mathrm{\&pi;}}\mathrm{\&sigma;}}{e}^{-\frac{1}{2}{\left(\frac{x-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)}^2}\mathrm{dx}---\left(4\right)$

Wherein ought be then

When then

σ is made as 1

The confidence desired value increases along with the increase of overlapping area A.This phenomenon points out that the result who uses the pattern of estimating to obtain approaches the result who uses reference model to obtain, thereby corresponding virtual measurement value is more reliable.Otherwise the fiduciary level of corresponding virtual measurement value reduces along with the minimizing of overlapping area A.When by

Estimated distribution With by

Estimated distribution

When overlapping fully, according to statistical theory of distribution, its confidence desired value equals 1; And when two distributed almost completely separately, its confidence desired value then leveled off to 0.

Below the explanation estimate mode computation virtual measurement value ( With ) the method for distribution.

In estimating pattern,, then can suppose [given if the condition of convergence is minimum error quadratic sum (SSE)

Down, The average that is assigned as equal variance and be

Distribution], promptly given Down,

And The NN estimator be

The NN estimator be

Before carrying out the modeling that NN estimates pattern, need carry out the step of technological parameter data normalization earlier.

It is as follows that NN estimates pattern technological parameter data-standardizing formula:

$Z_{x_{i,j}}=\frac{x_{i,j}-{\stackrel{\&OverBar;}{x}}_j}{{\mathrm{\&sigma;}}_{x_j}},i=\mathrm{1,2},...,n,n+1,...,m;j=\mathrm{1,2},...,p---\left(5\right)$

${\stackrel{\&OverBar;}{x}}_j=\frac{1}{n}(x_{1,j}+x_{2,j}+...+x_{n,j})---\left(6\right)$

${\mathrm{\&sigma;}}_{x_j}=\sqrt{\frac{1}{n-1}[{(x_{1,j}-{\stackrel{\&OverBar;}{x}}_j)}^2+{(x_{2,j}-{\stackrel{\&OverBar;}{x}}_j)}^2+...+{(x_{n,j}-{\stackrel{\&OverBar;}{x}}_j)}^2]}---\left(7\right)$

Wherein

x _{I, j}Be j technological parameter in the i group process data;

It is the technological parameter after j the standardization in the i group process data;

Be the mean value of j technological parameter data;

It is the standard deviation of j technological parameter data;

Use the process data after this n organizes standardization Actual measured value after the standardization of n group therewith

Come construction NN to estimate pattern.Then, the process data after the standardization of input m group Estimate in the pattern to NN, to obtain the virtual measurement value after the corresponding standardization

Therefore,

(promptly) estimated value and

(promptly) estimated value can be calculated by formula as follows:

${\widehat{\mathrm{\&mu;}}}_{Z_{y_i}}=Z_{{\hat{y}}_{N_i}},i=\mathrm{1,2},...,n,n+1,...,m---\left(8\right)$

${\stackrel{\&OverBar;}{Z}}_{{\hat{y}}_N}=\frac{1}{n}(Z_{{\hat{y}}_{N_1}}+Z_{{\hat{y}}_{N_2}}+...+Z_{{\hat{y}}_{N_n}})---\left(10\right)$

Wherein

Mean value for the virtual measurement value after the standardization

Below the explanation by multiple regression mode computation reference prediction value (

With

) method.

The basic assumption of multiple regression algorithm is [given Down,

The average that is assigned as equal variance and be

Distribution], promptly given

Down,

And

The multiple regression estimator be The multiple regression estimator

For trying to achieve the process data after n organizes standardization Actual measured value after the standardization of n group therewith Between relation, must definition utilizing in the multiple regression analysis the pairing weight of these p parameter is (β _R0, β _R1, β _R2..., β _Rp).Construction

With

Concern as follows:

${\mathrm{\&beta;}}_{r0}+{\mathrm{\&beta;}}_{r1}{Z}_{x_{\mathrm{1,1}}}+{\mathrm{\&beta;}}_{r2}{Z}_{x_{\mathrm{1,2}}}+...+{\mathrm{\&beta;}}_{\mathrm{rp}}{Z}_{x_{1,p}}=Z_{y_1}$

${\mathrm{\&beta;}}_{r0}+{\mathrm{\&beta;}}_{r1}{Z}_{x_{\mathrm{2,1}}}+{\mathrm{\&beta;}}_{r2}{Z}_{x_{\mathrm{2,2}}}+...+{\mathrm{\&beta;}}_{\mathrm{rp}}{Z}_{x_{2,p}}=Z_{y_2}---\left(11\right)$

${\mathrm{\&beta;}}_{r0}+{\mathrm{\&beta;}}_{r1}{Z}_{x_{n,1}}+{\mathrm{\&beta;}}_{r2}{Z}_{x_{n,2}}+...+{\mathrm{\&beta;}}_{\mathrm{rp}}{Z}_{x_{n,p}}=Z_{y_n}$

Suppose $Z_y=\left(\begin{array}{c}{Z}_{y_1}\\ {\mathrm{<figure-callout\; id="2"\; label="Z\; y"\; filenames="S071B0908820070620E000021.png"\; state="\{\{state\}\}">Z</figure-callout>}}_{y_{}}\\ .\\ .\\ .\\ Z_{y_n}\end{array}\right)---\left(12\right)$

Utilize the least square method in the multiple regression analysis on the statistics, can try to achieve parameter beta _rEstimator

Promptly

${\widehat{\mathrm{\&beta;}}}_r={\left(Z_x^T{Z}_x\right)}^{-1}{Z}_x^T{Z}_y---\left(14\right)$

Then, multiple regression pattern can obtain:

$Z_{{\hat{y}r}_i}={\widehat{\mathrm{\&beta;}}}_{r0}+{\widehat{\mathrm{\&beta;}}}_{r1}{Z}_{x_{i,1}}+{\widehat{\mathrm{\&beta;}}}_{r2}{Z}_{x_{i,2}}+...+{\widehat{\mathrm{\&beta;}}}_{\mathrm{rp}}{Z}_{x_{i,p}}$

i＝1，2，…，n，n+1，…，m (15)

Therefore, when estimating the stage, after the technological parameter data are come in, can obtain its pairing multiple regression estimated value standard variance according to formula (15)

