JP4308113B2 - Data analysis apparatus and method, and program - Google Patents

Description

  The present invention relates to a data analysis apparatus, a method thereof, and a program for deriving a more appropriate set value by adjusting a set value of a control parameter for controlling the apparatus.

  In general, the basic functions and manufacturing processes of manufacturing equipment have been designed in advance at the design and development stage of the equipment, and the design of control methods and control parameters for controlling the equipment has been completed. Specific adjustment guides have also been created through demonstrations using manufacturing process test facilities.

  However, even if the device has been adjusted to a level that achieves the desired performance, such as in a pre-shipment test of the device, control parameters, etc. are required to demonstrate its performance at the start-up stage installed on the actual production line. May need to be readjusted.

  This is because the installation environment differs between the production stage of the manufacturing apparatus and the production line where the apparatus is actually operated, or there are changes due to various stresses caused by accompanying the conveyance.

  For example, when targeting a fine process such as a semiconductor process device, there is generally a difference in the airframe even if the device is the same design, or the difference in installation environment as described above, even if it is the same airframe. There are subtle differences in manufacturing quality due to differences in materials and materials.

  These subtle differences are corrected by adjusting control parameters, etc., but the number of control parameters to be adjusted is very large, and each control parameter to be adjusted is individually shaken and adjusted while observing the behavior. The trial-and-error method of doing so is inefficient and causes the adjustment work to be prolonged.

  In order to improve the efficiency of adjustment work, parameters are adjusted based on empirical rules based on knowledge specific to the target device, but the knowledge is often personal and has empirical rules. There were implementation restrictions, such as the need to have an expert who is familiar with the target device and process.

  In addition, when there are multiple control systems such as feedback control by PID control such as temperature and pressure, sequence control, etc., control methods such as control methods are mixed. Is bad.

  For example, Patent Document 1 discloses a technique for correcting an estimation result while judging from a control parameter estimation process on the assumption that an adjustment rule that has been generalized to some extent exists (given). Yes.

Further, for example, in Patent Document 2, automatic classification is performed on a device that is already in operation using decision tree analysis based on control performance data, and control parameter values are recalculated for each classification. A technique for improving the accuracy of control is disclosed.
JP-A-4-123102 JP 2003-216203 A

  However, with the technique disclosed in Patent Document 1, there is a problem that a lot of experience (trials) is required to obtain a generalized adjustment rule in a new apparatus or process, or a control parameter As the number increases, there is a possibility of interaction between a plurality of control parameters, and it may be difficult for a person to set the adjustment rule itself.

  In addition, since the efficiency of correction of the estimation result depends on a given qualitative rule, even if the correction direction (such as increasing the parameter value) is known, a correction amount should be given quantitatively. I can't. Therefore, a plurality of trials are required until an appropriate correction result is obtained, which is not efficient.

  On the other hand, in the technique disclosed in Patent Document 2, it is necessary to separate control result data, which is important for improving accuracy, into data used for model creation and identification data for evaluating the error. There is. For this reason, when there are a large number of target control model parameters, a corresponding number of data is required, and a certain period of time is required after the operation is started. In other words, a data accumulation period is required to exhibit the effect of improving the accuracy of the control parameter value, and the efficiency is poor as control parameter adjustment for shortening the apparatus start-up period.

  The present invention has been made to solve the above-described problems, and provides a data analysis apparatus, method, and program capable of efficiently and universally deriving highly accurate control parameters. Objective.

In order to achieve the above object, a data analysis apparatus according to the present invention comprises the following arrangement. That is,
A data analysis device for adjusting a setting value of a control parameter for controlling the device to set a more appropriate setting value,
Rule deriving means for deriving a plurality of rules indicating the relationship between the observation information and the determination information regarding the device, based on reference data serving as a reference for adjusting the set value of the control parameter;
A plurality of rules derived by the rule deriving unit , and a reference information generating unit that generates reference information including a set value of a control parameter that satisfies the rule for each of the plurality of rules ;
Among the plurality of rules, the relationship between the observation information and the determination information about the device derived based on the adjustment process data obtained in the process of adjusting the setting value of the control parameter of the device according to the environmental change of the device A similar state estimating means for estimating a second rule similar to the first rule indicating:
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculating means for calculating a set value of the control parameter to be adjusted,
When the setting value of the control parameter is expressed as coordinates in an N-dimensional vector space, the calculation means sets the first vector indicating the control parameter setting value in the first rule and the control parameter setting in the second rule. A third vector indicating the set value of the control parameter used for the next trial is calculated from the second vector indicating the value .

