TW200811978A - Ranged fault signatures for fault diagnosis - Google Patents

Abstract

A method and apparatus for diagnosing faults. A fault is detected. One or more process variables that contributed to the fault are determined. A relative contribution of each of the one or more process variables is determined. A determination is made as to which fault signatures match the fault, a match occurring when the relative contributions of the one or more process variables are within relative contribution ranges of the matching fault signature. Each fault signature is associated with at least one fault class.

Description

200811978 IX. DESCRIPTION OF THE INVENTION: TECHNICAL FIELD OF THE INVENTION [0002] Embodiments of the present invention relate to error detection, particularly error detection using a variety of error characteristics. [Prior Art] Many companies use sophisticated equipment that includes multiple sensors and controllers that are carefully monitored for quality during processing. A means of monitoring statistical processing of such multiple sensors and controllers (a statistical analysis performed on sensor measurements and processing controls, rational variables), which enables automatic, error detection. A "fault" can be used to avoid an impending adjustment for the manufacturing equipment to be out of tune (for example, to maintain a desired indication of the error in the operation of a machine with the desired value. Therefore, one of the statistical processing monitoring targets In order to generate the above detection and/or detection error, during processing monitoring, when one or more statistical models of the most recently processed data deviate by an amount, and the amount is large enough to cause a threshold of another trust threshold, the detection Measure an error. A model metering value whose value indicates the statistical difference between the statistical features collected during the actual processing monitoring period and the statistical characteristics predicted by the model is the only mathematical method to eliminate this deviation. Common models include Squared Prediction Error (SPE, Qres, or Q), and Hotelling's T2. The method for making a tight manufacturing control to ensure that the method is a Ν value (detection and / or fault or poor), or For the fault or loss of defects, the previous statistical self-measurement excess is a purely rational data. Each model-type measurement package is generally referred to as 5 200811978 models. The quantity has an individual trust threshold, which is also referred to as a limit or control limit, where the value indicates that one of the model measures can be connected to an upper limit. If a molding meter exceeds its individual signal value during processing monitoring, the process should be inferred. The data has deviated from the threshold due to an error. Once errors are detected, the errors are detected by ignoring the error contribution of one of the processing variables. Some errors are difficult due to the lack of explicit (eg direct) correlation with a single number of points. Detection. Errors with complex and/or indirect correlations of polymorphic variables can be particularly difficult to measure. 0 Common methods of detecting errors generally require a multiple occurrence of an error before being classified. This will be a classification with multiple processing The problem of erroneous correlation of variables. [Invention] One aspect of the present invention relates to a method for detecting errors, including: detecting an error; determining contribution to the one or more processing variables; The contribution of each of the one or more processing variables; and the determination of which of the complex features of the complex number is consistent with the error The feature conforms to the error, if the respective contributions of the one or more processes are within a corresponding contribution range of the matching error feature, each of the error features is associated with at least one error classification. A system is a machine-readable medium that contains information that, when accessed by a machine, causes the machine to perform a method that includes a debt test error; determines that one or more trusts contributing to the error are subject to any failure. The relative rationality is in the error of detecting the error of the error, the access method in the variable, processing 6 200811978 variables; determining the error characteristics of a corresponding contribution of each of the one or more processing variables Which one is consistent with the error, and an error corresponds to the error, if the corresponding one or more processing variables are within the corresponding contribution range of the matching error feature, wherein each of the errors is related to at least A misclassification. Still another aspect of the present invention relates to a statistical processing monitoring system including an error detector coupled to at least one manufacturing machine for receiving processing data from the at least one manufacturing machine and for processing data Detecting an error, the processing data includes a plurality of points; a database for storing complex error features, each of which is associated with at least one error classification; and an error detection system The error detector and the database are coupled to determine one or more of the processing variables of the complex number to the error, determine a corresponding contribution of each of the or multiple processing variables, and determine a characteristic of the complex number Whichever corresponds to the error, an error signature that matches the error is if the corresponding contribution of the one or more processing variables is within the corresponding contribution range of the matching feature. [Embodiment] This description is a method and apparatus for detecting an error. In a particular embodiment, the identification contributes to one of the errors or the multi-processing variable has a measured value outside of a control limit, and a processing variable is assigned to the error. Determining one phase of each of the one or more processing variables, and the characteristic contribution of the erroneous system, by the rationality of the erroneous device, the contribution, etc., the error is in the same way as the tribute to the tribute 7 200811978 offer. The respective contributions are standardized and arranged in an arranged list, wherein the ranking is based on one of the error contributions. A fault feature is determined that matches the detected error. In a specific embodiment, if the corresponding contribution of the identified processing variable is within a corresponding contribution range of the matched error feature, an error feature conforms to the detected error. Each of the error characteristics is associated with at least one error classification identifying a particular cause of the error. In the following description, a number of details are set forth. However, it will be apparent to those skilled in the art that the present invention may be practiced without the following specific details. In the exemplification, the structures and devices are shown in the form of a block diagram (not a detail) in order to avoid obscuring the invention. Some of the detailed descriptions are represented by algorithms and representative symbolic symbols of the data bits in a computer memory. Those skilled in the art of data processing techniques use the narrative and presentation of these algorithms to convey the essence to other skilled practitioners in the most efficient manner. Algorithms, which are generally referred to herein as procedures for directing a self-consistent step or instruction of a desired result. These steps are for those who need physical manipulation of physical quantities. Although not necessarily, these quantities are typically in the form of electrical or magnetic signals capable of being stored, transferred, combined, compared, or otherwise manipulated in a computer system. It has proven convenient at present to express these signals as bits, values, elements, symbols, characters, terms, numbers, etc., primarily based on general usage factors. However, it should be noted that these and similar terms are related to the appropriate physical quantities and are merely convenient labels applied to these quantities. Unless otherwise stated, it can be clearly seen in the discussion that the terminology used in the article is 8 200811978 is "processing", "transfer" means "computer system" or class: calculation" or "definite" or "display", etc. 'The operation and processing of the electronic transfer device in the control and conversion, which is expressed as the physical (electronic) amount of the register and the data in the memory.