The multiple regression estimator be

Have:

${\widehat{\mathrm{\&sigma;}}}_{Z_{\hat{y}r}}=\sqrt{\frac{1}{n-1}[{(Z_{{\hat{y}}_{r_1}}-{\stackrel{\&OverBar;}{Z}}_{{\hat{y}}_r})}^2+{(Z_{{\hat{y}}_{r_2}}-{\stackrel{\&OverBar;}{Z}}_{{\hat{y}}_r})}^2+...+{(Z_{{\hat{y}}_{r_n}}-{\stackrel{\&OverBar;}{Z}}_{{\hat{y}}_r})}^2]}---\left(16\right)$

${\stackrel{\&OverBar;}{Z}}_{{\hat{y}}_r}=\frac{1}{n}(Z_{{\hat{y}}_{r_1}}+Z_{{\hat{y}}_{{\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="S071B0908820070620E000021.png"\; state="\{\{state\}\}">r</figure-callout>}}_2}}+...+Z_{{\hat{y}}_{r_n}})---\left(17\right)$

Estimate the estimator of pattern when trying to achieve NN

With

And the estimator of multiple regression pattern

With After, can draw normality distribution diagram as shown in Figure 2, calculate to use the distribution common factor area coverage values (overlapping area A) between the two of the prediction (reference measurement values) of the distribution of the prediction (virtual measurement value) that estimates pattern (for example adopting neural network (NN) algorithm) and reference model (for example adopting the multiple regression algorithm), can obtain each (confidence desired value of virtual measurement value.

After obtaining confidence desired value (RI), must stipulate a confidence metrics-thresholds (RI _T).If RI＞RI _T, then the degree of reliability of virtual measurement value is can be received.Decision confidence metrics-thresholds (RI is below described _T) method:

Stipulating confidence metrics-thresholds (RI _T) before, at first need stipulate out the maximum admissible error upper limit (E _L).The error of virtual measurement value (Error) is actual measured value y _iObtain with the pattern that estimates by NN

Difference, again divided by the percent of the absolute value behind the mean value of all actual measured value, promptly

${\mathrm{Error}}_i=\left|\frac{y_i-{\hat{y}}_{\mathrm{Ni}}}{\stackrel{\&OverBar;}{y}}\right|\&times;100\%---\left(18\right)$

Then, can specify the maximum admissible error upper limit (E according to the degree of accuracy specification of defined error of formula (18) and virtual measurement _L).Therefore, confidence metrics-thresholds (RI _T) be to be defined as corresponding to the maximum admissible error upper limit (E _L) confidence desired value (RI), as shown in Figure 3.That is,

${\mathrm{RI}}_T=2{\&Integral;}_{Z_{\mathrm{Center}}}^{\&infin;}\frac{1}{\sqrt{2\mathrm{\&pi;}}\mathrm{\&sigma;}}{e}^{-\frac{1}{2}{\left(\frac{x-\mathrm{\&mu;}}{\mathrm{\&sigma;}}\right)}^2}\mathrm{dx}---\left(19\right)$

μ and σ are defined in the formula (4); And

$Z_{\mathrm{Center}}=Z_{{\hat{y}}_{\mathrm{Ni}}}+[\stackrel{\&OverBar;}{y}\&times;(E_L/2)]/{\mathrm{\&sigma;}}_y---\left(20\right)$

σ wherein _yBe to be defined in the formula (3).

Overall similarity index (GSI)

As mentioned above, when applying virtual is measured, there is not actual measured value can obtain to verify the degree of accuracy of virtual measurement value.Therefore, with the multiple regression estimated value after the standardization

Actual measured value after the replacement standardization

Calculate confidence desired value (RI).Yet this kind replacement may cause the error of confidence desired value (RI), and in order to compensate this situation, the overall similarity index (GSI) that the present invention proposes technological parameter helps judge the degree of reliability of virtual measurement.

The notion of GSI proposed by the invention is that all the history parameters data when adopt apparatus and process supplemental characteristic when the input of virtual measurement system with modeling are at present compared, and obtains the technological parameter data of importing and the similarity degree index of all history parameters data.

The present invention can quantize similarity with various statistical distance algorithm (for example mahalanobis distance algorithm).Mahalanobis distance is the statistical distance algorithm of being introduced 1936 Christian eras by P.C.Mahalanobis.This kind technological means is based on relevance between variable with identification with analyze the kenel of different sample groups.Mahalanobis distance is the method in order to the similarity between decision unknown sample group and known sample group, and the method is considered the relevance between data set and have yardstick unchangeability (Scale Invariant) that promptly the size with measured value is not relevant.If data have high similarity, then the mahalanobis distance that is calculated will be less.

The present invention is a size of utilizing the GSI (mahalanobis distance) calculated, differentiates the technological parameter data newly advanced whether similar in appearance to all process datas of modeling.If the GSI that calculates is little, represent that then the technological parameter data class that newly advances is similar to the process data of modeling, therefore the virtual measurement value of the technological parameter data of newly advancing (high similarity) will be more accurate.Otherwise if the GSI that calculates is excessive, then some is different for the process data of the technological parameter data newly advanced of expression and modeling.Thereby the confidence degree of the accuracy of the virtual measurement value of the technological parameter data of newly advancing (low similarity) is lower.

Estimate the standardization technological parameter data of pattern

Computing formula be suc as formula shown in (5), (6) and (7).At first, definition example edition supplemental characteristic X _M=[x _{M, I}, x _{M, 2}..., x _{M, p}] ^T, x wherein _{M, j}Equal

So, then to be 0 (also be the modeling parameters Z after the standardization to each parameter of the modelled process data after the standardization _MjBe 0).In other words, Z _M=[Z _{M, 1}, Z _{M, 2}..., Z _{M, p}] ^TIn all parameters be 0.Next calculate the related coefficient between the modeling parameters after each standardization.