Preferably, the reference information generation unit calculates a representative value of the control parameter included in the reference data satisfying the rule for the plurality of rules, and satisfies the rule with the calculated representative value. It is characterized by generating reference information as a set value of a control parameter .

Preferably, the reference information generation unit divides the plurality of rules into similar clusters by a clustering method based on a set value of the control parameter, and the control parameter included in the reference data belonging to each cluster And generating reference information using the calculated representative value as a set value of a control parameter that satisfies the rule .

Preferably, the calculating means sets Vnext, which is the third vector , to Vnext = 2 × Vnow−V, where Vnow is the first vector and V is the second vector.
It is characterized by calculating by.

Preferably, the rule deriving means derives the plurality of rules using a decision tree composed of nodes and leaves, and manages the result as a rule table.

Preferably, the reference data includes at least a set value of the control parameter, the observation information, and the determination information.

In order to achieve the above object, a data analysis apparatus according to the present invention comprises the following arrangement. That is,
A data analysis device for adjusting a setting value of a control parameter for controlling the device to set a more appropriate setting value,
A plurality of rules indicating the relationship between the observation information and the determination information regarding the device derived based on reference data serving as a reference for adjusting the set value of the control parameter, and satisfying the rule for each of the plurality of rules Storing means for storing reference information including a set value of the control parameter ;
Observation derived based on adjustment process data obtained in the process of adjusting the set value of the control parameter according to the change in the environment of the device among the plurality of rules in the reference information stored in the storage means Similarity state estimation means for estimating a second rule similar to the first rule indicating the relationship between the information and the determination information;
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculating means for calculating a set value of the control parameter to be adjusted,
When the setting value of the control parameter is expressed as coordinates in an N-dimensional vector space, the calculation means sets the first vector indicating the control parameter setting value in the first rule and the control parameter setting in the second rule. A third vector indicating the set value of the control parameter used for the next trial is calculated from the second vector indicating the value .

Preferably, the calculating means sets Vnext, which is the third vector , to Vnext = 2 × Vnow−V, where Vnow is the first vector and V is the second vector.
It is characterized by calculating by.

In order to achieve the above object, a data analysis method according to the present invention comprises the following arrangement. That is,
A data analysis method for adjusting a set value of a control parameter for controlling an apparatus to set a more appropriate set value,
A rule derivation step for deriving a plurality of rules indicating the relationship between the observation information and the determination information regarding the device, based on reference data serving as a reference for adjusting the set value of the control parameter;
A plurality of rules derived in the rule derivation step , and a reference information generation step for generating reference information including a set value of a control parameter satisfying the rule for each of the plurality of rules ;
Among the plurality of rules, the relationship between the observation information and the determination information about the device derived based on the adjustment process data obtained in the process of adjusting the setting value of the control parameter of the device according to the environmental change of the device A similar state estimation step for estimating a second rule similar to the first rule indicating
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A third vector indicating the set value of the control parameter used for the next trial is calculated from the second vector indicating the value .

In order to achieve the above object, a data analysis method according to the present invention comprises the following arrangement. That is,
A data analysis method for adjusting a set value of a control parameter for controlling an apparatus to set a more appropriate set value,
A plurality of rules indicating the relationship between the observation information and the determination information regarding the device derived based on reference data serving as a reference for adjusting the set value of the control parameter, and satisfying the rule for each of the plurality of rules A storage step of storing reference information including a set value of the control parameter in a storage device;
Observation derived based on adjustment process data obtained in the process of adjusting the set value of the control parameter in accordance with the change in the environment of the device among the plurality of rules in the reference information stored in the storage device A similar state estimation step of estimating a second rule similar to the first rule indicating the relationship between the information and the determination information;
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A third vector indicating the set value of the control parameter used for the next trial is calculated from the second vector indicating the value .