The memory or other such resources: for the memory in the computer system, or the temporary invention is also related to the physical quantity of the ":: transmission or display device. The operation can be performed according to the demand and the operation described in the text - The device is selectively activated by the device. The computer can also be used in the computer::: stored in a computer readable medium, such as (but...), any kind of floppy disk, CD Read-only memory (CD-ROMs) with CD-ROM, read-only memory (rqMs), random access memory _) can be used to eliminate stylized read-only memory (EPROMs), electronically erasable stylization Read only (4) (EEPR〇Ms), magnetic or optical card or any kind of media suitable for storing electronic instructions. : The described/beautiful method and module are not related to any particular computer. Various general-purpose systems may be in parallel with the program as taught by the present invention, or may prove to facilitate the construction of more specialized equipment to perform the steps of the method. The architecture required for these various systems will be described in detail below. The invention is not through any particular programming language The disclosure of the invention described herein may be implemented in a variety of programming languages.The machine-readable medium includes any mechanism for storing or transmitting information in a form readable by a machine. For example, a machine readable The media I includes a machine readable storage medium (such as a read only memory (ROM), a random access memory (RAM), a disk storage medium, an optical storage medium, a 9200811978 flash memory device, etc.), and the machine can Read transmission media (electrical, audio or other forms of propagable signals (eg carrier waves, infrared digital signals, etc.)), etc. The following description provides monitoring and/or detection errors that operate on the manufacturing device (unstable Details of a statistical system of the manufacturing process. In a specific embodiment, the statistical processing is monitored for the manufacture of electronic devices, such as semiconductors. Manufacturing such as many manufacturing steps involving different types of manufacturing processes is required. , sputtering, chemical vapor deposition are three different types of processing, performed on different types of machines. In addition, the statistical processing Monitoring is used to monitor the manufacture of other products (such as cars). This other product also requires a number of different processing steps processed by various manufacturing machines. Figure 1 depicts one of the statistical processing monitoring systems 100. 00 includes a statistical processing monitoring device coupled to one or more manufacturing machines 110 or a plurality of processing controllers 150 via a data communication link 160. The statistical processing monitoring system can be included in a factory (eg, Manufacture of a manufacturing plant. In addition, the statistical processing monitoring system 100 may include a manufacturing machine 110 in a factory, for example, all manufacturing machines 110 may be in one or more specific processes. In one embodiment, each manufacturing machine 110 is a machine that manufactures electricity, such as an etcher, a chemical vapor deposition furnace, a photolithography device, an implanter (imp 1 anter), and the manufacturing machine 11 0 can be used to manufacture other products (such as steam, optical, signal, and motion detection systems). 105, and the sub-devices of the special 100-machines, etc. Car) 10 200811978 A type. In a specific embodiment, each of the manufacturing machines 110 can be of a single type. Alternatively, the manufacturing machine 110 can include a number of different types of equipment, each of which can perform different processes. Each manufacturing machine 110 can include multiple sensors for monitoring operation on the manufacturing machine 110. One type of sensor included in the manufacturing machine 110 may be a temperature sensor. Examples of other sensors include a pressure sensor, a flow rate sensor, or any other sensor that monitors the physical properties of a working component made by the manufacturing machine 110 or the physical condition of a manufacturing process. Each manufacturing process performed on a manufacturing machine 特征 is characterized by various physical conditions and attributes detected by the sensor and various operational parameters that are collected to correlate as processing data. Each distinct physical condition or attribute detected by the sensor, as well as each operational parameter, can be a distinct processing variable of the processed data. Examples of process variables representative of detector data include process chamber pressure, susceptor temperature, RF forward power, and RF reflected power. Examples of process variables indicative of operational parameters include flow rate settings (e.g., for chemical reagents) and throttling settings (e.g., for a chamber exhaust vacuum pump). The sensor, manufacturing machine, and process controller can be monitored during processing to collect the process variables in a timely manner at successive points. In a specific embodiment, each process variable is applied to a particular process. Alternatively, one or more processing variables can be applied to only a portion of a particular process. In a specific embodiment, sensor measurements and operational parameters at different steps in a process represent distinct processing variables (modeled as additional scale in the model space 11 200811978). This can be useful, for example, if multiple steps are performed at a manufacturing location in a machine with different operating parameter settings. Example In a three-step manufacturing process, one of the susceptor temperatures during the three-step process is considered to be three distinct processing variables. It may be beneficial to have individual scales of the processing steps, such as when multiple layers are deposited on a single processing work piece, or when different steps of a process are exposed to different processing conditions (eg, pressure, temperature) and many more). The processing controller 150 controls the operating parameters of the manufacturing machine 110. The process controller can control the process chamber temperature of the manufacturing machine 110, the vacuum, the gas injection system, and the like. The process controller 150 can store a one or process recipe 170. Each recipe 170 defines the operational parameters of the machine 110 on each step of the process. In a specific embodiment, recipe 1 70 is loaded into manufacturing machine 110 by process controller 150. The data communication link 160 can include a common communication link and it is wireless or wired. The data can be transferred between the machine 110, the processing controller 150, and the statistical processing monitoring device in a raw or processed format. In one embodiment, a Semiconductor Equipment Standard (SECS) interface can be used. In other embodiments, a statistical model, a high-speed SECS message service (HSMS) interface, etc., may be used. The statistical processing monitoring device 105 may be a single server that is analyzed from the manufacturing machine. The device 1 5 5 and the processing controller enter the processing data. In addition, the statistical processing monitoring device 105 can be a multi-server and/or a computer. In a specific embodiment, the statistical device 1 〇 5 includes an error detector 1 2 5, an error detector 1 30, and a tool such as, and the degree is divided into a plurality of air pumps. Manufacturing can also be manufactured 105 communication general name. For 150 contains Supervised Errors 12 200811978 Reporter 150. The statistical processing monitoring device 1〇5 175. In a specific embodiment, the statistical processing is monitored in one or more processing controllers 150. In addition, the device 105 can also be a distinguishable and/or independent device. The storage device 175 includes a multivariate statistical model 135 for processing measurement data, an error feature 140, and in a specific embodiment, The storage device 175 is one of the computers or servers of the system 105. The single storage device device 1 75 can be external to the statistical processing monitoring device embodiment. The storage device 175 includes multiple storage devices. The redundant processing measurement data (processing data) of the backed up data can be stored in the library 120. The stored processing data can be used for each of the displays 110 and for drift and propensity on such manufacturing machines. In a specific embodiment, the multivariate statistics stored to produce one or more of the following are generated, and the multivariate statistical model 丨3 5 can be stored. The error detector database 1 4 contains multiple signs, which are described further below. In a specific sensor database 1 40 is an associated database. For example, the repository 140 may include a misclassification table, which may store classes, and include a related error feature, which may store faults. 0 also includes the storage device device 105 being included in the statistical processing monitoring device library 120, one or more cautions. Category 1 4 5. In the processing of monitoring equipment. In addition, the storage 105. In a specific device, the storage copies. In the process of processing measurement data for the processing of such manufacturing machines, etc., the model 1 3 5 is used. Once in the misclassification and error specific case of the storage device 175, the error detection 'error detector' - the definition of the error classification feature of the list 13 1311117978 In one embodiment, a green period is used to receive multiple multivariate Statistical model data. The training produces one or a collection of processing operations on the machine that are known and/or controlled to include a particular manufacturing process. In training i specific manufacturing, + self-processing and processing data can be used to generate statistics (such as collecting variable arrays, etc.). These statistics collect collectable melons (3)1^), variables, and specific treatments on a particular machine—or y slaves

