CN101438251A - Adaptive multivariate fault detection - Google Patents
Adaptive multivariate fault detection Download PDFInfo
- Publication number
- CN101438251A CN101438251A CNA2007800165002A CN200780016500A CN101438251A CN 101438251 A CN101438251 A CN 101438251A CN A2007800165002 A CNA2007800165002 A CN A2007800165002A CN 200780016500 A CN200780016500 A CN 200780016500A CN 101438251 A CN101438251 A CN 101438251A
- Authority
- CN
- China
- Prior art keywords
- model
- data
- multivariate statistics
- deal
- adjusting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims description 46
- 230000003044 adaptive effect Effects 0.000 title description 7
- 238000000034 method Methods 0.000 claims abstract description 109
- 230000008569 process Effects 0.000 claims abstract description 33
- 238000000491 multivariate analysis Methods 0.000 claims description 85
- 238000004519 manufacturing process Methods 0.000 claims description 41
- 238000012549 training Methods 0.000 claims description 28
- 238000004422 calculation algorithm Methods 0.000 claims description 26
- 238000012423 maintenance Methods 0.000 claims description 17
- 239000013598 vector Substances 0.000 claims description 11
- 230000008859 change Effects 0.000 abstract description 15
- 238000013179 statistical model Methods 0.000 abstract description 7
- 238000012545 processing Methods 0.000 description 50
- 238000003860 storage Methods 0.000 description 22
- 238000005259 measurement Methods 0.000 description 14
- 230000015654 memory Effects 0.000 description 14
- 238000012544 monitoring process Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 7
- 238000000513 principal component analysis Methods 0.000 description 7
- 239000004065 semiconductor Substances 0.000 description 6
- 230000003068 static effect Effects 0.000 description 5
- BLRPTPMANUNPDV-UHFFFAOYSA-N Silane Chemical compound [SiH4] BLRPTPMANUNPDV-UHFFFAOYSA-N 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 229910000077 silane Inorganic materials 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000009434 installation Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000005229 chemical vapour deposition Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005530 etching Methods 0.000 description 2
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 230000003449 preventive effect Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 238000003070 Statistical process control Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 239000003153 chemical reaction reagent Substances 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 238000002347 injection Methods 0.000 description 1
- 239000007924 injection Substances 0.000 description 1
- 238000009342 intercropping Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000000206 photolithography Methods 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 238000004886 process control Methods 0.000 description 1
- NHDHVHZZCFYRSB-UHFFFAOYSA-N pyriproxyfen Chemical compound C=1C=CC=NC=1OC(C)COC(C=C1)=CC=C1OC1=CC=CC=C1 NHDHVHZZCFYRSB-UHFFFAOYSA-N 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 230000004083 survival effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 235000012431 wafers Nutrition 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- General Factory Administration (AREA)
Abstract
A method and apparatus for detecting faults. A set of data samples is received, the set of data samples including multiple process variables. One or more multivariate statistical models are adapted, wherein adapting includes applying a change to at least one univariate statistic of the one or more multivariate statistical models if the change is greater than a threshold value. The one or more multivariate statistical models are used to analyze subsequent process data to detect faults.
Description
Related application
The right of priority of the provisional application that the application requires to submit on May 7th, 2006 provisional application is submitted to number on May 7th, 60/746649 and 2006 number 60/746647.
Technical field
Specific embodiments of the invention are about error-detecting, especially about using the error-detecting of multiple error characteristic.
Background technology
The utilization of many enterprises comprises the accurate manufacturing equipment of multiple sensor and controller, and these sensors and controller are carefully monitored to guarantee the quality of product during handling.A kind of method of monitoring these multiple sensors and controller is statistical treatment monitoring (a kind of means of carrying out the statistical study on sensor measurement and processing controls numerical value (treatment variable)), its activation Auto-Sensing and/or error-detecting.One " mistake (fault) " can be the fault of manufacturing equipment or imbalance (for example with the error of operating parameter of a machine of desire numerical value), or is that a prevention is kept a required indication and avoided an imminent fault or an imbalance.Therefore, a target of statistical treatment monitoring is detecting and/or a detection mistake before producing above-mentioned defective.
Handling monitoring period, when one or more statistics of nearest deal with data departs from an amount from a statistics model, and this amount is detected a mistake enough greatly when causing a model measurement not trust threshold value above one.One model measurement is a scale number, the bias of its value representation between the statistical nature that statistical nature and this model of the collected deal with data of actual treatment monitoring period are predicted.Each model measurement is this unique mathematical method that departs from of cancellation.Common model measurement comprises square prediction error (Squared Prediction Error, its general denotion is SPE, Qres or Q), and Hotelling ' s T2.
Each model measurement has indivedual trust threshold values, and it is also censured is a trust restriction or control restriction, wherein one of this model measurement of the numeric representation acceptable upper limit.If a model measurement surpasses its indivedual trust threshold values at the processing monitoring period, answer this deal with data of deducibility to depart from threshold value because of a mistake.
Accurately the fact of an obstacle of error detection is handled general drift overtime for making, even if under the situation without any problem.For example, this operational circumstances in the semiconductor process chamber generally drifts about between the continuous replacement of the chamber component of the continuous cleaning room of this chamber and consumption.The common statistical treatment method for supervising of error detection is subjected to differentiating the shortcoming of a normal drift and a mistake.
In particular, some error detection method is used static model, and its hypothesis disposition remains unchanged in the survival of an instrument.A model like this can not be in time the change of expectation and a mistake differentiate between unexpected departing from of being caused.Trigger many false alarms for avoiding handling drift, this control restriction must be set to the width that is large enough to hold drift.Therefore, this model can't be detected trickle mistake.
Gallagher, Neal B. etc. " development and the evaluation of the multivariate statistics processing controls instrument that conductor etching is handled: promote strong by model modification; Development and benchmarking ofmultivariate statistical process control tools for a semiconductor etch process:improving robustness through model updating ", ADCHEM 1997, Banff, Canada; And Li, Weihua etc. " suitably handling the recurrence PCA of monitoring; Recursive PCA for adaptiveprocess monitoring ", J.Process Control, the 10th, 471-486 page or leaf (2000), its each description is via adjusting a model responds the drift in this treatment situation to the drift in the deal with data method periodically.This Gallagher delivers to describe and adjusts means (adaptation of mean) and oblique variance statistics (covariance statistics).Surpass one and trust restriction if measure at a Q of a module or T2, the Gallagher attempt is differentiated between mistake and normal drift via the generation of identification one mistake.This Li delivers the number of principal components in the means of adjusting of describing, oblique variance, major component matrix (Principal Component Matrix) and principal component analysis (PCA) (PCA) model.Detect the mistake that takes place gradually by the adapting method that Gallagher and Li advised.
Spitzlsperger, the Tokyo of Gerhard etc. (2004), " using the error detection of the through hole etch processes of adaptive multivariate method; Fault detection for a via etch process usingadaptive multivariate methods " of ISSM, it discloses making of human special knowledge and is used for adjusting only single argument means and the gage number system number (scaling coefficient) that is drifted about by expectation.Yet via adjusting only single argument means and gage coefficient, the method can't provide adjusting of these covariances between each variable in the model.
Each common adapting method described above is subject to the influence of the truncation error (cumulativecomputational rounding error) of cumulative calculation, and it is adjusted institute by this cycle and causes.This then causes this model to have coarse statistical value, its can make the mistake alarm and fault both detect mistake.