In order to achieve the above object, a program according to the present invention comprises the following arrangement. That is,
A program for adjusting a set value of a control parameter for controlling an apparatus and causing a computer to perform data analysis for setting a more appropriate set value,
A rule derivation step for deriving a plurality of rules indicating the relationship between the observation information and the determination information regarding the device, based on reference data serving as a reference for adjusting the set value of the control parameter;
A plurality of rules derived in the rule derivation step , and a reference information generation step for generating reference information including a set value of a control parameter satisfying the rule for each of the plurality of rules ;
Among the plurality of rules, the relationship between the observation information and the determination information about the device derived based on the adjustment process data obtained in the process of adjusting the setting value of the control parameter of the device according to the environmental change of the device A similar state estimation step for estimating a second rule similar to the first rule indicating
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A third vector indicating the set value of the control parameter used for the next trial is calculated from the second vector indicating the value .

In order to achieve the above object, a program according to the present invention comprises the following arrangement. That is,
A program for adjusting a set value of a control parameter for controlling an apparatus and causing a computer to perform data analysis for setting a more appropriate set value,
A plurality of rules indicating the relationship between the observation information and the determination information regarding the device derived based on reference data serving as a reference for adjusting the set value of the control parameter, and satisfying the rule for each of the plurality of rules A storage step of storing reference information including a set value of the control parameter in a storage device;
Observation derived based on adjustment process data obtained in the process of adjusting the set value of the control parameter in accordance with the change in the environment of the device among the plurality of rules in the reference information stored in the storage device A similar state estimation step of estimating a second rule similar to the first rule indicating the relationship between the information and the determination information;
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A third vector indicating the set value of the control parameter used for the next trial is calculated from the second vector indicating the value .

  ADVANTAGE OF THE INVENTION According to this invention, the data analysis apparatus which can implement | achieve derivation | leading-out of a highly accurate control parameter efficiently and universally, its method, and a program can be provided.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1A is a diagram showing a configuration example of a data analysis system according to an embodiment of the present invention.

  A parameter adjustment device (analysis computer) 200 that functions as a data analysis device is connected to each other via devices 300 and 102, which are control parameter adjustment target devices, and a host computer (for example, server) 100, and a network 300.

  Here, the apparatus includes, for example, a semiconductor process apparatus. However, the present invention can be applied to an apparatus or a manufacturing process that has been adjusted in advance and installed on a production line or mass-produced from a demonstration facility. It can be applied to all devices that can undergo environmental changes such as deployment to facilities.

  In the parameter adjustment device 200, the data management unit 202 manages various data such as data collected from the devices 101 and 102 and data generated by the control unit 203.

  For example, the data collection from the devices 101 and 102 may be received directly by the data I / F (interface) unit 201 via the network 300. Alternatively, data collection itself from the devices 101 and 102 may be performed by the host computer 100, and the parameter adjustment device 200 may receive data of the corresponding devices 101 and 102 from the host computer 100. Alternatively, a method of directly connecting the apparatus 101 or 102 and the parameter adjustment apparatus 200 (a form in which the parameter adjustment apparatus 200 is incorporated into the apparatus 101 or 102) and receiving data of the corresponding apparatus 101 or 102 may be used.

  The control unit 203 receives an operation instruction from the user I / F unit 204, and controls the reference information generation unit 205, the parameter setting unit 206, and the data management unit 202 based on the content of the instruction.

  The data management unit 202 includes a reference data 202a, reference information 202b, adjustment process data 202c, and a data storage unit 202d that temporarily stores data used in the reference information generation process of the reference information generation unit 205.

  Here, the reference data 202a includes control parameters for controlling a device used in a trial process in performance development at the time of device development or device manufacture, observation information on the device obtained in the trial process, This is determination information indicating evaluation results such as quality and condition.