135. - The error characteristics of the initial set are based on the training === model. The data can also be created and added to the error detector: 1 collection of 14 episodes. The processing error is characterized by - (the characteristics of the processing of each error::: error: the levy can be the processing data including the contribution to the (4) 4 temple error: two lists, forms, or other data structures. In a particular embodiment, each multivariate statistical model is applied to a manufacturing-only machine. The processing data from two or more manufacturing machines 110 of a matching machine type can be aggregated to establish an operationally usable Or a single error detection mode/type (multivariate statistical model) of one or more of the above manufacturing machines. Furthermore, one of the error detection models developed for a first manufacturing machine can be applied to one of the same types. Two machines (for example, the same model) Each variable pattern of peppers 1 3 5 contains one or more model measures. The model is measured as a scalar value, which characterizes a set of processing data and an offset between the models. In a specific embodiment, the model metering includes a squared prediction error (generally referred to as SPE, Qres, or Q), and Hotelling's T2. Model metering also includes Combined measurements (Example 14 200811978 such as Combined Multivariate Index (CMI)). Each of these measurements corresponds to an estimate of the probability that the monitored data being processed has the same statistics as the training data used to establish the model. Different methods. The above statistics and measurements can be calculated according to general statistical algorithms. One or more multivariate models can use the main component analysis (PCA } to convert an M-dimensional processing variable space to the main components that are perpendicular to each other. One of the N-dimensional spaces, where Μ is the number of processing variables, and N is smaller than Μ. PCA calculates a set of Μ eigenvectors and μ eigenvalues, where each eigenvector transforms the variables Data to an individual dimension of one of the main component spaces, and each feature value is a variable that is often expressed by a corresponding feature value. To simplify the main component space (reducing the dimension of the main component space), corresponding to the N largest The N eigenvector of the feature value is held in the model: the other eigenvector is discarded or ignored. Staying in the model The number of components to be N is a parameter of the model selected by the user. The number of the main components (n) can be based on a model that interprets less data changes when a smaller value N is used and when using a comparison The large value N is selected by the transaction between the specified models. Once one or more multivariate statistical models have been generated, they can be used by the error detector 125 to monitor the processing run on the manufacturing machine. The error detector 125 analyzes the processed data by performing various statistical processing monitoring methods, each of which is based on at least one variable statistical model. In one embodiment, the error detector 125 receives directly from the manufacturing machine. 110, sensor 155 and / or processing controller 15 processing measurement 15 200811978 quantity data (processing data). In another embodiment, error detector 125 can receive processing data from processing measurement database 120. In yet another embodiment, the error detector 1 25 receives processing data from both sources. In order to detect an error, the error detector 125 calculates various statistics for the processed data of each process being monitored, and compares the calculated statistics with the corresponding statistics of the appropriate multivariate statistical model. This statistic is compared for a model metric or for multiple model metrics (eg T2, SPE, CMI). An error can be detected if one or more of the models are measured beyond a predefined threshold (referred to as a credit limit or control limit). In a specific embodiment, each model meter has a threshold value selected for the user. The threshold value of the selection may indicate a compromise between the risk of an error warning (if the threshold is too low) and the risk of not detecting an error (if the threshold is too high). Where multiple measures are calculated, if any of the measurements exceeds the threshold value, an error is caused. In addition, a specific error can be indicated only if the specific metering exceeds the threshold value or if the multiple metering exceeds the threshold value. Once an error has been identified by the error detector 1 2 5, the error is analyzed by the error detector 130. Error Detector 1 3 0 compares one of the error and error characteristics collected. Each error signature represents the processing of a particular error. In one embodiment, the error signature 140 is a ranked list of processing variables having a larger statistical contribution to one of the specific errors. The processing variables can be arranged in the order of their respective amounts of individual contributions. Alternatively, the error signature can be a form, tree or other data structure that arranges processing variables based on statistical contributions to an error. Error detection 16 200811978 征 库 该 该 该 该 、 、 、 1 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在 在When there is a high level similarity between the one of the error exceptions and the current error, the report derives a 0. Each error signature is associated with one or more error classifications stored in the error detection material 140. This misclassification can indicate the actual problem that caused one of the errors or the possible cause of the current error. For example, if the error characteristic indicates that the maximum contribution processing variable is a decane flow rate, the misclassification may indicate that a value that feeds decane into a processing chamber has become abnormal. Error Reporting 1 65 produces an error report indicating which of the error categories 1 45 corresponds to the current error. The error report can be transmitted to one or more clients that are networked to the statistical processing monitoring device 105 (not shown, and can be like a local computer, a remote computer, a personal digital assistant (PDAs), a pager mobile phone, etc. Wait). Error Reporting 1 65 can also shut down the manufacturing machine 1 10 0. Warning a machine, or perform other appropriate actions. Figure 2 depicts a flow diagram of one embodiment of a method 200 of generating an error classification. The method can be performed by processing logic, which can include hardware (eg, circuitry, dedicated logic, programmable logic, microcode, etc., such as instructions running on a processing device), or a combination of the above. In an example, method 200 can be performed by statistical processing monitoring device 105 of Figure 1. Referring to Figure 2, method 200 begins by obtaining information indicating an error (block 205). The processing data can be manufactured from one or more. Obtained by the instrument, sensor, processing controller, and a processing measurement database. 