Suppose that the related coefficient between s parameter and t the parameter is r _St, and k group data are wherein arranged, then

$r_{\mathrm{st}}=\frac{1}{k-1}\underset{l=1}{\overset{k}{\mathrm{\&Sigma;}}}{z}_{\mathrm{sl}}\&CenterDot;z_{\mathrm{tl}}=\frac{1}{k-1}(z_{s1}\&CenterDot;z_{t1}+z_{s2}\&CenterDot;{\mathrm{<figure-callout\; id="2"\; label="z\; t"\; filenames="S071B0908820070620E000021.png"\; state="\{\{state\}\}">z</figure-callout>}}_t+...+z_{\mathrm{sk}}\&CenterDot;z_{\mathrm{tk}})---\left(21\right)$

After finishing the related coefficient of calculating between each parameter, it is as follows to obtain correlation matrix:

Suppose the inverse matrix (R of R ^-1) be defined as A, then

$A=R^{-1}=\left[\begin{array}{cccc}{a}_{11}& a_{12}& ...& a_{1p}\\ a_{21}& a_{22}& ...& a_{2p}\\ ...& ...& ...& ...\\ a_{p1}& a_{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="S071B0908820070620E000021.png"\; state="\{\{state\}\}">p</figure-callout>}2}& ...& a_{\mathrm{pp}}\end{array}\right]---\left(23\right)$

So, the standardized technological parameter data of λ pen (Z _λ) and standardized example edition supplemental characteristic (Z _M) between mahalanobis distance Computing formula is as follows:

$D_{\mathrm{\&lambda;}}^2={(Z_{\mathrm{\&lambda;}}-Z_M)}^T{R}^{-1}(Z_{\mathrm{\&lambda;}}-Z_M)---\left(24\right)$

$=Z_{\mathrm{\&lambda;}}^T{R}^{-1}{Z}_{\mathrm{\&lambda;}}$

Can get

$D_{\mathrm{\&lambda;}}^2=\underset{j=1}{\overset{p}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{a}_{\mathrm{ij}}{z}_{\mathrm{i\&lambda;}}{z}_{\mathrm{j\&lambda;}}---\left(25\right)$

And the GSI value of λ process data is

After obtaining the GSI value, should define GSI threshold value (GSI _T).Usually, GSI threshold value decided at the higher level but not officially announced is historical technological parameter data GSI _a(wherein on behalf of each, a organize historical technological parameter data) peaked 2 to 3 times.

Please continue with reference to Fig. 1.Finish the statistical distance pattern of the reference model of the pattern of estimating 60, confidence index module 40 and similarity index module 50 in foundation after, just can carry out virtual measurement to a plurality of workpiece of card casket 80.Generally speaking, only there is a workpiece 82 (for example: unit one) can be taken a sample test and deliver to and measure board 30 and carry out actual measurement, and the result of the actual measurement of this unit one 82 often need pass through a few hours (for example: 6 hours) and just can obtain afterwards in a plurality of workpiece of card casket 80 to be measured.Estimate pattern 60, confidence index module 40 and similarity index module 50 and arranged two stages: phase one and subordinate phase at the virtual measurement value that a plurality of workpiece produced in the card casket 80, the virtual measurement value in each stage also is attended by the confidence index and the overall similarity index, judges with the confidence degree of auxiliary virtual measurement value.Confidence degree index and overall similarity index that so-called [phase one] virtual measurement value is followed with it are after the complete technological parameter data aggregation of each workpiece is finished, and the technological parameter data of importing workpiece immediately obtain to estimating pattern 60, confidence index module 40 and similarity index module 50.Confidence index that so-called [subordinate phase] virtual measurement value is followed with it and similarity index then are when measuring board and obtain the actual measured value of the workpiece 82 that this quilt taken a sample test, the technological parameter data and the actual measured value of this workpiece of being taken a sample test 82 are added historical technological parameter data and historical measurements, again train or adjustment estimates the reference model and the similarity index module 50 statistical distance patterns of pattern 60, confidence index module 40, recomputate the subordinate phase virtual measurement value (VM of each workpiece in the card release casket 80 again _II) confidence index and the overall similarity index followed with it.

Below of the present invention pair of stage computing of explanation (virtual measurement) mechanism 62.

Please refer to Fig. 1 and Fig. 4, it is the schematic flow sheet according to the two-stage virtual Measurement Algorithm of preferred embodiment of the present invention.

Finish the statistical distance pattern of the reference model of the pattern of estimating 60, confidence index module 40 and similarity index module 50 in foundation after, present embodiment begins to wait for collects the technological parameter data of producing each workpiece that board 20 sent.After the complete technological parameter data aggregation of each workpiece is finished, carry out (triggering) phase one immediately to estimate step 100; When measuring board 30 and obtain the actual measured value of this workpiece of being taken a sample test 82, then carry out (triggering) subordinate phase and estimate step 200.

Below be that illustration illustrates that the phase one estimates step 100 and subordinate phase estimates step 200 with all workpiece in the single card casket.

Estimate in the step 100 in the phase one, carry out step 102, to collect the technological parameter data of a certain workpiece (can be workpiece 82 or other the arbitrary workpiece taken a sample test).Then, carry out step 110, whether collect with the technological parameter data of inspecting this workpiece and finish.If the result of step 110 then proceeds step 102 for not; If the result of step 110 for being, then carry out step 120, confidence index and the overall similarity index followed with it with the virtual measurement value of calculating this workpiece, i.e. phase one virtual measurement value (VM _I) confidence index (RI) and the overall similarity index (GSI) followed with it.

Estimate in the step 200 in subordinate phase, carry out step 202, to collect the actual measurement data of a certain workpiece (i.e. the workpiece of being taken a sample test 82).Then, carry out step 210, whether collect with the actual measurement data of inspecting this workpiece 82 and finish.If the result of step 210 then proceeds step 202 for not; If the result of step 210 for being, then carry out step 220, with the actual measurement data of inspecting this workpiece 82 relevance of the pairing technological parameter data of workpiece 82 therewith.Then, carry out step 230 and whether belong to same workpiece (i.e. the workpiece of being taken a sample test 82) with decision actual measurement data and technological parameter data.If the result of step 230 then proceeds step 202 for not; If the result of step 230 then carry out step 240 for being, to judge whether produce board 20 leaves unused above one period schedule time.If the result of step 240 then carries out step 250 to inspect whether manual indication is arranged for not.The result of step 250 then carries out step 270 and estimates the reference model of pattern 60, confidence index module 40 and the statistical distance pattern of similarity index module 50 with adjustment for not.So-called [adjustment] is weighted value or the parameter value of adjusting each pattern according to the present one group of actual measurement data that obtains and technological parameter data, generally only needs the several seconds to finish.If step 250 or step 240 result are for being, i.e. representative has manual indication (usually keep in repair or during part exchanging); Or the characteristic of producing board 20 after one period schedule time just must carry out step 260, to train the reference model of pattern 60, confidence index module 40 and the statistical distance pattern of similarity index module 50 of estimating again when producing bigger variation.So-called [training again] is with the present one group of actual measurement data that obtains and the technological parameter data add historical technological parameter data and historical measurements is trained each pattern again, generally needs consumption just can finish in several minutes.