  The adjustment process data 202c is control parameters, observation information, and determination information obtained when adjustment work (control parameter readjustment) is performed again after an environmental change such as installation of the apparatus on the production line.

  The reference information 202a and the adjustment process data 202c are used as information for calculating a correction amount for correcting the parameter setting value in the control parameter readjustment process.

  The reference information generation unit 205 includes a rule derivation unit 205a and a representative value calculation unit 205b.

  In the reference information generating unit 205, the rule deriving unit 205a automatically generates (derived) a relationship (rule) between data in the device to be controlled based on the reference data 202a, and uses this as reference information (rule derivation result) 202b. And stored in the data management unit 202. The representative value calculation unit 205b calculates a representative value of the derived rule based on the reference data 202a and the reference information 202b, and additionally stores this as reference information 202b.

  The parameter setting unit 206 includes a similar state estimation unit 206a and a parameter correction amount calculation unit 206b.

  In the parameter setting unit 206, the similar state estimation unit 206a compares the control parameter to be adjusted with the reference information 202b among the control parameters, and estimates control parameter information similar to the current state. The parameter correction amount calculation unit 206b calculates the control parameter correction amount from the processing result of the similar state estimation unit 206a.

  Note that the configuration of the data analysis system in FIG. 1A is an example, and various configurations can be employed depending on the application and purpose. For example, the reference information generation unit 205 and the parameter setting unit 206 in the parameter adjustment device 200 can function as different terminals (reference information generation device and parameter setting device), respectively.

  Hereinafter, a configuration example in this case will be described with reference to FIG. 1B.

  FIG. 1B is a diagram illustrating another configuration example of the data analysis system according to the embodiment of this invention.

  In FIG. 1B, the reference information generation apparatus 400 includes a data I / F unit 401, a data management unit 402, a control unit 403, a user I / F unit 404, a reference information generation unit 405 (a rule derivation unit 405a and a representative value calculation unit 405b. And realizes the function of the reference information generation unit 205 in FIG. 1A.

  In the reference information generation apparatus 400, the data management unit 402 includes reference data 402a acquired from the data I / F unit 401, reference information 402b generated by the reference information generation unit 405, and data used in the process of generating reference information The data storage unit 402c for primary storage is managed.

  The parameter setting device 500 includes a data I / F unit 501, a data management unit 502, a control unit 503, a user I / F unit 504, and a parameter setting unit 505 (similar state estimation unit 505a and parameter correction amount calculation unit 505b). And implements the function of the parameter setting unit 206 in FIG. 1A.

  In the parameter setting device 500, the data management unit 502 manages the adjustment process data 502a acquired through the data I / F unit 501 and the reference information 502b created by the reference information generation unit 405 of the reference information generation device 400. Yes.

  Next, the hardware configuration of various terminals such as the parameter adjustment device 200 of FIG. 1A, the reference information generation device 400 of FIG. 1B, and the parameter setting device 500 will be described with reference to FIG. 1C.

  FIG. 1C is a diagram illustrating a hardware configuration of each terminal according to the embodiment of this invention.

  In FIG. 1C, the hardware configuration of the parameter adjustment device 200 in FIG. 1A will be described. However, the reference information generation device 400 and the parameter setting device 500 in FIG. 1B also have the same hardware configuration. To do.

  A CPU 1201 controls the entire terminal (computer) using programs and data stored in the RAM 1202 and the ROM 1203, and executes each process described later performed by the parameter adjustment apparatus 200 to which the present computer is applied. .

  A RAM 1202 has an area for temporarily storing programs and data loaded from an HDD (Hard Disk Drive) 1204 and a work area used by the CPU 1201 to perform various processes.

  A ROM 1203 stores various programs such as a computer boot program and BIOS. Reference numeral 1204 denotes an HDD (Hard Disk Drive), which stores various data such as an OS (Operating System) and programs and data for causing the CPU 1201 to execute processes to be described later performed by the parameter adjustment apparatus 200 to which the computer is applied. These are read into the RAM 1202 and executed under the control of the CPU 1201 as necessary.