17 200811978 The processing data includes, for example, process chamber temperature, pressure, gas such as 'if the temperature is too high or too low, the gas flow rate force is different The processing data is currently processed, etc., and the processing data can be collected during a training session or during processing monitoring. The erroneous data can be intentionally caused or the error can be generated unintentionally. The processing data can be triggered by an analysis of the processing data before an error is found to have occurred. In block 210, 'create a new error. Class (In a specific embodiment, the new misclassification is based on one or more of the misclassifications in the library error detection database. The one or more parameters may define a cause of sufficient specificity error. The error classification is established after one of the new error specific errors is established. At block 215, the contribution contribution to the error is determined. The contribution processing variables may be arranged in accordance with their individual contributions, and then referred to as Error contribution. An example of how the processing error contribution can determine the corresponding contribution of the error decision processing variable via any common statistical method. J. Chemometrics 200 1, Chapter 15 7 1 5-742 ) Sergio Valle, Michael J· "Multiblock Analysis with Application to Process Monitoring" by Piovoso et al., which incorporates additional body flow rates, etc., as a corresponding contribution to the decision processing variables of the test error. Unstable, the pressure may indicate a mistake. The actual product is produced in the processing embodiment, the acquisition, and based on the block 2 0 5). The definition of a solid parameter in the collection Was built to identify one or more of misclassification can be individually Gong related quantity corresponding sequence of the variable of a variable. The method for detecting based upon disclosed in [, S · Joe Qin, "On Unifying Decentralized illuminated. The Detective Model Methodology Department 18 200811978 is disclosed in J. Chemometrics 2000, Vol. 14, pp. 725-736, Conlin, EB Martin, AJ Morris et al., "ConHdence Limits For Contribution Plots", This is incorporated by reference. A further exemplary method for determining the corresponding contribution of a detected error to a detected error is disclosed in Chemometrics and Intelligent Laboratory Systems 2000, Vol. 51, pp. 95-1 1 4, Johan A. Westerhuis, Stephen G Gurden "Generalized Contribution Plots in Multivariate Statistical Process Monitoring" by Age K. Smilde et al., which is incorporated herein by reference. Other methods of determining the corresponding contribution can also be used. In a specific embodiment, the error contribution can be independent of one of the statistical methods used to determine the error contribution. Thus, parameters specific to a statistical method (e.g., covariance measure, primary component feature vector, etc.) are not included in the error classification and/or error characteristics associated with the error classification. Therefore, the error classification can be equally applied to appropriate statistical methods (such as statistical methods with adapted models (such as models that time-adjust to a particular parameter). In the case of a statistical method using one of the adaptation models, the model includes a main component analysis (p c A ) in which the number of primary components is adapted and/or a transition from the processing variable space to the primary component space is adapted. In a specific embodiment, the error contribution of the processing variable can be independently determined by two or more different statistical methods (e.g., using a static model and an adaptive model). Because of the different statistical models, it can help to more accurately determine the error contribution of different errors. In block 220, the assignment contribution ranks to the contribution processing variable, whereby a new error signature is generated by 19 200811978. In a specific embodiment, a subset of the contribution variables is selected. The subset may contain such processing variables that the error contribution is greater than a contribution threshold (hereinafter referred to as the significant limit, significantCe limit). The significance limit can be calculated according to various methods, such as Qin, c〇nlin, and Westerhuis as discussed above. The ranking of the scoring numbers can then be assigned to the respective processing variables in the subset based on the order of the relevant amounts of their individual contributions. The boundary is erroneously limited. In addition, if it is used to determine the new error, the number of errors is limited, and each of the processing variables can be limited to a difference of at least. The change

The processing variables outside the subset are selected (the error contribution is less assigned to one or none of the error contributions or may be omitted from the error. In a specific embodiment, the significance boundary is predetermined] the significant limit may be The use of one or more statistical methods (eg, one of the individual error contributions of each of the processing variables) is used to produce the eigenvalues. The use of a significant limit can be given to the error (all processing variables for the error) The detection of the contribution system) enhances the noise immunity. An example of including a significant limit is described in Tables 1 and 2 below. In a specific embodiment, it can be included in a new number of error features. There are limits on the number. Therefore, as long as the processing variables combine the error contribution of the significant limit, any of the processing variables are included in a new error feature. Alternatively, an upper and/or lower bound is contributed. Each processing variable is on a number of error features. Ghost 2 2 5 processing logic determines whether the contribution processing variable has an individual error contribution to a threshold value, and then refers to the change boundary boundary can be ' The user selects or automatically selects. The change 20 200811978 is a fixed value, or it may be a correlation value (eg, a ratio of the error contribution based on one of the processing variables). In a particular embodiment A statistical trust range is calculated for the error contribution of each processing variable. The change bound may be based on the calculated statistical trust range of the processing variables. In a specific embodiment, if the processing variables have overlapping trust ranges The process variables vary to be less than the change limit. If the process variables differ to be less than the change limit (e.g., have overlapping trust ranges), the method proceeds to block 23 0. If the process variables are not different to less than the change Limit, the method proceeds to block 2 3 5. In a particular embodiment, the process variables are or not different to the change limit, the method proceeds to block 2 3 5. At block 23 0, the contribution ranking range is assigned To one or more contribution processing variables. Each contribution ranking range contains a contribution to each process variable that is less than the variation limit. Each of the processing variables is assigned a ranking range that includes both the processing variable itself contribution ranking and the other contribution ranking including the processing variables. In a particular embodiment, the ranking range is a continuous numbering A range of contribution rankings. For example, a ranking range may be 1-2, which includes a contribution ranking 1 of a first processing variable and a contribution ranking 2 of a second processing variable. Different processing variables may have the same or overlapping rankings. Scope. Defining an error feature by substituting an absolute contribution ranking or ranking in addition to an absolute contribution ranking may improve noise immunity (eg, where one of the correlation variables between the various processing variables may be exchanged by expected statistical changes) An example of an error feature with a range of ranges is described in Tables 3 and 4 below. 21 200811978 At block 2 3 5, the new error signature is stored in the error library. The stored error signature is associated with the new error. In a specific embodiment, the contribution values (e.g., 〇·9, 0.5, etc.) of the processing variables of the error feature are not stored, and the ranking is instead contributed. (eg 1, 2, 3, etc.). Alternatively, the actual contribution is stored, or the contribution value and the contribution ranking can be stored. Dealing with variables contributing to errors Rank A 0.9 1 B 0.8 2 C 0.4 3 D 0.2 none E 0.07 none F < 0.02 none G < 0.02 none Η < 0.02 none Table 1: First error characteristics Table 1 Description One of the first features produced by the method 200 of FIG. According to a first-order record analysis method, the statistical contributions of the process variables A, ED, and E are individually determined to be 0 · 9 ' 0 · 8, 0 · 4, 0.2, and the remaining processing variables F, G, and The statistical contribution was determined to be 0.02. These processing variables are in the order of their contribution to the wrong amount. The first error signature has a significant limit of 〇 · 3, so the variables A, B, C are considered to contribute to the error and are part of the first fault. The variables D to 处理 are excluded from the first erroneous feature because their error contribution is less than the limit. Measurement class. The actual existence can be erroneously C, 0.07° less than the row, processing t characteristics: significant 22 200811978 processing variables contribution to the error contribution A 0.9 1 B 0.8 2 C 0.4 -3 D 0.2 none E 0.07 none F < 0.02 None G < 0.02 none Η < 0.02 none Table 2: Second Error Characteristics Table 2 describes one of the second error characteristics produced by the method 200 of Figure 2. The same processing material used to generate the first error signature of Table 1 is used to generate the second error signature. These processing variables are ranked in the order in which they contribute to the amount of an error. The second error signature has a significant limit of 〇 · 1, so the processing variables A, B, C, D are considered to contribute to the error. The contribution range of the processing contribution to the error trust contribution ranking range is 0.9 0.92-0.98 1 1-2 B 0.8 0.73-0.87 2 1-2 C 0.4 0.45-0.55 3 3 D 0.2 0.15-0.25 4 4 E 0.07 0.06-0.11 none None F < 0.02 N/A none none G < 0.02 N/A none none 23 200811978 Η < 0.02 Ν/Α none none Table 3: Third error feature with ranking range Table 3 describes the method according to Figure 2. 