Summary of the invention
An aspect of of the present present invention is about a kind of method of detecting mistake, it comprises the reception deal with data, this deal with data comprises a plurality of treatment variables, adjust one or more multivariate statistics model according to this deal with data, the process of wherein adjusting comprises and applies an at least one univariate statistics that changes to this one or more multivariate statistics model, if it in the time should changing greater than a threshold value and for it, and the deal with data of using this one or more multivariate statistics model through adjusting to analyze to continue is with the detecting mistake.
But another aspect of the present invention is about a kind of machine access medium that comprise data, it is when by the access of a machine institute, cause this machine to carry out a method, this method comprises the reception deal with data, this deal with data comprises a plurality of treatment variables, and adjust one or more multivariate statistics model according to this deal with data, the process of wherein adjusting comprises and applies an at least one univariate statistics that changes to this one or more multivariate statistics model, if it in the time should changing greater than a threshold value and for it, and the deal with data of using this one or more multivariate statistics model through adjusting to analyze to continue is with the detecting mistake.
Of the present invention again on the other hand about a kind of statistical treatment supervisory system, it comprises a database, it is used to store one or more multivariate statistics model, an and error detection device, its be coupled at least one manufacturing machine and this database, this error detection device is in order to receive the deal with data from this at least one manufacturing machine, wherein this deal with data comprises a plurality of treatment variables, with adjust this one or at least one of multivariate statistics model, the process of wherein adjusting comprises and applies an at least one univariate statistics that changes to this one or more multivariate statistics model, if it is in the time should changing greater than a threshold value and for it, and in order to the deal with data using this one or more multivariate statistics model to analyze to continue through adjusting with the detecting mistake.
Description of drawings
Fig. 1 descriptive statistics is handled the specific embodiment of monitoring system;
Fig. 2 describes a kind of process flow diagram of detecting a specific embodiment of wrong method via the multivariate of adjusting one or more;
Fig. 3 describes a kind of process flow diagram of detecting a specific embodiment of wrong method;
Fig. 4 describes a kind of at machine maintenance after by resetting one or multiple statistics model and detect the process flow diagram of a specific embodiment of wrong method;
Fig. 5 describes the graphic representation of the machine in the computing system of an exemplary form, wherein has any one or more method that one group of instruction can be carried out and be discussed in order to cause this machine to carry out in this.
The primary clustering symbol description:
110 make machine
155 sensors
170 prescriptions
150 processing controllers
160 data communication links
125 error detection devices
130 error detectors
165 wrong report devices
120 handle measured database
135 multivariate statistics models
140 error characteristics
145 mis-classifications
175 storage devices
105 statistical treatment watch-dogs
210 receive deal with data
Whether 215 decision deal with data point out faults
Are 220 drifts at treatment variable and measured?
Does 225 are predetermined the interval reach?
230 adaptive multivariate statistical models
Will do 235 adjust and change at least one threshold value of univariate statistics?
240 change univariate statistics
245 uses are added up through adaptive multivariate and are analyzed the deal with data that continues
310 receive deal with data
315 apply the first error detection algorithm
Are 320 mistakes pointed out?
325 adaptive multivariate statistical models
330 apply the second error detection algorithm
335 report mistakes
400 detecting machine maintenances
410 automatic replacement multivariate statistics models
415 reply the multivariate statistics model to the state that meets a virgin state
420 adaptive multivariate statistical models are to new operational circumstances
425 initialization are reset the training period
502 processors
526 processing logics
504 primary memorys
522 softwares
506 static memories
508 Network Interface Units
520 networks
510 videos show
512 letters and digital input media
514 finger control devices
516 signal generation devices
518 supplementary storages
But 531 machine access Storage Medias
522 softwares
Embodiment
Be described as a kind of wrong method and apparatus that is used to detect in this.In one embodiment, receive the deal with data that comprises plural treatment variable.The example of treatment variable comprises temperature, pressure silane flow (silane flow) etc.One or more multivariate statistics model is adjusted according to this deal with data.Can comprise one and change and can not surpass a threshold value if adjust, apply this at least one unitary variant that changes to this multivariate statistics module and add up.In one embodiment, carry out on the interval based on being predetermined of one or more treatment variable and adjust once measuring drift.This multivariate statistics model through adjusting can then be used to the continue deal with data of dissecting needle to the detecting of each mistake.
In following description, a plurality of details are proposed.Yet, be familiar with this skill person and can understand that the present invention can be in no following specific detail and implemented.In specific illustration, known structure and device show (but not details) by block diagram form, use and avoid fuzzy the present invention.
Described part details with algorithm and in a computer memory representative symbol symbol of calculation data bit show.The skill person who has the knack of data processing technique uses the narration of these algorithms and presents in the mode of full blast and passes on essence to know this skill person to other.Algorithm can be considered guiding one result's that wants the step of self or the program of instruction herein usually.These steps need have the physical property operator for physical quantity for those.Though be not inevitable, that the common employing of this tittle can store in a computer system, transmit, makes up, relatively reaches or otherwise operate is electric, the form of magnetic signal.Confirmed these signal indications are position, numerical value, assembly, symbol, character, term, digital or the like sometimes very convenient, mainly based on the factor of common usage.
Yet, it should be noted, these and similarly term is all relevant with suitable physical magnitude, and only be the label of applying mechanically to this tittle that makes things convenient for.Unless stated otherwise, otherwise can know under discussion and learn, utilizing term in the literary composition similarly is " processing ", " computing ", " calculating " or " decision " or " demonstration " or the like, represent a computer system or the similarly action and the processing of electronic operation device, its manipulation and conversion are expressed as physics (electronics) amount in the buffer and the data in the storer of this computer system, become in the storer of this computer system or the physical quantity in buffer or other this information storage, transmission or the display device.
The present invention is also relevant for a device of carrying out computing described herein.This device can be in response to demand to be formed, a common computer that perhaps also can be the computer program that is stored in this computing machine selected property activation or reset.This computer program can be stored in a computer-readable media, for example (but being not limited to), the disk of any kind of, the medium that it comprises floppy diskette, CD, ROM (read-only memory) (CD-ROMs) and Magneto Optical MO, ROM (read-only memory) (ROMs), random access memory (RAMs), can eliminate program read-only memory (EPROMs), electronic type can be eliminated program read-only memory (EEPROMs), magnetic or optical card or any kind of, it is applicable to that stored electrons instruction and each medium all are furnished with a computer system bus.
Algorithm as herein described and module are not relevant to any certain computer or other device.Various general service system can be according to teachings of the present invention and program parallelization, and perhaps susceptible of proof helps the how special equipment of construction to implement the step of this required method.The required framework of these various systems will be in hereinafter describing in detail.In addition, the present invention describes by any specific program language.Should know that various program languages can implement announcement of the present invention as herein described.
One machine-readable medium is included in one can be by any mechanism that is used to store or transmit information in the form that machine read.For example, a machine-readable medium comprises machine readable storage media (for example ROM (read-only memory) (ROM), random-access memory (ram), disc storage medium, optical storage media, flash memory device or the like), machine readable transmission medium (but transmitting signal (for example carrier wave, infrared signal, digital signal or the like) of electric, optics, message or other form) etc.