  Reference numeral 1205 denotes a network I / F (interface) for connecting a computer to the network 300. The computer can perform data communication with an external device via the network I / F 1205.

  A storage medium drive 1206 includes a CD-ROM, CD-R / RW, DVD-ROM, DVD-R / RW, DVD-RAM, and the like.

  Reference numeral 1207 denotes a pointing device, and 1208 denotes a keyboard, which can input various instructions to the CPU 1201. These function as an input unit (user I / F unit).

  Reference numeral 1209 denotes a video I / F (interface), to which a display device 1210 is connected. The display device 1210 includes a CRT, a liquid crystal screen, and the like, and has a function of displaying information such as characters and images on the display screen based on a signal sent via the video I / F 1209.

  Reference numeral 1211 denotes an external device I / F (interface), which is a port for connecting a peripheral device to the computer. Through this external device I / F 1211, the computer can be configured with various interfaces such as SCSI, USB, IEEE 1394, etc. capable of transmitting / receiving data to / from peripheral devices, and usually a plurality of external devices I / F. The connection form with the peripheral device may be wired or wireless.

  A system bus 1212 functions as a bus for connecting the above-described various components to each other.

  Next, with reference to FIG. 2, the concept of the adjustment target device that is the processing target of the parameter adjustment device 200 will be described.

  In the following description, a case where processing is executed by the parameter adjustment apparatus 200 will be described as an example.

  FIG. 2 is a diagram for explaining the concept of the control parameter adjustment target device of the parameter adjustment device according to the embodiment of the present invention.

  FIG. 2A schematically shows input / output information with respect to the adjustment target apparatus when there is one control parameter adjustment target apparatus (for example, the apparatus 101).

  In FIG. 2A, the control parameter is information for giving (inputting) a control instruction to the apparatus, and processing is performed by a predetermined control method based on the control parameter. However, at this time, since there is an input of an error (effect of environment, disturbance, etc.), quality variation occurs. In addition, observation information such as a processing state, various control amounts, and environmental conditions inside and outside the apparatus is collected and output by a sensor. Furthermore, quality measurement / determination information obtained as a processing result is output.

  As for the quality measurement / determination information, information output from an inspection device (not shown) installed in a subsequent process may be used in addition to the case of being output from the control parameter target device.

  On the other hand, FIG. 2B schematically shows input / output information for each control parameter adjustment target device when there are a plurality of control parameter adjustment target devices (for example, the devices 101 and 102).

  In FIG. 2B, as in the case of FIG. 2A, there are input / output of control parameters, errors, and observation information. Further, the quality measurement / determination information is output as a processing result of all the control parameter adjustment target devices, not for each control parameter adjustment target device.

  In this case, the quality measurement / determination information may be information output from an inspection apparatus (not shown) installed in a subsequent process, as well as output from the control parameter target apparatus.

  Next, a configuration example of the reference data 202a and the adjustment process data 202c managed by the data management unit 202 will be described with reference to FIG.

  FIG. 3 is a diagram showing a configuration example of the reference data and the adjustment process data according to the embodiment of the present invention.

  FIGS. 3A and 3B show examples of the configuration of the input / output information obtained in FIGS. 2A and 2B, respectively. Each record shown in FIG. 3A and FIG. 3B means one trial (test run). As attributes, an ID for identifying the record, a control parameter that is an essential attribute, and an observation It consists of operating conditions for recording information, determination information, and supplementary information.

  Here, the operation condition is an attribute relating to a product ID such as a device ID for identifying a control parameter adjustment target device, a manufacturing method (process) classification, and a material classification.

  Observation information is information collected by detectors such as sensors installed inside and outside the device, and is obtained by directly and indirectly measuring the amount of control and the factors that may affect the quality. It shows the actual state of the process.

  Further, the determination information is a result of evaluating a control parameter adjustment target device, determination of process behavior and control amount, quality of processed product, and the like. Here, the state values of “NG” and “OK” are shown as the determination information, but a category type composed of numerical values or a plurality of category values, or a combination thereof may be used.

  Next, the concept of the reference data and the adjustment process data will be described with reference to FIG.