Behind completing steps 260 or 270, just can carry out step 280 and estimate the reference model of pattern 60, confidence index module 40 and the statistical distance pattern of similarity index module 50 with renewal.The pattern that estimates 60 that these are new, reference model and statistical distance pattern also are provided to step 120, to calculate the phase one virtual measurement value (VM of next workpiece _I) confidence index (RI) and the overall similarity index (GSI) followed with it.Simultaneously, carry out step 290, to use the new pattern that estimates 60, reference model and statistical distance pattern, recomputate confidence index and overall similarity index that the virtual measurement value of each workpiece in the card casket 80 under the workpiece taken a sample test 82 is followed with it, i.e. subordinate phase virtual measurement value (VM _II) confidence index (RI) and the overall similarity index (GSI) followed with it.

Owing to produce phase one virtual measurement value (VM _I) the confidence index of following with it (RI) and the pattern that estimates 60, reference model and the statistical distance pattern of overall similarity index (GSI) needn't wait for the actual measured value of workpiece, only need the pairing technological parameter data of workpiece to get final product, so can obtain phase one virtual measurement value (VM immediately _I) confidence index (RI) and the overall similarity index (GSI) followed with it, thereby can satisfy the real-time demand of virtual measurement.When the actual measured value of the workpiece 82 that obtains to be taken a sample test; estimate pattern 60, reference model and statistical distance pattern and can be re-used the subordinate phase virtual measurement value (VM that these new patterns are estimated each workpiece in the card casket 80 again by adjustment according to this or training again _II) confidence index (RI) and the overall similarity index (GSI) followed with it.These all patterns through adjustment or training again also are used and upgrade original pattern, to predict the phase one virtual measurement value (VM of next workpiece _I) confidence index (RI) and the overall similarity index (GSI) followed with it, thereby can guarantee the accuracy demand of virtual measurement.

Please refer to Fig. 5, phase one virtual measurement value that it is obtained for application preferred embodiment of the present invention and subordinate phase virtual measurement value are to the synoptic diagram of R2R control system, wherein produce the board that board 20a is ongoing technology (as deposition machine), producing board 20b is the board (as work-table of chemicomechanical grinding mill) of next technology.Embodiments of the invention can be applicable to for example a plurality of R2R control system 94a and the 94b of wafer factory, and wherein 94a of R2R system and 94b can be L2L control system or W2W control system.When the 94a of R2R system and 94b are the W2W control system, because phase one virtual measurement value (VM _I) can produce immediately by the two-stage virtual measuring system 90a (90b) of for example thickness prediction, so can offer feedback (Feedback) input of the W2W control system 94a (94b) that produces board 20a (20b), to satisfy the demand of its real-time.And subordinate phase virtual measurement value (VM _II) have quite good accuracy, so can offer feedforward (Feedward) input of W2W control system 94b of the production board 20b of next technology.

In addition, when the 94a of R2R system and 94b are the L2L control system, because the L2L control system does not need tool such as the required real-time demand of measuring piecewise of W2W control system, so can be with subordinate phase virtual measurement value (VM _II) offer the feedback input of the L2L control system 94a that produces board 20a, with the feedforward input of the L2L control system 94b of the production board 20b of next technology.Please refer to Fig. 6, the phase one virtual measurement value (VM that it is obtained for application preferred embodiment of the present invention _I), subordinate phase virtual measurement value (VM _II) and the result schematic diagram of actual measured value, wherein, represents phase one virtual measurement value (VM _I) result schematic diagram, zero expression subordinate phase virtual measurement value (VM _II) result schematic diagram,

The result schematic diagram of expression actual measured value.Wherein with mean absolute error number percent (Mean Absolute Percentage Error; MAPE) and maximum error (Max Error) assess VM _I, VM _IIAccuracy.Phase one virtual measurement value (VM _I) MAPE and Max Error be respectively 1.248% and 0.603%, and subordinate phase virtual measurement value (VM _II) MAPE and Max Error be respectively 0.281% and 0.070%.As shown in Figure 5, subordinate phase virtual measurement value (VM _II) almost consistent with actual measured value, and phase one virtual measurement value (VM _I) error also quite little, but still compare VM _IIError big.In addition, the confidence index of phase one virtual measurement point A (sample 14) and overall similarity index are all above its threshold value, and the confidence level of the phase one virtual measurement value of representative sample 14 is lower.

By the preferred embodiment of the invention described above as can be known, two-stage virtual measurement method of the present invention can be taken into account immediacy simultaneously (by VM _I) with accuracy (by VM _II), so can satisfy the demand of W2W control.

Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; being familiar with those of ordinary skill in the art ought can make various corresponding changes and distortion according to the present invention, but these corresponding changes and distortion all should belong to the protection domain of the appended claim of the present invention.