200 produces one of the third error features. The third error signature includes a ranking range determined by the individual trust ranges of the processing variables. The line named "trust range" shows the upper and lower bounds of the trust contribution of the error contribution of the various variables. Since the lower bound (0.8 2 ) of the trust range of the variable A is lower than the upper bound of the trust range of the variable B (0 · 8 7 ), the trust ranges of the processing variables A and B overlap. Thus, variables A and B are processed, which have an individual contribution ranking of 1 and 2 based on the calculated error contribution, and are assigned a ranking range that includes the rankings of the contributions of each of the others (each is assigned a ranking of 1-2) range). The third error signature has a significant limit of 0·1, and thus the processing variables A, B, C, and D are considered to contribute to the error and are part of the first error feature. The variables E to 处理 are excluded from the first erroneous feature because their erroneous contribution is less than the saliency limit. In a specific embodiment, the processing variable is included in the error signature because the boundary above its trust range is greater than the significance limit of 0.1. In a specific embodiment, at least one list of contribution processing variables (identified by a name or other indicator) and a ranking range of each contribution processing variable are included in the error signature. Alternatively, one or more calculated error contributions, trust ranges, and contribution rankings are also included as part of the error feature. 24 200811978 Processing Contributions to Error Trust Contribution Ranking Variables Range of Terms A 0.9 0.82-0.98 1 1-2 B 0.8 0.73-0.87 2 1-3 C 0.7 0.65-0.75 3 2-3 D 0.2 0.15-0.25 4 4 E 0.07 0.06-0.11 none none F < 0.02 Ν/Α none none G < 0.02 Ν/Α none none Η < 0.02 Ν/Α . none none Table 4 ··The fourth error feature with ranking range Table 4 Description The second error feature generated by the method 2 00 of Figure 2. The fourth error signature is the same as the error feature of Table 3, except that the calculated error contribution and ranking range of the variable C are processed. In a specific embodiment, the trust range of the error B of the processing variable B overlaps both the trust range of the variable A and the trust range of the variable C. Therefore, the ranking range of variable B includes the contribution ranking of variable A (1) and the contribution of variable C (.3) and its contribution ranking (2). Therefore, the processing variable variable B has a ranking range of 1-3. Since the processing variable A has a trust range that overlaps the trust range of the variable B, the rational variable A has a ranking range of 1 - 2. Similarly, since the trust range of the processing variable C overlaps the trust range of the processing variable B, the processing variable C has a ranking range of 2-3. Figure 3 depicts a flow chart of a particular implementation of the method 300 of detecting errors via the use of error signatures. The method can be performed by processing logic, which can include hardware (eg, circuitry, dedicated logic, programmable logic, microcode, etc.), software (eg, instructions running on a processing device), 25 200811978 or a combination thereof . In a specific embodiment, method 300 can be performed by statistical processing monitoring device 105 of FIG. Referring now to Figure 3, method 300 begins with processing logic detecting an error (block 3 0 5). In one embodiment, the error is detected based on processing data received from one or more manufacturing machines, sensors, processing controllers, and a processing measurement database. In block 301, the processing variable contributing to the error is determined. If a control limit is exceeded, a process variable can contribute to an error that would otherwise contribute to an unexpected and/or unexpected result. At block 3 15 5, processing logic determines the contribution of the contribution processing variable. In one embodiment, the processing logic ranks the processing variables in the order of the correlation amount of their individual contributions to the detected errors. The detected error ranking may or may not include the value of the error contribution of the processing variable. In a specific embodiment, the detected error ranking is only a sequential list of the processing variables. For example, if you are not able to deal with the abilities of the bEOs of variables A, B, C, D and E, the contribution is limited and the AD tribute is wrong, 4, the error is wrong. The wrong or fruit row 5 'A' should be divided as 0., to the ranks, 9'B is set in the name E ο table, the number of small rows is limited. 0. The sequence of the wrong circle is ^ _ _ _ _ _ _ ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° ° In the name of the tribute, in the name of C, the order of B can make the sex can be ranked. 2 is the miscalculation of the decimal error G should be wrong with the name of the micro-error, the limit of the measurement limit Everywhere in the investigation, the wrongness of the investigation, the amount of each sexual error, the test of the test, the value of the test, the value of the test, the zero-detection, the detection of one or the other. 26 200811978

At block 3 2 0, the processing logic determines 4 to detect an error. It will be achieved by comparing the debts. In a specific embodiment, one of the error characteristics is one of the error characteristics, and the ranking of the error classification name and one of the processing variables having one of the ranking ranges is one of the wrong rankings. Within the scope, each one that exactly matches the error characteristic that the detected error can meet multiple error characteristics is reported. The error is advertised and ranked if it is the subject of the investigation. When comparing the levies, if the traits of the erroneous features are found to be detected, the error is detected, the error is detected, and the error is detected. The same processing variable is detected in a specific embodiment. In one such occurrence, if a match is detected between one of the current error error rankings and an existing error feature, the method proceeds to square 35325. If no match is detected with an existing error feature, the method proceeds to block 34. In a specific embodiment, at least one of the coincident error features is a mixed error feature. A mixed error feature for an inclusion is targeted to different statistical models

Independently calculated independent contribution rankings and/or ranking ranges. For example, one of the mixed sigma error features, the first processing variable, or a contribution ranking that is 1 by a first statistical model "circumference model; and has a second statistical model (eg, a static force) The model is calculated as a ranking of 2 beneficiaries. π ^ . Therefore, during processing monitoring, the detected error contribution will include a ranking that is determined by each of the appropriate statistical models of the skewed processing variables. This will be facilitated by the fact that different statistical models can more accurately determine the erroneous contribution of these processing variables in response to the ^ and + cheek errors. 27 200811978 Processing Variables First Model Ranking Range Second Model Ranking Range A 1-2 3-4 B 1-2 3-4 C 3 1 D none 2 E none none F none none Table 5: Hybrid Error Feature Processing A model second model number ranking range of the range A 2 3 B 1 4 C 3 1 D 4 2 E none none F none none