The details of supervisory system is handled in the processing that this following description provides monitoring to operate on manufacturing installation with a statistics of detecting and/or detection wrong (unsettled manufacture process).In one embodiment, this statistical treatment supervisory system is used for the manufacturing of electronic installation (for example semiconductor).Make such device and generally need many manufacturing steps that dissimilar manufacturings are handled that relate to.For example, etching, sputter, chemical vapor deposition are three kinds of dissimilar processing, and each person carries out on dissimilar machines.Person in addition, this statistical treatment supervisory system can be used to monitor the manufacturing (for example automobile) of other products.The manufacturing of this other products also needs many different treatment steps by various manufacturing machine processing.
Fig. 1 descriptive statistics is handled a specific embodiment of supervisory system 100.This statistical treatment supervisory system 100 comprises a statistics and handles watch-dog 105, and it makes machine 110 via data communication links 160 and one or more and one or more processing controller 150 is coupled.This statistical treatment supervisory system 100 can be included in all manufacturing machines 110 of (for example manufacturing works) in the factory.Person in addition, this statistical treatment supervisory system 100 can comprise the only specific manufacturing machine 110 in the factory, and all that for example can move on one or more particular procedure are made machines 110.
In one embodiment, each makes machine 110 for making the machine of electronic installation, and for example etcher, chemical vapor deposition stove, micro-photographing process device (photolithography devices), cloth are planted machine (implanter) or the like.Person in addition, this manufacturings machine 110 can be a type of manufacturing other products (for example automobile).In one embodiment, each person of this manufacturing machine 110 can be a single type.Person in addition, this manufacturing machine 110 can comprise the outfit of number of different types, and each person of these outfits can carry out different disposal.
Each makes machine 110 can comprise the multiple sensor that is used to monitor operation on this manufacturing machine 110.The sensor that is included in the type in this manufacturing machine 110 can be a temperature sensor.The example of other sensor comprises the sensor that pressure transducer, flow rate sensor or any physical attribute or that other monitors a service part of these manufacturing machine 110 manufacturings are made the physical conditions of handling.
Making each that carry out on the machine 110 one makes and handles the attribute detected by various physical conditions and this sensor and various through collecting with association as the operating parameter of deal with data and characterization.Physical conditions that each is clear and definite or the attribute of detecting by this sensor, and each operating parameter can be the distinctive treatment variable of a tool of this deal with data.Each example of the treatment variable of expression detector data comprises chamber pressure, receiver (susceptor) temperature, RF forward power and RF reflective power.The example of treatment variable of expression operating parameter comprises that (for example chemical reagent) flow rate is set and (at a process chamber vacuum pump for vent Pu (chamber exhaust vacuum pump)) throttling valve is set.This sensor, manufacturing machine and processing controller can be monitored in time to collect this treatment variable on continuity point during handling.
In one embodiment, each treatment variable is applied to a particular procedure.Person in addition, one or more treatment variable can be applied to the only part of a particular procedure.In one embodiment, the sensor measurement and the operating parameter of the different step in handles are represented distinguishing treatment variable (being modeled as the extra scale in the modular space).For example, if the manufacturing processing that is executed in the machine has the multiple step that contains the different operating parameter setting, this will be for useful.For example, in one or three step manufacturings are handled, the receiver temperature during three steps will be regarded as three distinctive treatment variables of tool.Indivedual scales of these treatment steps being divided into the modular space are useful, it is for example when a single processing deposits multiple layer on a service part, or when a different step of handling exposes this service part to the different disposal situation to the open air (for example pressure, temperature or the like).
The operating parameter of machine 110 is made in processing controller 150 controls.For example, the chamber temperature of processing controller may command manufacturing machine 110, vacuum pump, gas injection system or the like.Processing controller 150 can store one or more process recipe 160.The operating parameter of the manufacturing machine 110 on each step that each 160 definition one of filling a prescription is handled.In one embodiment, prescription 160 can be loaded via processing controller 150 and make machine 110.
Data communication links 160 can comprise common communication linkage, and it also can be wireless or wired.Data can be made machine 110, this processing controller 150 and the 105 intercropping transmission of this statistical treatment watch-dog at this by pure (raw) or treated form.In one embodiment, can use semiconductor devices communicating standard (SECS) interface.In other specific embodiment, can use the traffic model of a common name, a high speed SECS message to serve (HSMS) interface or the like.
This statistical treatment watch-dog 105 can be a single server, and it is used to analyze the deal with data that enters from this manufacturing machine 110, sensor 155 and processing controller 150.Person in addition, this statistical treatment watch-dog 105 can comprise multiple server and/or computing machine.In one embodiment, this statistical treatment watch-dog 105 comprises error detection device 125, error detector 130 and wrong report device 150.This statistical treatment watch-dog 105 also comprises storage device 175.In one embodiment, this statistical treatment watch-dog 105 is comprised in one or more processing controller 150.Person in addition, this statistical treatment watch-dog 105 also can be a property distinguished and/or separate equipment.
This storage device 175 comprises one and handles measured database 120, one or more multivariate statistics model 135, error characteristic 140 and mis-classification 145.In one embodiment, this storage device 175 computing machine that is this statistical treatment watch-dog 105 or a single storage device of server.Person in addition, this storage device 175 can be outside in this statistical treatment watch-dog 105.In one embodiment, this storage device 175 comprises multiple storage device, and the specific person of these storage devices comprises the redundant copy that is used for backed up data.
Handle measurement data (deal with data) and can be stored in processing measured database 120.This deal with data through storing can be used to show at these makes each person of machines 110 and at drift and the tendency of making processing of moving on machines or the like at these.In one embodiment, this is through storing the multivariate statistics model 135 that deal with data is used for producing as described below one or more.In case through producing, this multivariate statistics model 135 can be stored in the storage device 175.
In one embodiment, use a training period to collect the data that produce one or more multivariate statistics model.This training period comprises the collection of respectively handling operation that the specific manufacturing known and/or that finished on the specific manufacturing machine is handled under the control situation.Can be used to produce statistics (number (mean), variable, variable array or the like for example) in the training period from handling the collected deal with data of operation.These statistics collection property ground are used to produce general at one or more the changeable statistical model that operates in the particular procedure on the particular machine.
Each multivariate statistics model 135 comprises one or more model metering.The model metering is scalar values, the side-play amount between one a group of deal with data of its characterization and the model.In one embodiment, this model metering comprises square prediction error (Squared Prediction Error, its general denotion is SPE, Qres or Q), and Hotelling ' s T2.The model metering also comprises multiple measurement (for example combined type multivariate index (CMI)).Each person of these measurements is corresponding to estimating that monitored deal with data has the distinct methods as the probability of the identical statistics of the training data that is used for setting up this model.Above-mentioned statistics and measurement can be calculated according to general statistic algorithm.
One or more multivariate model can utilize critical piece analysis (PCA) to change a M-dimension treatment variable space N-dimensional space of vertical critical piece extremely each other, and wherein M is the number of treatment variable, and N is smaller than M.PCA calculates one group of M proper vector (M eigenvectors) and M character numerical value (eigenvalues), wherein each other proper vector conversion process variable data is to an other dimension in this critical piece space, and each character numerical value often becomes than in the represented variable of an individual features numerical value.In order to simplify this critical piece space (reducing the dimension in this critical piece space), be maintained in this model corresponding to this N proper vector of the maximum character numerical value of this N: this further feature vector is abandoned or is ignored.Remain on the parameter of the number N of the critical piece in this model for selected this model of user.The number of this critical piece (N) can be exceeded transaction between the model of appointment based at a model of explaining less data variation when using a less numerical value of N and as use one bigger numerical N the time and selected.