Claims (39)

1. a virtual measurement method that utilizes the two-stage virtual measuring system is characterized in that, this virtual measurement method comprises at least:

This two-stage virtual measuring system is obtained many groups of historical technological parameter data of producing board, and wherein a plurality of historical workpiece are according to should the historical technological parameter data of many groups producing;

This two-stage virtual measuring system is measured board from one and is obtained a plurality of historical measurements, and wherein those historical measurements are respectively the measured value of these a plurality of historical workpiece;

Seeing through this two-stage virtual measuring system uses those to organize historical technological parameter data and those historical measurements to set up one first and estimate pattern, this first foundation that estimates pattern is to estimate algorithm according to one, and this estimates algorithm is to be selected from a group that is made up of a multiple regression algorithm and a neural network algorithm;

This two-stage virtual measuring system is waited for the technological parameter data of a plurality of workpiece that this production board of collection is sent;

After one technological parameter data aggregation of those workpiece that this production board is sent is finished, see through this two-stage virtual measuring system and carry out a phase one immediately and estimate step, this phase one estimates step and comprises at least:

The technological parameter data of importing this person of those workpiece that this production board sent first estimate pattern to this, and calculate this person's of those workpiece a phase one virtual measurement value;

This two-stage virtual measuring system obtains an actual measured value of a workpiece of being taken a sample test from this measurement board;

This two-stage virtual measuring system obtains the technological parameter data of the workpiece that this quilt takes a sample test;

This two-stage virtual measuring system obtains the affiliated technological parameter data of blocking all workpiece in the casket of workpiece that this quilt is taken a sample test; And

During this actual measured value of the workpiece of taking a sample test when this quilt of obtaining those workpiece from this measurement board, see through this two-stage virtual measuring system and carry out a subordinate phase and estimate step, wherein this subordinate phase estimates step and comprises at least:

The technological parameter data of the workpiece that this quilt is taken a sample test and actual measured value add those and organize historical technological parameter data and those historical measurements, rebulid this and first estimate pattern and become one second and estimate pattern; Or the technological parameter data of the workpiece of taking a sample test with this quilt and actual measured value are come adjustment, and this first estimates the parameter value of pattern and becomes this and second estimate pattern;

The technological parameter data of importing all interior workpiece of this card casket under the workpiece that this quilt takes a sample test second estimate pattern to this, and recalculate a subordinate phase virtual measurement value of each workpiece in this card casket; And

Second estimate pattern and replace this and first estimate pattern with this, calculate the follow-up phase one virtual measurement value that enters the workpiece of this production board.

2. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 1 is characterized in that, this two-stage virtual measuring system also comprises a confidence index module, and this virtual measurement method also comprises at least:

This two-stage virtual measuring system is used those to organize historical technological parameter data and those historical measurement data and is set up one first reference model, and the foundation of this first reference model is according to one with reference to algorithm, and it is different with reference to algorithm with this that this estimates algorithm; And

This phase one estimates step and also comprises at least:

See through this person's technological parameter data that this two-stage virtual measuring system imports those workpiece that this production board sent to this first reference model, and calculate this person's of those workpiece one first reference prediction value; And

See through that this confidence index module is calculated the overlapping area between the distribution of this person's the distribution of this phase one virtual measurement value of those workpiece and this first reference prediction value respectively and the confidence desired value of this phase one virtual measurement value that produces this person of those workpiece, when overlapping area is healed big, then the confidence desired value is higher, represents the confidence level of this phase one virtual measurement value that corresponds to higher.

3. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 2 is characterized in that, this is to be selected from a group that is made up of a multiple regression algorithm and a neural network algorithm with reference to algorithm.

4. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 2 is characterized in that, this subordinate phase estimates step and also comprises at least:

See through the technological parameter data of the workpiece that this two-stage virtual measuring system takes a sample test this quilt and actual measured value and add those and organize historical technological parameter data and those historical measurements, rebulid this first reference model and become one second reference model; Or the technological parameter data and the actual measured value of the workpiece of taking a sample test with this quilt are come the parameter value of this first reference model of adjustment and are become this second reference model; And

See through technological parameter data that this two-stage virtual measuring system imports all interior workpiece of this card casket under the workpiece that this quilt takes a sample test to this second reference model, and recalculate one second reference prediction value of each workpiece in this card casket;

See through that this confidence index module is calculated the overlapping area between the distribution of the distribution of this subordinate phase virtual measurement value of each interior workpiece of this card casket and this second reference prediction value respectively and the confidence desired value that produces this subordinate phase virtual measurement value of each workpiece in this card casket, when overlapping area is healed big, then the confidence desired value is higher, represents the confidence level of this second virtual measurement value that corresponds to higher.

5. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 4 is characterized in that, also comprises at least:

See through this two-stage virtual measuring system and replace this first reference model, calculate the confidence desired value of the phase one virtual measurement value of the follow-up workpiece that enters this production board with this second reference model.

6. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 4, it is characterized in that, producing board when this has left unused above after the schedule time, see through the technological parameter data of the workpiece that this two-stage virtual measuring system takes a sample test this quilt and actual measured value and add those and organize historical technological parameter data and those historical measurements, train this first to estimate pattern and this first reference model and become this and second estimate pattern and this second reference model again.

7. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 1 is characterized in that, this two-stage virtual measuring system also comprises a similarity index module, and this virtual measurement method also comprises at least:

See through this two-stage virtual measuring system and use those to organize historical technological parameter, and, set up one first statistical distance pattern according to a statistics distance algorithm; And

This phase one estimates step and also comprises at least:

See through this two-stage virtual measuring system and import this person's the technological parameter data of those workpiece that this production board sent to this first statistical distance pattern; And

See through the overall similarity desired value of the pairing technological parameter data of this phase one virtual measurement value that this similarity index module calculates this person of those workpiece, wherein when the overall similarity desired value littler, represent this person's of those workpiece technological parameter data more to be similar to those and organize historical technological parameter data.

8. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 7 is characterized in that, this statistical distance algorithm is a mahalanobis distance algorithm.

9. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 7 is characterized in that, this subordinate phase estimates step and also comprises at least:

The technological parameter data that see through the workpiece that this two-stage virtual measuring system takes a sample test this quilt add those and organize historical technological parameter data, train this first statistical distance pattern again and become one second statistical distance pattern; Or the technological parameter data of the workpiece of taking a sample test with this quilt are come this first statistical distance pattern of adjustment and are become this second statistical distance pattern;

See through this two-stage virtual measuring system and import the technological parameter data of all interior workpiece of this card casket under the workpiece that this quilt takes a sample test to this second statistical distance pattern; And

See through the overall similarity desired value that this similarity index module recalculates the pairing technological parameter data of subordinate phase virtual measurement value of each workpiece in this card casket, wherein when the overall similarity desired value littler, represent the technological parameter data of each interior workpiece of this card casket under the workpiece that this quilt takes a sample test more to be similar to those and organize historical technological parameter data.

10. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 9, it is characterized in that, producing board when this has left unused above after the schedule time, the technological parameter data that see through the workpiece that this two-stage virtual measuring system takes a sample test this quilt add those and organize historical technological parameter data, train this first statistical distance pattern again and become this second statistical distance pattern.

11. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 9 is characterized in that, also comprises at least:

See through this two-stage virtual measuring system and replace this first statistical distance pattern with this second statistical distance pattern, calculate the overall similarity desired value of the pairing technological parameter data of phase one virtual measurement value of the follow-up workpiece that enters this production board, it is littler wherein to work as the overall similarity desired value, represents these follow-up technological parameter data that enter the workpiece of this production board more to be similar to those and organizes historical technological parameter data.

12. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 1 is characterized in that, this two-stage virtual measuring system comprises at least: a technological parameter data pre-processing module, and this virtual measurement method also comprises at least:

See through this technological parameter data pre-processing module and carry out a technological parameter data pre-treatment, to delete all unusual technological parameter data.

13. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 1 is characterized in that, this two-stage virtual measuring system comprises at least: a measurement data pre-processing module, and this virtual measurement method also comprises at least:

See through this measurement data pre-processing module and carry out a measurement data pre-treatment, with the exceptional value in the actual measured value that screens out the workpiece that this quilt takes a sample test.

14. virtual measurement method that utilizes the two-stage virtual measuring system, be applied to a wafer factory batch to batch control system, it is characterized in that, this wafer factory batch to run-to-run control system comprise at least one first workpiece to workpiece control system and one second workpiece to the workpiece control system, this two-stage virtual measuring system comprises at least: a technological parameter data pre-processing module and a measurement data pre-processing module, and this virtual measurement method comprises at least:

This two-stage virtual measuring system is obtained many groups of historical technological parameter data of producing board, and wherein a plurality of historical workpiece are according to should the historical technological parameter data of many groups producing;

This two-stage virtual measuring system is measured board from one and is obtained a plurality of historical measurements, and wherein those historical measurements are respectively the measured value of these a plurality of historical workpiece;

Seeing through this two-stage virtual measuring system uses those to organize historical technological parameter data and those historical measurements to set up one first and estimate pattern, this first foundation that estimates pattern is to estimate algorithm according to one, and this estimates algorithm is to be selected from a group that is made up of a multiple regression algorithm and a neural network algorithm;

This two-stage virtual measuring system is waited for the technological parameter data of a plurality of workpiece that this production board of collection is sent;

After one technological parameter data aggregation of those workpiece that this production board is sent is finished, see through this two-stage virtual measuring system and carry out a phase one immediately and estimate step, this phase one estimates step and comprises at least:

The technological parameter data of importing this person of those workpiece that this production board sent first estimate pattern to this, and calculate this person's of those workpiece a phase one virtual measurement value;

This two-stage virtual measuring system obtains an actual measured value of a workpiece of being taken a sample test from this measurement board;

This two-stage virtual measuring system obtains the technological parameter data of the workpiece that this quilt takes a sample test;

This two-stage virtual measuring system obtains the affiliated technological parameter data of blocking all workpiece in the casket of workpiece that this quilt is taken a sample test; And

During this actual measured value of the workpiece of taking a sample test when this quilt of obtaining those workpiece from this measurement board, see through this two-stage virtual measuring system and carry out a subordinate phase and estimate step, wherein this subordinate phase estimates step and comprises at least:

The technological parameter data of the workpiece that this quilt is taken a sample test and actual measured value add those and organize historical technological parameter data and those historical measurements, rebulid this and first estimate pattern and become one second and estimate pattern; Or the technological parameter data of the workpiece of taking a sample test with this quilt and actual measured value are come adjustment, and this first estimates the parameter value of pattern and becomes this and second estimate pattern;

The technological parameter data of importing all interior workpiece of this card casket under the workpiece that this quilt takes a sample test second estimate pattern to this, and recalculate a subordinate phase virtual measurement value of each workpiece in this card casket;

Provide the phase one virtual measurement value of each workpiece in this card casket to import to this first workpiece to the feedback of workpiece control system of this production board; And

Provide the subordinate phase virtual measurement value of each workpiece in this card casket to import to this second workpiece to the feedforward of workpiece control system of the board of next production technology.

15. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 14 is characterized in that, this two-stage virtual measuring system also comprises a confidence index module, and this virtual measurement method also comprises at least:

This two-stage virtual measuring system is used those to organize historical technological parameter data and those historical measurement data and is set up one first reference model, and the foundation of this first reference model is according to one with reference to algorithm, and it is different with reference to algorithm with this that this estimates algorithm; And

This phase one estimates step and also comprises at least:

See through this person's technological parameter data that this two-stage virtual measuring system imports those workpiece that this production board sent to this first reference model, and calculate this person's of those workpiece one first reference prediction value; And

See through that this confidence index module is calculated the overlapping area between the distribution of this person's the distribution of this phase one virtual measurement value of those workpiece and this first reference prediction value respectively and the confidence desired value of this phase one virtual measurement value that produces this person of those workpiece, when overlapping area is healed big, then the confidence desired value is higher, represents the confidence level of this phase one virtual measurement value that corresponds to higher.

16. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 15 is characterized in that, this is to be selected from a group that is made up of a multiple regression algorithm and a neural network algorithm with reference to algorithm.

17. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 15 is characterized in that, this subordinate phase estimates step and also comprises at least:

See through the technological parameter data of the workpiece that this two-stage virtual measuring system takes a sample test this quilt and actual measured value and add those and organize historical technological parameter data and those historical measurements, rebulid this first reference model and become one second reference model; Or the technological parameter data and the actual measured value of the workpiece of taking a sample test with this quilt are come the parameter value of this first reference model of adjustment and are become this second reference model; And

See through technological parameter data that this two-stage virtual measuring system imports all interior workpiece of this card casket under the workpiece that this quilt takes a sample test to this second reference model, and recalculate one second reference prediction value of each workpiece in this card casket;

See through that this confidence index module is calculated the overlapping area between the distribution of the distribution of this subordinate phase virtual measurement value of each interior workpiece of this card casket and this second reference prediction value respectively and the confidence desired value that produces this subordinate phase virtual measurement value of each workpiece in this card casket, when overlapping area is healed big, then the confidence desired value is higher, represents the confidence level of this second virtual measurement value that corresponds to higher.

18. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 17 is characterized in that, also comprises at least:

See through this two-stage virtual measuring system and replace this first reference model, calculate the confidence desired value of the phase one virtual measurement value of the follow-up workpiece that enters this production board with this second reference model.

19. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 17, it is characterized in that, producing board when this has left unused above after the schedule time, see through the technological parameter data of the workpiece that this two-stage virtual measuring system takes a sample test this quilt and actual measured value and add those and organize historical technological parameter data and those historical measurements, train this first to estimate pattern and this first reference model and become this and second estimate pattern and this second reference model again.

20. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 14 is characterized in that, this two-stage virtual measuring system also comprises a similarity index module, and this virtual measurement method also comprises at least:

See through this two-stage virtual measuring system and use those to organize historical technological parameter, and, set up one first statistical distance pattern according to a statistics distance algorithm; And

This phase one estimates step and also comprises at least:

See through this two-stage virtual measuring system and import this person's the technological parameter data of those workpiece that this production board sent to this first statistical distance pattern; And

See through the overall similarity desired value of the pairing technological parameter data of this phase one virtual measurement value that this similarity index module calculates this person of those workpiece, wherein when the overall similarity desired value littler, represent this person's of those workpiece technological parameter data more to be similar to those and organize historical technological parameter data.

21. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 20 is characterized in that, this statistical distance algorithm is a mahalanobis distance algorithm.

22. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 20 is characterized in that, this subordinate phase estimates step and also comprises at least:

The technological parameter data that see through the workpiece that this two-stage virtual measuring system takes a sample test this quilt add those and organize historical technological parameter data, train this first statistical distance pattern again and become one second statistical distance pattern; Or the technological parameter data of the workpiece of taking a sample test with this quilt are come this first statistical distance pattern of adjustment and are become this second statistical distance pattern;

See through this two-stage virtual measuring system and import the technological parameter data of all interior workpiece of this card casket under the workpiece that this quilt takes a sample test to this second statistical distance pattern; And

See through the overall similarity desired value that this similarity index module recalculates the pairing technological parameter data of subordinate phase virtual measurement value of each workpiece in this card casket, wherein when the overall similarity desired value littler, represent the technological parameter data of each interior workpiece of this card casket under the workpiece that this quilt takes a sample test more to be similar to those and organize historical technological parameter data.

23. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 22, it is characterized in that, producing board when this has left unused above after the schedule time, the technological parameter data that see through the workpiece that this two-stage virtual measuring system takes a sample test this quilt add those and organize historical technological parameter data, train this first statistical distance pattern again and become this second statistical distance pattern.

24. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 22 is characterized in that, also comprises at least:

See through this two-stage virtual measuring system and replace this first statistical distance pattern with this second statistical distance pattern, calculate the overall similarity desired value of the pairing technological parameter data of phase one virtual measurement value of the follow-up workpiece that enters this production board, it is littler wherein to work as the overall similarity desired value, represents these follow-up technological parameter data that enter the workpiece of this production board more to be similar to those and organizes historical technological parameter data.

25. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 14, this two-stage virtual measuring system comprises at least: a technological parameter data pre-processing module, and this virtual measurement method is characterized in that, also comprises at least:

See through this technological parameter data pre-processing module and carry out a technological parameter data pre-treatment, to delete all unusual technological parameter data.

26. the virtual measurement method in according to claim 14 pair of stage is characterized in that, this two-stage virtual measuring system comprises at least: a measurement data pre-processing module, and this virtual measurement method also comprises at least:

See through this measurement data pre-processing module and carry out a measurement data pre-treatment, with the exceptional value in the actual measured value that screens out the workpiece that this quilt takes a sample test.

27. virtual measurement method that utilizes the two-stage virtual measuring system, be applied to a wafer factory batch to batch control system, it is characterized in that, this wafer factory batch to run-to-run control system comprise at least a first consignment to unit of cargo control system and one second unit of cargo to the unit of cargo control system, comprise at least:

This two-stage virtual measuring system is obtained many groups of historical technological parameter data of producing board, and wherein a plurality of historical workpiece are according to should the historical technological parameter data of many groups producing;

This two-stage virtual measuring system is measured board from one and is obtained a plurality of historical measurements, and wherein those historical measurements are respectively the measured value of these a plurality of historical workpiece;

Seeing through this two-stage virtual measuring system uses those to organize historical technological parameter data and those historical measurements to set up one first and estimate pattern, this first foundation that estimates pattern is to estimate algorithm according to one, and this estimates algorithm is to be selected from a group that is made up of a multiple regression algorithm and a neural network algorithm;

This two-stage virtual measuring system is waited for the technological parameter data of a plurality of workpiece that this production board of collection is sent;

After one technological parameter data aggregation of those workpiece that this production board is sent is finished, see through this two-stage virtual measuring system and carry out a phase one immediately and estimate step, this phase one estimates step and comprises at least:

The technological parameter data of importing this person of those workpiece that this production board sent first estimate pattern to this, and calculate this person's of those workpiece a phase one virtual measurement value;

This two-stage virtual measuring system obtains an actual measured value of a workpiece of being taken a sample test from this measurement board;

This two-stage virtual measuring system obtains the technological parameter data of the workpiece that this quilt takes a sample test;

This two-stage virtual measuring system obtains the affiliated technological parameter data of blocking all workpiece in the casket of workpiece that this quilt is taken a sample test; And

During this actual measured value of the workpiece of taking a sample test when this quilt of obtaining those workpiece from this measurement board, see through this two-stage virtual measuring system and carry out a subordinate phase and estimate step, wherein this subordinate phase estimates step and comprises at least:

The technological parameter data of the workpiece that this quilt is taken a sample test and actual measured value add those and organize historical technological parameter data and those historical measurements, rebulid this and first estimate pattern and become one second and estimate pattern; Or the technological parameter data of the workpiece of taking a sample test with this quilt and actual measured value are come adjustment, and this first estimates the parameter value of pattern and becomes this and second estimate pattern;

The technological parameter data of importing all interior workpiece of this card casket under the workpiece that this quilt takes a sample test second estimate pattern to this, and recalculate a subordinate phase virtual measurement value of each workpiece in this card casket;

Provide the subordinate phase virtual measurement value of each workpiece in this card casket to import to this first consignment to the feedback of unit of cargo control system of this production board; And

Provide the subordinate phase virtual measurement value of each workpiece in this card casket to import to this second unit of cargo to the feedforward of unit of cargo control system of the board of next production technology.

28. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 27 is characterized in that, this two-stage virtual measuring system also comprises a confidence index module, and this virtual measurement method also comprises at least:

This two-stage virtual measuring system is used those to organize historical technological parameter data and those historical measurement data and is set up one first reference model, and the foundation of this first reference model is according to one with reference to algorithm, and it is different with reference to algorithm with this that this estimates algorithm; And

This phase one estimates step and also comprises at least:

See through this person's technological parameter data that this two-stage virtual measuring system imports those workpiece that this production board sent to this first reference model, and calculate this person's of those workpiece one first reference prediction value; And

See through that this confidence index module is calculated the overlapping area between the distribution of this person's the distribution of this phase one virtual measurement value of those workpiece and this first reference prediction value respectively and the confidence desired value of this phase one virtual measurement value that produces this person of those workpiece, when overlapping area is healed big, then the confidence desired value is higher, represents the confidence level of this phase one virtual measurement value that corresponds to higher.

29. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 28 is characterized in that, this is to be selected from a group that is made up of a multiple regression algorithm and a neural network algorithm with reference to algorithm.

30. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 28 is characterized in that, this subordinate phase estimates step and also comprises at least:

See through the technological parameter data of the workpiece that this two-stage virtual measuring system takes a sample test this quilt and actual measured value and add those and organize historical technological parameter data and those historical measurements, rebulid this first reference model and become one second reference model; Or the technological parameter data and the actual measured value of the workpiece of taking a sample test with this quilt are come the parameter value of this first reference model of adjustment and are become this second reference model; And

See through technological parameter data that this two-stage virtual measuring system imports all interior workpiece of this card casket under the workpiece that this quilt takes a sample test to this second reference model, and recalculate one second reference prediction value of each workpiece in this card casket;

See through that this confidence index module is calculated the overlapping area between the distribution of the distribution of this subordinate phase virtual measurement value of each interior workpiece of this card casket and this second reference prediction value respectively and the confidence desired value that produces this subordinate phase virtual measurement value of each workpiece in this card casket, when overlapping area is healed big, then the confidence desired value is higher, represents the confidence level of this second virtual measurement value that corresponds to higher.

31. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 30 is characterized in that, also comprises at least:

See through this two-stage virtual measuring system and replace this first reference model, calculate the confidence desired value of the phase one virtual measurement value of the follow-up workpiece that enters this production board with this second reference model.

32. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 30, it is characterized in that, producing board when this has left unused above after the schedule time, see through the technological parameter data of the workpiece that this two-stage virtual measuring system takes a sample test this quilt and actual measured value and add those and organize historical technological parameter data and those historical measurements, train this first to estimate pattern and this first reference model and become this and second estimate pattern and this second reference model again.

33. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 27 is characterized in that, this two-stage virtual measuring system also comprises a similarity index module, and this virtual measurement method also comprises at least:

See through this two-stage virtual measuring system and use those to organize historical technological parameter, and, set up one first statistical distance pattern according to a statistics distance algorithm; And

This phase one estimates step and also comprises at least:

See through this two-stage virtual measuring system and import this person's the technological parameter data of those workpiece that this production board sent to this first statistical distance pattern; And

See through the overall similarity desired value of the pairing technological parameter data of this phase one virtual measurement value that this similarity index module calculates this person of those workpiece, wherein when the overall similarity desired value littler, represent this person's of those workpiece technological parameter data more to be similar to those and organize historical technological parameter data.

34. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 33 is characterized in that, this statistical distance algorithm is a mahalanobis distance algorithm.

35. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 33 is characterized in that, this subordinate phase estimates step and also comprises at least:

The technological parameter data that see through the workpiece that this two-stage virtual measuring system takes a sample test this quilt add those and organize historical technological parameter data, train this first statistical distance pattern again and become one second statistical distance pattern; Or the technological parameter data of the workpiece of taking a sample test with this quilt are come this first statistical distance pattern of adjustment and are become this second statistical distance pattern;

See through this two-stage virtual measuring system and import the technological parameter data of all interior workpiece of this card casket under the workpiece that this quilt takes a sample test to this second statistical distance pattern; And

See through the overall similarity desired value that this similarity index module recalculates the pairing technological parameter data of subordinate phase virtual measurement value of each workpiece in this card casket, wherein when the overall similarity desired value littler, represent the technological parameter data of each interior workpiece of this card casket under the workpiece that this quilt takes a sample test more to be similar to those and organize historical technological parameter data.

36. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 35, it is characterized in that, producing board when this has left unused above after the schedule time, the technological parameter data that see through the workpiece that this two-stage virtual measuring system takes a sample test this quilt add those and organize historical technological parameter data, train this first statistical distance pattern again and become this second statistical distance pattern.

37. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 35 is characterized in that, also comprises at least:

See through this two-stage virtual measuring system and replace this first statistical distance pattern with this second statistical distance pattern, calculate the overall similarity desired value of the pairing technological parameter data of phase one virtual measurement value of the follow-up workpiece that enters this production board, it is littler wherein to work as the overall similarity desired value, represents these follow-up technological parameter data that enter the workpiece of this production board more to be similar to those and organizes historical technological parameter data.

38. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 27 is characterized in that, this two-stage virtual measuring system comprises at least: a technological parameter data pre-processing module, and this virtual measurement method also comprises at least:

See through this technological parameter data pre-processing module and carry out a technological parameter data pre-treatment, to delete all unusual technological parameter data.

39. the virtual measurement method that utilizes the two-stage virtual measuring system according to claim 27 is characterized in that, this two-stage virtual measuring system comprises at least: a measurement data pre-processing module, and this virtual measurement method also comprises at least:

See through this measurement data pre-processing module and carry out a measurement data pre-treatment, with the exceptional value in the actual measured value that screens out the workpiece that this quilt takes a sample test.

Images