Errors Detecting Logic Errors Table 6: Detected Errors Using First and Second Statistical Models Table 5 describes an example of a hybrid feature in accordance with one embodiment of the present invention. Table 6 describes an error in a hybrid error ranking generated by the simultaneous use of two models in accordance with another embodiment of the present invention. As shown, using two different models to independently determine the ranking range and/or contribution ranking of the same material can produce different results. The error characteristics of Table 5 show that all of the detected errors in Table 6 are true, and the error contribution rankings in Table 6 of the processing variables of the 2008 200811978 model are based on the model of the mixed error feature in Table 5 and The corresponding processing variable is within the ranking range. In a particular illustration, it may be detected by a first model, but not by a second model. Therefore, the error contribution of the processing variable of the second model may be small (e.g., below the significance limit) and thus not used for classification. In this case, the processing logic can establish a hybrid error signature, wherein each processing variable of the second model has a none ranking range and/or a contribution ranking. Referring to Figure 3, in block 3 2 5, the processing logic identifies an error classification associated with the error-compliant feature. In a specific embodiment, each error feature is associated with a single error classification. In addition, multiple error classifications can be associated with an error signature. This will happen if the two error classifications have the same error characteristics. At block 303, processing logic determines whether any of the compliant error characteristics is associated with multiple error classifications. If an error feature is associated with multiple error classifications, the method proceeds to block 3 3 5. If no error features are associated with multiple error classifications, the method ends. Alternatively, the method can end because any error signature is associated with multiple misclassifications. At block 3 3 5, a record (tally) is reported, the record containing the number of times each of the error categories associated with the error signature is the actual cause of the error. This will have the actual cause of assisting a user in identifying a current error. The record can be stored in the error detection database. In a specific embodiment, after identifying the actual cause of the current error, the actual cause of the 200811978 is entered into the record in the error detection database. At block 340, a new error classification is established. At block 345, a new error signature is generated that can be associated with the new error classification. This new misclassification and the new error signature can be generated in accordance with method 200 of FIG. Alternatively, other methods of generating this new misclassification and new error characteristics can be used. Therefore, when new errors are encountered during the actual processing of the product, new misclassifications and new error features can be added (eg to a multivariate statistical model). In one embodiment, only a single instance of a single error is required to incorporate a new error classification and a related new error feature. Figure 4 depicts a flow chart of a particular implementation of a method 400 of detecting errors via the use of erroneous features. The method can be performed by processing logic, which can comprise hardware (e.g., circuitry, special purpose logic, programmable logic, microcode, etc.), software (e.g., instructions executed on a processing device), or a combination of the above. In a specific embodiment, method 400 can be performed by statistical processing monitoring device 105 of Figure 1. Referring to Figure 4, method 400 begins with processing logic detecting an error (block 4 0 5 ). At block 410, the processing variable contributing to the error is determined. At block 4 15 5, the contribution of the contribution processing variable is determined. At block 420, an error signature that matches the error is determined. In a specific embodiment, an error feature that fully complies with the error is considered. In addition, consider some of the error features that match a detected error. This can be useful, for example, when multiple errors occur simultaneously. In such a case, although one or more error signatures may be associated with an incorrect classification indicating the actual cause of the detected error, the error-free feature will fully comply with the detected error. In a specific 30 200811978 embodiment, if there is no forest, the person is born, and the discovery is fully paid, only some of the error characteristics that meet the detected error are considered. In addition, whether a complete match is found Considering a 4-knife compliance would be useful, for example, to ensure that a user is wary of a serious error associated with a partially compliant error feature (eg, a bug that causes a serious yield loss). At block 425, for the complete / or partially conform to the error feature to determine the match close knives (match cl〇seness s (3) π). A match close to the knife represents a similarity between the feature and a detected error. The match close score can be calculated in various ways. Specifically, the one-shot proximity score is determined by assigning a value of 2 to each processing variable, wherein the detected error ranking of the processing variables is within a proper ranking range of one of the given error characteristics, and via Others have a significant error contribution in both the actual ranking of the currently detected error and the given error feature. A erroneous contribution above the sexual limit) is determined by the value of the variable assignment - 1. In other embodiments, the previous values 2 and i are multiplied by a weighting factor for each processing variable. The ranking is higher for the game. At block 430, the error feature with the lowest matching proximity score is discarded. In a particular embodiment, all error features are discarded except for the X error feature with the highest matching proximity score. In other embodiments, all error features having an error contribution of less than one threshold are discarded. In another embodiment, no error features are discarded. At block 43 5, the identification is related to the error. The traits of the traits are classified as eight categories. At block 440, the identified erroneous features are associated with the faulty severity value (fauU severity values) of the 200811978 class and the coincidence scores are reported. In a particular embodiment, Each error classification contains an error severity value. An error classification with a low error severity value can result in little to no damage to the manufactured product, and The misclassification of the severity value or the product yield is significantly reduced. Therefore, the error severity value can indicate the importance of a user's error. For example, an error feature with one high-match close score but one low error severity value A misclassification or not being concerned. However, a misclassification of an error feature with a low match close score but a high error severity value will cause concern. What does the machine do? Dividing the error feature into multiple error features. For example, if the error classification associated with the error feature can be classified into a misclassification 'or if the original error feature is found to be capable of: correcting the error feature and being corrected - no Correct error detection, #ij分=Features or can be expected. A specific embodiment of the method for splitting an error is a main τ 奵 刀 及 及 及 及 及 及 及 及 及 及 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 。 Figure 5 depicts a flow chart for implementing the error by using the error detection feature. The gastric method can be performed by processing logic: The processing logic can include (d) (eg, circuitry, dedicated logic, mega, microcode, etc.), software (eg, in processing device logic or a combination of the above. In a particular embodiment, the method 5. The execution can be performed by the statistical processing monitoring device i 05 of the Vth. Referring to Figure 5, the method 500 begins with processing logic detection_block 5〇5). At block 510, a process variable contributing to the error is determined. At 32 200811978 block 5 1 5, the contribution of the contribution processing variable is determined. Shiwan block 5 2 0, determines the error characteristics that match the error. At block 525, processing logic determines if the error feature includes a contribution range. If one of the error characteristics is ambiguous, and the phase includes a relevant contribution range, the method proceeds to block 535. If there is no error in the supply - the person has a range of related contributions, the method proceeds to block 530. At block 530, the category classification is identified with respect to the erroneous feature # hoof, and the method then ends. The determined error signature is divided into multiple error signatures in method 5 3 5 '. For example, in one embodiment, the determined metrics are divided into two distinct erroneous features. The two distinctive erroneous features are different by having at least one processing variable, wherein the φ 乂 processing variable has an individual ranking range among the two new erroneous features, the variable in the original erroneous feature Different subsets of the ranking range are shown, for example, in the examples in Tables 7, 8, and 9. At block 540, a new error classification is added for one or more multiple error signatures. The new misclassification can be a subset of the misclassification associated with the original error feature being divided. For example, if the original error is classified as a "defective electrostatic chuck," the new misclassification will be "Defective electrostatic chuck because of particles on wafer backside." "." At block 545, at least one of the original