In case one or more multivariate statistics model produces, they can be made by error detection device 125 to be used for monitoring the processing that is moved on the machine 110 making.Error detection device 125 is analyzing and processing data via carrying out various statistical treatment method for supervising, and each person of these methods is based at least one statistics of variable model.In one embodiment, error detection device 125 directly receives the processing measurement data (deal with data) of making machine 110, sensor 155 and/or processing controller 150 from this.In another specific embodiment, error detection device 125 can receive from the deal with data of handling measured database 120.In another specific embodiment again, this error detection device 125 receives the deal with data in both sources from this.
In order to detect mistake, 125 calculating of error detection device are at each statistics of each monitored deal with data of handling, and relatively this statistics as calculated and the suitably corresponding statistics of multivariate statistics model.This statistics compares at model metering or at multiple model metering (for example T2, SPE, CMI).If one or more this model metering surpasses a predefined threshold value (denotion is a trust restriction or control restriction), can detect a mistake.In one embodiment, each model metering has the threshold numerical value of selecting for the user.The threshold numerical value of this selection can be represented the risk (if this threshold value is too low) of a false alarm and can't detect compromise between the risk (if this threshold value is too high) of a mistake.Wherein multiple metering is calculated, if any one that should measure surpasses threshold numerical value, then causes mistake.Person in addition is if surpass threshold numerical value and can point out particular error if only specific metering surpasses threshold numerical value or only multiple metering.
In case a mistake is discerned by this error detection device 125, analyzes this mistake by error detector.Error detector 130 relatively should mistake and collecting of error characteristic.Each error characteristic represents to represent the disposition of one (respectively) particular error.In one embodiment, error characteristic 140 is for having tabulating through arranging to each treatment variable of the bigger statistics contribution of one (respectively) particular error.Error-detecting 130 can compare each through store error characteristic with have at each treatment variable of the maximum contribution of present mistake through arranging tabulation.When the high-level similarity between one and this present mistake of having these error characteristics 140, report out one and meet.
Each error characteristic 140 is associated in one or more mis-classification 145.This mis-classification 145 can point out to produce a mistake a practical problems or should be wrong at present possible make because of.For example, be the silane flow rate if this error characteristic is pointed out this maximum contribution treatment variable, this mis-classification can be pointed out to feed, and to enter a numerical value of a process chamber not normal for silane.
Mistake report 165 produces the corresponding mistake report to a present mistake of what person of misdirection classification 145.This mistake report can be transferred into one or the multi-client (show, and for example local computer, remote computer, personal digital assistant (PDAs), calling set, mobile phone or the like) that is connected to this statistical treatment watch-dog 105 by net.Mistake report 165 also can make to be made machine 110 shutdown, warning one machine or carries out other suitable action.
Even if the shortage mistake is made and handled usually drift in time, for example, this operational circumstances in the semiconductor process chamber is generally drifting about between this process chamber of cleaning and between the continuous replacement of the process chamber components that exhausts continuously.Handle drift via adjusting, the change in the treatment variable that drift is caused can literal not translated mistakenly and is each mistake.
Fig. 2 describes a concrete process flow diagram of implementing profit of detecting wrong method via adjusting one or more multivariate statistics model.This method can be carried out by processing logic, and this processing logic can comprise hardware (for example circuit, special logic, FPGA (Field Programmable Gate Array), microcode or the like), software (instruction that for example moves) or above combination on treating apparatus.In one embodiment, method 200 can be carried out by the statistical treatment watch-dog 105 of Fig. 1.
With reference to Fig. 2, method 200 starts from order to receive the processing logic (square 210) of deal with data.This deal with data can be comfortable one to make the processing that moves on the machine, and can comprise multiple treatment variable.At square 215, analyze this deal with data to determine whether a mistake is pointed out at one or more multivariate statistics model.In this described specific embodiment, before carrying out any adjusting, analyze this deal with data to detect a mistake.Person in addition can be after execution be adjusted, analyzing and processing data.When deal with data pointed out that a threshold value of one or many meterings (for example T2, SPE, CMI or the like) of this multivariate statistics model has surpassed, a mistake was pointed out at one of this multivariate statistics model person.
In one embodiment, two or above multivariate statistics model use at error detection at present.If at least one of these models detected a mistake, can therefore discern a mistake.Even if do not identify mistake, report also can be transferred into a user, if for example a model is detected a mistake and alternate model does not have.Person in addition, unless at least two models are pointed out a possible mistake, otherwise a mistake will can not reported.
In one embodiment, two or above multivariate statistics model at least one mode, make differentiation with another person.For example, can or use the deal with data of difference amount to keep this model and distinguish each model via utilization different disposal variable, the different numbers that use critical piece, the different trust restrictions of use.For example, one first model comprises all deal with data of handling at one, one second model comprises because a up-to-date preventive maintenance (preventive maintenance, all deal with data that PM) produced, and one the 3rd model can comprise only last 1000 wafers.Model also can be adjusted the method for drifting about and distinguishes (for example what person's treatment variable is adjusted, what person's statistics which kind of is adjusted and uses adjust threshold or the like) via utilization is different, and it has further with reference to square 220 explains.In one embodiment, at least one multivariate statistics model is not adjusted with drift, and at least one multivariate statistics model is really adjusted with drift.The multivariate statistics model also can as above change in the NM mode of institute.
Each person of this multivariate statistics model can use one single group training data one the training period during and side by side be established.Person in addition, different models can use different training datas or extra training data.For example, if a model comprises more treatment variable compared to other person, if a statistics model need incorporate into hang oneself design experiment or from one the excessive data of long training period comprising additional operation standard, but above-mentionedly will be the anticipator.
Referring now to Fig. 2, in square 220, whether processing logic decision drift is at one or multiprocessing variable and measured.If the certain statistical of a treatment variable (number, standard deviation etc. for example) is regulated gradually, measure drift.If measure not drift, this method finishes.If at one or more treatment variable detecting drift, this method advances to square 225.In another specific embodiment, this method advance to square 225 and no matter the drift whether measured.
In square 225, whether processing logic determines one to be predetermined situation and to take place, and is expressed as this afterwards and adjusts triggering.In one embodiment, this is adjusted to trigger and comprises a specified time interval (for example a hour, a day or the like).Adjust when this and to trigger the processing operation, comprise a given number and be predetermined data sampling of number or the like.In case the processing of this given number operation is done, produces this and is predetermined data sampling of number or the like, this is adjusted triggering and can take place.Capable of being combinedly for example produce a data setting that is predetermined number if one or more adjusts to trigger to cause, perhaps from one before adjusted one be predetermined the time interval and expire, adjust and carry out one.Do not take place if this adjusts to trigger, then method finishes.Take place if this adjusts triggering, this method advances to square 230.
In square 230, adjust one of this multivariate statistics model or many persons.Available algorithms of different is adjusted with drift, and the person comprises index weight moving average (exponentially weightedmoving average (EWMA)) one of in these algorithms.Other is suitable adjusts the use that algorithm comprises forgetting factor (forgettingfactor), windowization (windowing) and recurrence moving average (recursive moving average).Also can use other to adjust algorithm.