error signatures and their associated error classifications is updated. The update process involves narrowing down one of the error characteristics or multiple locations. The ranking range of the 200811978 arguments, and the narrowing-error classification description. For example, if the original misclassification was originally a defective electrostatic seat, the original misclassification was narrowed to a defective electric station due to residual accumulation. It has an error feature for dividing the basic error of the two processing variables into one error and the most common one, wherein one of the two different processing variables affects one of the two processing variables. An example is the common mis-mode for sharing an electrostatic seat of one of the semiconductor workpieces on a cathode in a half-guide chamber. The two variables that are preferably monitored to detect errors in an electrostatic seat are the turbulence rates injected into a workpiece and a hole in between, and one of the connections between the cathode and an RF power supply. An error in a process control parameter value whose impedance meets one of the capacitances of an adjustable tuning capacitor in the path is caused by excessive residual accumulation on the block will generally be related to both of the process variables. However, one of the errors caused by the particles on the lower end of the workpiece is generally only associated with the turbulence rate. Therefore, the first error sign associated with one of the first misclassifications of a defective electrostatic seat will be split into a second error defined as one of the defective electrostatic blocks created by the particles on the lower end of the workpiece. One of the second errors associated with the classification. This first misclassification will then be redefined as one of the defective electrostatic pads caused by the residual product on the seat. Such as or static of the body erroneous table network 微 微 定 特 特 特 特 特 特 特 A A A A A A A A A A A A A A A A 表 表 表 表 表 表 表 表 表 表 表 表 表 表 表 表 表 表 表1 B 2 C 3 D 4 Table 8 • New error feature based on error feature processing variable ranking range A 2 B 1 C 3 D 4 Table 9: Updated error signature based on error feature partitioning Table 7 description has one One of the wrong range features. Tables 8 and Tables describe two new error characteristics resulting from the error signature of Table 7 in accordance with Method 5 of Figure 5. In the original error signature of Table 7, the variables A and B are assigned a ranking range of 1 - 2, and the original error signature is associated with a misclassification identified as X. The new error in Table 8 is associated with a new error classification Y, and has rankings 1 and 2 for one of the processing changes A and B, respectively. The updated error signatures of Table 9 are linked to the updated error classification X and have rankings 2 and 1 for the processing variable A and one of the contributions, respectively. Figure 6 depicts a diagram of an exemplary form of computing system 600 and a graphical representation of 200811978, in which a set of instructions is executable and used to cause the machine to execute therein. Any one or more of the methods discussed. In alternative embodiments, the machine can be connected (e.g., networked) to a local area network, a corporate network, or other machine in the Internet. The machine can operate as a server or a client in a client-server network environment, or as a peer machine in a peer-to-peer (or decentralized) network environment. The machine can be a personal computer, a desktop computer, a set-top box (STB), a personal digital assistant PDA, a mobile phone, a web application, a server, a network router, a switch or a bridge. Or any machine capable of executing a set of instructions (sequence or non-sequence) specifying the actions taken by the machine. Moreover, when only a single machine is described, the item "machine" should also be employed to include executing a set (sets) of instructions individually or in connection to perform any one or more of the methods described herein. Any collection of machines. The exemplary computer system 600 includes a processing device (processor) 602, a main memory 604 (eg, a read-only memory, a flash memory, a dynamic random access memory (eg, synchronous dynamic random access memory, or Rambus). Dynamic random access memory (etc.), static memory 606 (eg, flash memory, static random access memory, etc.), and data storage device 618, which pass through busbar 63 0 and others Communicate. Processor 602 represents one or more processing devices (e.g., microprocessors, central processing units, etc.) that are generally intended. In particular, the processor 602 can be a Complex Instruction Set Computing (CISC) microprocessor, a Reduced Instruction Set Computing (RISC) microprocessor, or a Very Long Instruction Set (VLIW) microprocessing 36 200811978, or A processor that implements other instruction sets or a processor that can be implemented as a set. The processor 602 can also be one or more specific devices (eg, an application specific integrated circuit (ASIC)), a field programmable logic-digital signal processor, a network processor, or The above similar device 602 is configured to execute processing logic 626 for performing the operations and steps. The computer system 60 includes a network interface device 6 〇 8 The system 600 also includes a video display unit 61 0 (eg one (LCD) or cathode ray tube (CRT)), letters and numbers 6 1 2 (eg keyboard), indicator control devices (eg mouse), to produce device 6 16 (eg speaker). The storage device 618 can include a machine accessible {63 1 'on which can store one or more sets of instructions that can utilize any of the methods or functions described herein. The software 622 can also be 'completed' or during execution via the computer. At least partially resident in the primary memory and/or processor 602, the primary memory 604 and the processing may constitute a machine-accessible storage medium. The software 622 is further permeable to the interface I d 2 in a network 620 It is transmitted and received. ^ Machine Accessible Media 631 can also be used to store a set of negative data structures that define a and a user profile that defines the user directory. The focus and user preferences can also be stored in the computer system in his area 2, such as static. Memory 006. Tian 亥 机器 machine can access the storage medium $ 3 丨 in a sample-specific instruction map processing specific gate array, I. The processing is described in the process. The computer LCD display input device and a signal The storage medium one or more i system 600 memory 604 602 also passed the user preference of the network. The embodiment of the system is shown as a single medium in 200811978, and the item "machine accessible storage medium should be Can be employed to include a single media heavy medium (eg, a centralized or decentralized database, and/or related and server) that can store one or more sets of instructions. The item "machine accessible storage media" should also be One or more methods of the present invention can be performed by a person and media containing a set of instructions capable of storing, encoding or carrying. The item "machine accessible storage" should therefore be employed to include However, it is not limited to solid state memory, magnetic media, and carrier signals. It should be understood that the above description is only intended to be illustrative and not limiting. Many other specific embodiments are known to those skilled in the art. The scope of the present invention is determined by the following description of the invention, and the scope of the invention is to be determined by the scope of the appended claims. 1 is a specific embodiment of a statistical processing monitoring system; FIG. 2 is a flow chart describing a specific example of a method for generating an error classification; FIG. 3 depicts a specific implementation of detecting an error by using an error feature. Flowchart of an example; Figure 4 depicts a flow diagram of another embodiment of detecting errors via the use of error signatures; Figure 5 depicts a flow diagram of yet another embodiment for detecting errors via the use of error signatures; "Body" or more caches using the method of storing media optics as a method of reading should also be implemented in the law 38 200811978 6 The Figure depicts a graphical representation of a machine in an exemplary form of computing system in which a set of instructions is executable and used to cause the machine to perform any one or more of the methods discussed herein. [Main component symbol description] 11 0 manufacturing machine 1 5 5 sensor 1 7 0 recipe 1 5 0 processing controller 1 6 0 data communication link 1 2 5 error detector 1 3 0 error detector 165 error reporter 120 processing measurement database 1 3 5 multivariate statistical model 140 error detection database 175 storage device 1 0 5 statistical processing monitoring device 205 receives processing data indicating error 2 1 0 to establish a new error classification 2 1 5 determination contribution processing variable The corresponding contribution 220 assigns a contribution ranking to the contribution processing variable to generate a new error feature 225 that overlaps the trust range? 