In one embodiment, adjust all statistics of each person of these treatment variables.Person in addition, the particular procedure variable do not adjusted and/or one or the certain statistical of multiprocessing variable can not adjusted.In one embodiment, processing logic is discerned one first group one or multiprocessing variable, and it respectively adds up by expection to drift about in the normal running of a manufacturing machine.This treatment variable of first group is adjusted in this model, keeps all other treatment variable statistics simultaneously.Processing logic also can be discerned one second group one or multiprocessing variable, and its each statistics is come not remain unchanged under the mistake having by expection.This second group all outer treatment variables can be adjusted in this model, keep this second group outer treatment variable simultaneously.In one embodiment, this treatment variable of first group and this second group outer treatment variable can use together.This will allow processing logic detect gradually wrong (gradual faults) and suddenly wrong (sudden faults) both, and can avoid the mistake gradually that is adapted to of mistake.
Can at each through adjusting treatment variable and adjust one or many statistics.The example of each statistics that can be adjusted at treatment variable comprises middle number (mean), variation, covariance, relevance (for example relevance array), critical piece proper vector and character numerical value, reaches the number of critical piece.In one embodiment, can adjust different statistics at different treatment variables.Can import selection based on the user and add up at what person of treatment variable (if any) and adjust, perhaps can user's input and (for example based on a selection algorithm) automatically for it.For example, the middle number of particular procedure variable and variation can be drifted about by expection, and the covariance of these treatment variables of other treatment variable of tool can not drifted about by expection simultaneously.Therefore, the middle number of this suitable treatment variable and variation can be drifted about, and the statistics of the covariance between these treatment variables and other treatment variable is kept fixing simultaneously.In other example, at specific other treatment variable, all statistics can be drifted about by expection.Therefore, all statistics of those other treatment variables can be adjusted.
Referring now to Fig. 2, in square 235, processing logic judges to adjust whether to change at least one threshold value of one or more univariate statistics, censures to adjusting threshold value afterwards.When adjusting a multivariate statistics model, it has the accumulation calculating mistake (cumulative computational rounding errors) that rounds off will cause this make mistake risk of change of numerical value as calculated to reality and unaltered univariate statistics.The mistake that adds up in this univariate statistics like this can cause unbecoming (disproportionate) mistake in the calculating of this multivariate statistics (for example covariance).In order to improve such generation, this adjusts threshold at least with not changing one or more single argument if adjust, and this method advances to square 245.If one or more univariate statistics will be changed this and adjust threshold value, this method advances to square 240.
In one embodiment, to adjust threshold value be a fixed numbers to this of each other univariate statistics.Person in addition, it is a respective value that this of a univariate statistics adjusted threshold value, for example a mark that is predetermined (fraction) of a present numerical value.For example, in one embodiment, this of each other univariate statistics adjust threshold value can be these indivedual univariate statistics present numerical value part per billion (10
-9).One or many univariate statistics can share the identical threshold value of adjusting.Person in addition, the specific or owner of this univariate statistics has their threshold value of adjusting.
In square 240, these will change more than or equal to this adjust threshold value one the amount univariate statistics be changed.Can be postponed up to this change not changing adjusting of this univariate statistics of adjusting threshold value really above this threshold value.
At square 245, this (respectively) multivariate statistics model through adjusting is used to analyze ensuing deal with data with the detecting mistake.This method then finishes.
In one embodiment, after judging that this nearest deal with data does not depart from this model of enough pointing out a mistake, this nearest deal with data is used to adjust a multivariate statistics model.Person in addition, this nearest deal with data can be used to adjust this model before execution error detection.In other specific embodiment, according to the described method of the following Fig. 3 of reference, in case before adjusting and in case after adjusting, execution error detection is twice on deal with data.
Fig. 3 describes a kind of process flow diagram of detecting a specific embodiment of wrong method 300.This method can be carried out by processing logic, and this processing logic can comprise hardware (for example circuit, special logic, programmable logic, microcode or the like), software (instruction that for example moves) or above combination on treating apparatus.In one embodiment, method 300 can be carried out by the statistical treatment watch-dog 105 of Fig. 1.
Referring now to Fig. 3, method 300 starts from receiving the processing logic (square 310) of deal with data.In square 315, one first error detection algorithm is applied to this deal with data.In one embodiment, this first error detection algorithm uses loose relatively (insensitive) error detection threshold value.In square 320, processing logic determines whether a mistake is pointed out.In one embodiment, if point out a mistake, this method advances to square 335, and if do not point out faults, this method advances to square 325.In a substituting specific embodiment, this method advances to square 325, and it points out whether a mistake is pointed out.
In square 335, report a mistake.Reporting a mistake can comprise and notify a user via transmitting message to a client, send a warning, stop one and handle or the like making on the machine.This method then finishes.
In square 325, one or more multivariate statistics model can be adjusted.Then, can use one second error detection algorithm (square 330).In one embodiment, this second detecting algorithm uses this through adjusting model to determine whether a mistake takes place.In one embodiment, this second error detection algorithm uses corresponding tight (sensitivity) error detection threshold.The use of two error detection algorithms can reduce probability that triggers false alarm and the probability that increases the detecting factual error.
The general experience one of this operational circumstances in the semiconductor process chamber in the machine reparation, proofread and correct the unexpected skew after keeping (for example substituting or the parts of classifying) or prevention and keeping (for example clearing up a process chamber), it is machine maintenance that its owner censures with being collected.For avoiding this unexpected skew that is identified as a mistake, the model of its desire " replacement " this all or part behind machine maintenance.
Fig. 4 describes a kind of at machine maintenance after by resetting one or multiple statistics model and detect the process flow diagram of a specific embodiment of wrong method.This method can be carried out by processing logic, and this processing logic can comprise hardware (for example circuit, special logic, programmable logic, microcode or the like), software (instruction that for example moves) or above combination on treating apparatus.In one embodiment, method 400 can be carried out by the statistical treatment watch-dog 105 of Fig. 1.
With reference to Fig. 4, method 400 starts from detecting machine maintenance (square 405).Can detect this machine maintenance at a specific manufacturing machine or at multiple manufacturing machine.In square 410, one or more multivariate statistics model is automatically reset.One model is reset in the time of can changing in the mode that the hint machine maintenance has been performed at the numerical value when a treatment variable and quilt initialization automatically.The example that these treatment variables change comprises a counter (for example pointing out that parts replace) that is reset to zero, stop work (out of service) to be longer than one and to be predetermined one of period time and to make machine or the change in the particular procedure set-point value.Person in addition, one of one or more multivariate statistics model is reset can be by user initialization manually, for example makes machine and is operated by answer behind machine maintenance when one.
In one embodiment, this model is finished a training during period and carrying out a state of a virgin state that is produced before any adjusting and be reset (square 415) via returning back to be coincident with.In another specific embodiment, this model is reset (square 420) via the model of adjusting all or part to new operational circumstances, and is as above described with reference to Fig. 2.For example, at the operational circumstances that can be adjusted through the statistics of selecting treatment variable to reflect that this is new.In one embodiment, through selecting to comprise those variablees that changed by expection because of the actual execution of the maintenance of a particular type for the treatment variable of adjusting.The person can be maximum wrong contribution and/or wrong those treatment variables contributed that have greater than a threshold numerical value through selection for the treatment variable of adjusting in addition.This calculating can be based on adding up via being applied to deal with data to the mismatch that this model produced as its existence before this machine maintenance that produces behind this machine maintenance.In one embodiment, reduce under the threshold value that is predetermined through selecting a number for each treatment variable of adjusting repeatedly to be increased up to one or more model error statistics.
In one embodiment, this multivariate statistics model of resetting comprises initialization one and resets the training period (square 425).Can be used to recomputate the multivariate statistics model of all or part from the deal with data of this replacement training period.In one embodiment, this training period of resetting is used the deal with data from the actual treatment of the product on the manufacturing machine.In one embodiment, the multivariate error detection is made inefficacy during this replacement training period.This will be avoided the generation of many false alarms.Person in addition, the multivariate error detection makes inefficacy at particular error classification and/or error characteristic during period is trained in this replacements.Therefore, may can be suppressed for each mistake of false alarm, yet factual error still can be monitored.In case enough deal with data have been collected to rebulid at least one multivariate statistics model, this replacement training period is moved to end.
In one embodiment, when deal with data pointed out that a new multivariate statistics model has converged to the statistics of a stable set, this replacement training period was moved to end, and error detection restarts.This will occur in when the certain statistical of this new multivariate statistics model is changed less than a threshold numerical value via the new deal with data of introducing.Person in addition, this replacements training period can be comfortable reached when being merged in this new model by this manufacturing machine generation and finishes when a training data sampling that is predetermined number.In another specific embodiment, this replacement training period can be when relatively this new model and deal with data the time be added up when one or more model error and is reduced to one when being predetermined threshold value and finish.In another specific embodiment again, this replacement training period can be when relatively this new model and this training data the time when wherein one or more model error statistics surpasses a frequency reduction by that is predetermined threshold value and is predetermined threshold value and finish.In another specific embodiment again, this replacement training period can reduce by one in the frequency that the middle numerical value that is different from them when the treatment variable through selecting wherein is bigger than their standard deviation when being predetermined threshold value and finish.
In one embodiment, this following existing multivariate statistics model to one second manufacturing machine that is used to apply one first manufacturing machine with reference to the described technology of a multivariate statistics model of resetting.For conversion second is made machine from this first multivariate statistics model of making machine to this, a duplicate of this multivariate statistics model is produced and is associated in this and second made machine.In one embodiment, be a current state of this model on this first machine at this second init state of making this model of machine.One adjusts and/or the training of resetting can then be initialised and to adjust this model to this second machine the period.
Fig. 5 describes the graphic representation of the machine in the computing system 500 of an exemplary form, wherein has any one or more method that one group of instruction can be carried out and be discussed in order to cause this machine to carry out in this.In each alternative formula specific embodiment, can connect other machine in (for example network connection) this machine to one LAN, enterprise network or the world-wide web.This machine is operable in the server in one client-server-side network environment or the ability of a client, or is operating as a bit (peer) machine in one point-to-point (or distributed) network environment.This machine can be box on a personal computer, desktop PC, the machine (set-top box, STB), personal digital assistant PDA, mobile phone, network application device, server, network router, switch or bridge or any machine that can carry out one group of instruction (sequence or non-sequence) of the action that appointment taked by machine.Moreover when only describing a single machine, this project " machine " should also may be utilized and carry out and appoint any collection of machines of closing one of described in this or multi-method to carry out one group of (many groups) instruction with comprising individually or connect.
This example formula computer system 500 comprises a treating apparatus (processor) 502, main storer 504 (for example ROM (read-only memory), flash memory, dynamic RAM (for example Synchronous Dynamic Random Access Memory or Rambus dynamic RAM) or the like), static memory 506 (for example flash memory, static RAM or the like) and data memory device 518, and it communicates with other person by bus 530.
The treating apparatus (for example microprocessor, CPU (central processing unit) or the like) of processor 502 one or more general intentions of expression.Special, this processor 502 can be a sophisticated vocabulary and calculates (CISC) microprocessor, reduced instruction set computer and calculate (RISC) microprocessor or very long instruction word group (VLIW) microprocessor or can make the processor of other instruction set in fact or can make the processor of a combined type instruction set in fact.This processor 502 also can be one or the treating apparatus of many specific intended (for example using specific integrated circuit (applicationspecific integrated circuit (ASIC)), a field programmable gate array, a digital signal processor, network processing unit or above fellow).This processor 502 through configuration to carry out processing logic 526 for carrying out operation and the step described in this.
This computer system 500 more comprises a Network Interface Unit 508.This computer system 500 also comprises input media 512 (for example keyboard), finger control device (for example mouse) and a signal generation device 516 (for example loudspeaker) of a video display unit 510 (for example a liquid crystal display (LCD) or cathode ray tube (CRT)), letter and numeral.
But this data memory device 518 can comprise machine access Storage Media 531, can store on it one or many groups can use any one or more method described in this or the instruction of function.This software 522 also can be resided on fully or at least in part in this main storer 504 and/or the processor 502, but this main storer 504 and this processor 502 also can constitute machine access Storage Media via this computer system 500 term of execution.This software 522 more can be transmitted and receive on a network 520 by this Network Interface Unit 508.
But these machine access medium 53l also can be used for storing the set of data structures of definition user user state and user's hobby of definition user catalogue.Set of data structures and user's hobby also can be stored in other section of computer system 500, and for example static memory 506.
When but this machine access Storage Media 531 is shown as a single medium in an exemplary specific embodiment, this project " but machine access Storage Media " should may be utilized with comprise can store one or a single medium of many group instructions or multiple medium (for example centralization or distributed data base, with and/or relevant getting soon and server).People and medium that this project " but machine access Storage Media " should also may be utilized and can store, encodes or carry one group of instruction to comprise are in order to carry out of the present invention one or multi-method.Therefore this project " but machine access Storage Media " should be used with including (but not limited to) solid-state memory, optics and magnetic medium, and carrier signal.
Should recognize above description only for illustration do not regard it as the restriction.Many other specific embodiments all can be after being familiar with that this skill person reads and understanding foregoing description and real work the in addition.Therefore, category of the present invention should be determined with reference to following claims of enclosing, and the present invention also comprises all categories as each equipollent of these claims.
Claims (21)
1. detect wrong method for one kind, it comprises:
Receive deal with data, this deal with data comprises a plurality of treatment variables;
Adjust one or more multivariate statistics model according to this deal with data, the process of wherein adjusting comprises and applies an at least one univariate statistics that changes to this one or more multivariate statistics model, if it in the time should changing greater than a threshold value and for it; And
The deal with data of using this one or more multivariate statistics model through adjusting to analyze to continue is with the detecting mistake.
2. the method for claim 1, it more comprises:
Before adjusting this one or more multivariate statistics model or after adjusting these multivariate statistics models, judge whether this deal with data points out the mistake at this one or more multivariate statistics model.
3. the method for claim 1, wherein this one or more multivariate statistics model adjusting according to one of this a plurality of treatment variables or many persons once measuring to drift about.
4. the method for claim 1 is wherein adjusted from the treatment variable of one first subclass of these a plurality of treatment variables, and is not wherein adjusted from the treatment variable of one second subclass of these a plurality of treatment variables.