39 200811978 23 0 Assignment contribution ranking range to contribution processing variable 235 Store new error characteristics 305 Detection error 3 1 0 Decision contribution processing variable 3 1 5 Determining the corresponding contribution of the contribution processing variable 320 Compliance error feature? 3 2 5 Identify misclassifications that match the characteristics of the error 3 3 0 What are the error characteristics of multiple error classifications? 33 5 indicates the error record (Tally) 3 40 adds a new error classification 345 adds a new error feature 405 detects the error 4 1 0 determines the contribution processing variable 4 1 5 determines the corresponding contribution of the contribution processing variable 420 determines which error feature meets the error 425 Judging the coincidence score that matches the error feature 43 0 discarding the error feature with the lowest match score 43 5 Identifying the misclassification of the coincident error feature 440 Reporting the error classification, coincidence score, and severity value 5 0 5 Detection error 5 1 0 Decision contribution processing variable 5 1 5 Determining the contribution of the contribution processing variable 5 20 Judging the coincidence error feature 5 2 5 The error feature contains the corresponding contribution range? 40 200811978 5 3 0 Identifying misclassifications for coincident error features 53 5 dividing coincidence error features into multiple error features 5 40 adding new error classifications for one or more error signatures 545 updating the partitioned error signatures and associated error classifications 602 processors 626 processing logic 6 04 main memory 622 software 6 06 static memory 608 network interface device 620 network 6 1 0 video display 612 letters and numbers input device 6 1 index control device 6 1 6 signal generating device 6 1 8 Auxiliary memory 631 machine accessible storage medium 622 software 41

Claims (1)

200811978 X. Patent application scope: 1 · A method for detecting errors, comprising: detecting an error; determining one or more processing variables contributing to the error; determining each of the one or more processing variables a corresponding contribution; and determining which of the complex features of the complex number corresponds to the error, an error characteristic conforming to the error, if the corresponding contribution of the one or more processing variables is within the corresponding contribution range of the matching error feature , wherein each of the error characteristics is related to at least one error classification. 2. The method of claim 1, wherein the method further comprises: if the complex error features are not consistent with the error, adding a new error feature for one of the errors. 3. The method of claim 2, further comprising: adding a new error classification; and associating the new error feature with the new error classification. 4. The method of claim 2, wherein the new error feature is added after one of the errors occurs. 5. The method of claim 1, further comprising: dividing a first error feature including a first corresponding contribution range for a first processing variable into a multiple error feature, the multiple error Each of the features has a different corresponding contribution of the first processing variable, and each of the multiple error features is associated with a different error classification. 6. The method of claim 1, wherein the multiple error feature is associated with a single error feature. 7. The method of claim 1, further comprising: tallying each associated error classification as a number of actual errors for one of the specific error characteristics. 8. The method of claim 1, further comprising: determining which of the complex features of the complex number partially coincides with the error, and a portion of the coincidence occurs if the one or more processing variables are At least one of the respective contributions is not within a corresponding contribution range of the erroneous feature; and assigning a match closeness score to one or more of the erroneous features of the plural. 9. The method of claim 1, wherein at least one of the complex error features is a combined error feature having a first error based on a first statistical model The feature is based on a second error feature of one of the second statistical models. 43 200811978 10. A machine-accessible medium containing data that, when accessed by a machine, causes the machine to perform a method, the method comprising: detecting an error; determining one or more contributions contributing to the error Processing a variable; determining a corresponding contribution of each of the one or more processing variables; and determining which of the complex features of the complex number is consistent with the error, an error characteristic conforming to the error, if the one or more processing The corresponding contribution of the variable is within the corresponding contribution range of the matching error feature, wherein each of the error features is related to at least one error classification. 1 1. The machine-accessible medium of claim 10, wherein the method further comprises: if the complex error features are not consistent with the error, adding a new error feature for one of the errors. 1 2. The machine-accessible medium of claim 11, wherein the method further comprises: adding a new error classification; and associating the new error feature with the new error classification. 1 3: The machine-accessible medium as claimed in claim 10, the method further comprising: 44 200811978 comprising dividing a first error feature for a first corresponding contribution range of a first processing variable A plurality of error features, each of the multiple error features having a different corresponding contribution of the first processing variable, and each of the multiple error features being associated with a different error classification. 1 4 - The machine-accessible medium of claim 10, wherein the method further comprises: determining which of the complex features of the plurality of parts is partially in accordance with the error, and a part of the coincidence occurs if the The corresponding contribution of at least one of the one or more processing variables is not within a corresponding contribution range of the erroneous feature; and one or more of the erroneous features that match the score to the complex number are assigned. The machine-accessible medium of claim 10, wherein at least one of the plurality of complex features is a combined error feature, the combined error feature having a first statistical One of the first error features of the model and a second error feature based on one of the second statistical models. 1 6 - a statistical processing monitoring system comprising: an error detector coupled to at least one manufacturing machine for receiving processing data from the at least one manufacturing machine and for processing based on the processing 45 200811978 Detecting an error, the processing data includes a plurality of processing variables; a database for storing complex error features, each of which is associated with at least one error classification; and an error detector Corresponding to the error detector and the database to determine a plurality of processing variables contributing to the complex number of the error, determining a corresponding contribution of each of the one or more processing variables, and determining a complex number Which of the error characteristics corresponds to the error, an error sign that matches the error, if the corresponding contribution of the one or more processing variables is within the corresponding contribution range of the matching error feature. 1 7 · If the error handling feature of the plural is not consistent with the error in the statistical processing monitoring system described in claim 16 of the patent application, the error detector is configured to store a new error feature in the database. 1 8. The statistical processing monitoring system according to claim 17, wherein the error detector is configured to store a new error classification in the data, and correlate the new error feature with the new error classification. In the statistical processing monitoring system described in claim 16, the error detector is further configured to determine that the error feature of the plurality is partially consistent with the error, and a part of the coincidence occurs. The corresponding contribution to at least one of the one or more processing variables is not within a corresponding contribution range of the erroneous feature, and assigning a match close to the score to levy or to confuse the cumber thereof is what is wrong 46 200811978 One or more of the wrong features of plurals. 20. The statistical processing monitoring system of claim 16, wherein at least one of the intermediate complex error features is a combined error signature, the combined error signature having a first based on a first statistical model The error feature and the second error feature based on one of the second statistical models. Its special mistake 47