5. the method for claim 1, it more comprises:
Via applying one first error detection algorithm, before adjusting this one or more multivariate statistics model, judge whether this treatment variable points out a mistake; And
Via applying one second error detection algorithm, after adjusting this one or more multivariate statistics model, judge whether this deal with data points out a mistake, wherein the control of this first error detection algorithm restriction is extensively in the control restriction of this second error detection algorithm.
6. the method for claim 1, wherein this one or more multivariate statistics model comprises at least one first model and one second model, by the number of a quantity of the historical data of being considered, employed treatment variable, employed critical piece, trust restriction, forgetting factor, be adapted to this a method, treatment variable of being adjusted through measuring drift, be used for the training data that model produces and adjust threshold value at least one and this second model is different from this first model.
7. the method for claim 1, it more comprises:
During machine maintenance on a detecting instrument relevant with this deal with data, at least one of these multivariate statistics models of automatically resetting.
8. method as claimed in claim 7, wherein reset process comprise recomputate at have statistics greater than the treatment variable of a wrong contribution of a threshold value, adjust a multivariate statistics model to small part to new operational circumstances, and reply this multivariate statistics model at least one of a state that is coincident with a virgin state that is produced during the period in a training.
9. method as claimed in claim 7, wherein reset process comprises initialization one and resets the training period, wherein should replacement trains period to be used to collect data to upgrade at least one of these multivariate statistics models.
10. the method for claim 1, the process of wherein adjusting more comprise and apply at least one of a number, load vector (loading vectors), middle number (mean), variation, covariance, critical piece proper vector and character numerical value that changes to a relevance array, critical piece.
11. but machine access medium that comprise data, it causes this machine to carry out a method when by the access of a machine institute, and this method comprises:
Receive deal with data, this deal with data comprises a plurality of treatment variables;
Adjust one or more multivariate statistics model according to this deal with data, the process of wherein adjusting comprises and applies an at least one univariate statistics that changes to this one or more multivariate statistics model, if it in the time should changing greater than a threshold value and for it; And
The deal with data of using this one or more multivariate statistics model through adjusting to analyze to continue is with the detecting mistake.
But 12. machine access medium as claimed in claim 11, this method more comprises:
Before adjusting this one or more multivariate statistics model or after adjusting these multivariate statistics models, judge whether this deal with data points out the mistake at this one or more multivariate statistics model.
13. but machine access medium as claimed in claim 11, wherein this one or more multivariate statistics model adjusting according to one of these a plurality of treatment variables or many persons once measuring drift.
But 14. machine access medium as claimed in claim 11, this method more comprises:
Via applying one first error detection algorithm, before adjusting this one or more multivariate statistics model, judge whether this treatment variable points out a mistake; And
Via applying one second error detection algorithm, after adjusting this one or more multivariate statistics model, judge whether this deal with data points out a mistake, wherein the control of this first error detection algorithm restriction is extensively in the control restriction of this second error detection algorithm.
15. but machine access medium as claimed in claim 11, wherein this one or more multivariate statistics model comprises at least one first model and one second model, by the number of a quantity of the historical data of being considered, employed treatment variable, employed critical piece, trust restriction, forgetting factor, be adapted to this a method, treatment variable of being adjusted through measuring drift, be used for the training data that model produces and adjust threshold value at least one and this second model is different from this first model.
But 16. machine access medium as claimed in claim 11, this method more comprises:
During machine maintenance on a detecting instrument relevant with this deal with data, at least one of these multivariate statistics models of automatically resetting.
But 17. machine access medium as claimed in claim 11, the process of wherein adjusting more comprises and applies at least one of a number, load vector (loading vectors), middle number (mean), variation, covariance, critical piece proper vector and character numerical value that changes to a relevance array, critical piece.
18. a statistical treatment supervisory system, it comprises:
One database, it is used to store one or more multivariate statistics model, and
One error detection device, its be coupled at least one manufacturing machine and this database, this error detection device is in order to receive the deal with data from this at least one manufacturing machine, wherein this deal with data comprises a plurality of treatment variables, with adjust this one or at least one of multivariate statistics model, the process of wherein adjusting comprises and applies an at least one univariate statistics that changes to this one or more multivariate statistics model, if it is in the time should changing greater than a threshold value and for it, and in order to the deal with data using this one or more multivariate statistics model to analyze to continue through adjusting with the detecting mistake.
19. statistical treatment supervisory system as claimed in claim 18, wherein this error detection device more in order to before adjusting this one or more multivariate statistics model or after adjusting these multivariate statistics models, judges whether this deal with data points out the mistake at this one or more multivariate statistics model.
20. statistical treatment supervisory system as claimed in claim 18, wherein this one or more multivariate statistics model adjusting according to one of these a plurality of treatment variables or many persons once measuring drift.
21. statistical treatment supervisory system as claimed in claim 18, wherein this error detection device is more in order to when the machine maintenance of detecting on this at least one manufacturing machine, at least one of these multivariate statistics models of automatically resetting.
Applications Claiming Priority (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US74664706P | 2006-05-07 | 2006-05-07 | |
| US60/746,649 | 2006-05-07 | ||
| US60/746,647 | 2006-05-07 | ||
| US11/800,471 | 2007-05-04 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101438251A true CN101438251A (en) | 2009-05-20 |
Family
ID=40711656
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007800165002A Pending CN101438251A (en) | 2006-05-07 | 2007-05-07 | Adaptive multivariate fault detection |
| CNA2007800165017A Pending CN101438249A (en) | 2006-05-07 | 2007-05-07 | Ranged fault signatures for fault diagnosis |
Family Applications After (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007800165017A Pending CN101438249A (en) | 2006-05-07 | 2007-05-07 | Ranged fault signatures for fault diagnosis |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN101438251A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102844721A (en) * | 2010-02-26 | 2012-12-26 | 株式会社日立制作所 | Failure source diagnosis system and method |
| CN104601584A (en) * | 2009-07-17 | 2015-05-06 | 美国运通旅游有关服务公司 | Systems, methods, and computer program products for adapting security measures of a communication network based on feedback |
| CN105492982A (en) * | 2013-06-07 | 2016-04-13 | 科学设计公司 | System and method for monitoring a process |
| US9847995B2 (en) | 2010-06-22 | 2017-12-19 | American Express Travel Related Services Company, Inc. | Adaptive policies and protections for securing financial transaction data at rest |
| US9973526B2 (en) | 2009-12-17 | 2018-05-15 | American Express Travel Related Services Company, Inc. | Mobile device sensor data |
| US10360625B2 (en) | 2010-06-22 | 2019-07-23 | American Express Travel Related Services Company, Inc. | Dynamically adaptive policy management for securing mobile financial transactions |
| US10395250B2 (en) | 2010-06-22 | 2019-08-27 | American Express Travel Related Services Company, Inc. | Dynamic pairing system for securing a trusted communication channel |
| US10432668B2 (en) | 2010-01-20 | 2019-10-01 | American Express Travel Related Services Company, Inc. | Selectable encryption methods |
| US10997571B2 (en) | 2009-12-17 | 2021-05-04 | American Express Travel Related Services Company, Inc. | Protection methods for financial transactions |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7596718B2 (en) | 2006-05-07 | 2009-09-29 | Applied Materials, Inc. | Ranged fault signatures for fault diagnosis |
| US8010321B2 (en) | 2007-05-04 | 2011-08-30 | Applied Materials, Inc. | Metrics independent and recipe independent fault classes |
| WO2011048661A1 (en) * | 2009-10-19 | 2011-04-28 | 日立建機株式会社 | Diagnosis system and diagnosis method of construction machine |
| US8838413B2 (en) * | 2011-05-12 | 2014-09-16 | Saudi Arabian Oil Company | Valve actuator fault analysis system |
| EP3403152B1 (en) | 2016-03-09 | 2023-07-12 | Siemens Aktiengesellschaft | Smart embedded control system for a field device of an automation system |
| US11112449B2 (en) * | 2018-04-06 | 2021-09-07 | Bendy Nevada, LLC | Flexible and scalable monitoring systems for industrial machines |
| CN111857092B (en) * | 2020-06-22 | 2024-04-30 | 杭州群核信息技术有限公司 | Real-time error detection system and method for household parameterized model |
| CN112734641B (en) * | 2020-12-31 | 2024-05-31 | 百果园技术(新加坡)有限公司 | Training method and device for target detection model, computer equipment and medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5446664A (en) * | 1992-10-07 | 1995-08-29 | Spx Corporation | Method and apparatus for diagnosing faulty cylinders in internal combustion engines |
| JP3050065B2 (en) * | 1994-10-18 | 2000-06-05 | 三菱自動車工業株式会社 | Power failure detection device |
| CN1120366C (en) * | 1999-03-22 | 2003-09-03 | 西安交通大学 | Fault detecting and diagnosing method based on non-linear spectral analysis |
| US6829527B2 (en) * | 2002-08-26 | 2004-12-07 | Honeywell International Inc. | Relational database for maintenance information for complex systems |
| US7263632B2 (en) * | 2003-05-07 | 2007-08-28 | Microsoft Corporation | Programmatic computer problem diagnosis and resolution and automated reporting and updating of the same |
| US7239978B2 (en) * | 2004-03-31 | 2007-07-03 | Wu-Tung Cheng | Compactor independent fault diagnosis |
-
2007
- 2007-05-07 CN CNA2007800165002A patent/CN101438251A/en active Pending
- 2007-05-07 CN CNA2007800165017A patent/CN101438249A/en active Pending
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104601584A (en) * | 2009-07-17 | 2015-05-06 | 美国运通旅游有关服务公司 | Systems, methods, and computer program products for adapting security measures of a communication network based on feedback |
| US9848011B2 (en) | 2009-07-17 | 2017-12-19 | American Express Travel Related Services Company, Inc. | Security safeguard modification |
| CN104601584B (en) * | 2009-07-17 | 2018-05-11 | 美国运通旅游有关服务公司 | System, method and computer program product for the security measures based on feedback adjustment communication network |
| US10735473B2 (en) | 2009-07-17 | 2020-08-04 | American Express Travel Related Services Company, Inc. | Security related data for a risk variable |
| US10218737B2 (en) | 2009-12-17 | 2019-02-26 | American Express Travel Related Services Company, Inc. | Trusted mediator interactions with mobile device sensor data |
| US10997571B2 (en) | 2009-12-17 | 2021-05-04 | American Express Travel Related Services Company, Inc. | Protection methods for financial transactions |
| US9973526B2 (en) | 2009-12-17 | 2018-05-15 | American Express Travel Related Services Company, Inc. | Mobile device sensor data |
| US10931717B2 (en) | 2010-01-20 | 2021-02-23 | American Express Travel Related Services Company, Inc. | Selectable encryption methods |
| US10432668B2 (en) | 2010-01-20 | 2019-10-01 | American Express Travel Related Services Company, Inc. | Selectable encryption methods |
| CN102844721B (en) * | 2010-02-26 | 2015-11-25 | 株式会社日立制作所 | Failure cause diagnostic system and method thereof |
| CN102844721A (en) * | 2010-02-26 | 2012-12-26 | 株式会社日立制作所 | Failure source diagnosis system and method |
| US9847995B2 (en) | 2010-06-22 | 2017-12-19 | American Express Travel Related Services Company, Inc. | Adaptive policies and protections for securing financial transaction data at rest |
| US10360625B2 (en) | 2010-06-22 | 2019-07-23 | American Express Travel Related Services Company, Inc. | Dynamically adaptive policy management for securing mobile financial transactions |
| US10395250B2 (en) | 2010-06-22 | 2019-08-27 | American Express Travel Related Services Company, Inc. | Dynamic pairing system for securing a trusted communication channel |
| US10715515B2 (en) | 2010-06-22 | 2020-07-14 | American Express Travel Related Services Company, Inc. | Generating code for a multimedia item |
| US10104070B2 (en) | 2010-06-22 | 2018-10-16 | American Express Travel Related Services Company, Inc. | Code sequencing |
| TWI648609B (en) * | 2013-06-07 | 2019-01-21 | 美商科學設計股份有限公司 | Program monitoring system and method |
| US9892238B2 (en) | 2013-06-07 | 2018-02-13 | Scientific Design Company, Inc. | System and method for monitoring a process |
| CN105492982A (en) * | 2013-06-07 | 2016-04-13 | 科学设计公司 | System and method for monitoring a process |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101438249A (en) | 2009-05-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101438251A (en) | Adaptive multivariate fault detection | |
| US11487252B2 (en) | Process model identification in a process control system | |
| US20190384255A1 (en) | Autonomous predictive real-time monitoring of faults in process and equipment | |
| US10733536B2 (en) | Population-based learning with deep belief networks | |
| JP2023017888A (en) | Computer systems and methods for performing root cause analysis and building predictive model for rare event occurrences in plant-wide operations | |
| CN1860487B (en) | System and method for using first-principles simulation to analyze a process performed by a semiconductor processing tool | |
| CN100555136C (en) | Unusual condition prevention method and system in the processing factory | |
| CN101517505B (en) | Method and system for detecting abnormal operation in a hydrocracker | |
| CN100476733C (en) | System and method for semiconductor simulation on tool | |
| EP1419424B1 (en) | Shared-use data processing for process control systems | |
| Zhou et al. | Differential feature based hierarchical PCA fault detection method for dynamic fault | |
| US20190227504A1 (en) | Computer System And Method For Monitoring Key Performance Indicators (KPIs) Online Using Time Series Pattern Model | |
| CN101842756A (en) | Be used for chemical plant or refinery continuously, the System and method for of in-service monitoring | |
| US20100100521A1 (en) | Diagnostic system | |
| WO2007131185A2 (en) | Training a support vector machine with process constraints | |
| WO2003046770A1 (en) | System and method for historical database training of support vector machines | |
| WO2003042918A2 (en) | Analysing events | |
| US7242995B1 (en) | E-manufacturing in semiconductor and microelectronics processes | |
| US10928807B2 (en) | Feature extraction and fault detection in a non-stationary process through unsupervised machine learning | |
| Yang et al. | Performance monitoring method based on balanced partial least square and statistics pattern analysis | |
| US10139788B2 (en) | Remote data analytics to predict system components or device failure | |
| US10733533B2 (en) | Apparatus and method for screening data for kernel regression model building | |
| Benqlilou | Data reconciliation as a framework for chemical processes optimization and control | |
| Chamness | Multivariate fault detection and visualization in the semiconductor industry | |
| Suzuki et al. | An Online Anomaly Detection System Supporting Batch-Process Operator Decision-Making |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20090520 |
|
| C20 | Patent right or utility model deemed to be abandoned or is abandoned |