  FIG. 4 is a diagram for explaining the concept of the reference data and the adjustment process data according to the embodiment of the present invention.

  The reference data is data obtained by accumulating trial processes in device development and performance tests during device manufacture.

  On the other hand, the adjustment process data is data obtained when an adjustment operation (readjustment of control parameters) is performed again after an environmental change such as installation of the apparatus on the production line.

  These two data are collected with almost the same configuration as shown in FIG. In addition, the control parameters, observation information, and determination information of essential attributes are common to the two data.

  Here, due to the influence of the environmental change, it is assumed that the reference information and the adjustment process data have different observation information and determination information even if the control parameter values are exactly the same. This is because there is a slight difference in the behavior of the device depending on the control parameter.

  Therefore, in the present invention, in a control parameter readjustment phase after an environmental change, it is possible to provide a control parameter readjustment operation with a smaller number of trials as a general adjustment work support device. The details will be described below.

  Next, reference information generation processing executed by the parameter adjustment apparatus 200 will be described with reference to FIG.

  FIG. 5 is a flowchart showing reference information generation processing executed by the parameter adjustment apparatus according to the embodiment of the present invention.

  The reference information generation process is executed by the CPU 1201 of the parameter adjustment apparatus 200 under the control of a program that implements the rule derivation unit 205a and the representative value calculation unit 205b of the reference information generation unit 200.

  As described with reference to FIG. 4, the parameter adjustment device 200 first references the control parameter adjustment process data (control parameter adjustment <first time> shown in FIG. 4) in the performance test at the time of device development and device manufacture in step S <b> 101. The data is stored in the data management unit 202 as data. Next, in step S102, the rule deriving unit 205a automatically derives a relationship (rule) between the observation information and the determination information based on the reference data. In step S103, the representative value calculation unit 205b calculates a representative value corresponding to each derived rule.

<Rule derivation method>
Next, details of the rule derivation method in step S102 will be described.

  Considering the observation information and the determination information included in the reference data, it is assumed that the relationship (rule) between the two is estimated. The observation information and the determination information are used because they can be regarded as having an objective common scale even when the environment changes. Of course, the observation information and the determination information should be designed and implemented so that they can be objectively evaluated originally.

  Note that control parameters are not used in rule derivation. This is because there is no objective evaluation scale for control parameters.

  The rule derivation in this embodiment is generally performed using one of data mining techniques called decision tree analysis, regression tree analysis, and the like. In these methods, the difference in the values of the dependent variables can be automatically learned (machine learning) from among a large number of explanatory variable groups, and finally classified into a tree shape. The classification result can be expressed as an IF THEN rule.

  The difference between decision tree analysis and regression tree analysis is generally called decision tree analysis when the dependent variable is a category, and regression tree analysis when the dependent variable is a continuous value. Hereinafter, these are collectively referred to as a decision tree.

  In addition, the following are references for decision trees and regression trees.

L. Breiman, JHFriedman, RAOlshen and CJStone (1984)
"Classification and Regression Tree" Wadsworth
JRQuinlan (1993)
"C4.5: Programs for Machine Learning" Morgan Kaufmann

  In the present invention, a dependent variable is assigned from the determination information and an explanatory variable is assigned from the observation information, and the difference in the state of the determination information is described (classified) as an IF THEN rule that represents the condition of the observation information. In addition, many methods have been proposed for decision tree analysis / regression tree analysis. However, in the present invention, the method is not specified, and the reference data and adjustment process data 2 are used in the form of IF THEN rules. Anything can be used as long as it derives the relationship between the two.

  Here, supplementary explanation will be given for observation information which is an explanatory variable used for rule derivation.

  As the observation information, information that is considered to be related to the target process and the operation of the apparatus, such as a control amount of each control target and an environment monitor inside and outside the apparatus, may be used. In many cases, sensor information from sensors is recorded as time-series data. In this case, the sensor information may be converted into a form that allows decision tree analysis. For example, it is possible to use a technique described in Japanese Patent Application No. 2003-108186 by the applicant of the present application and convert it into a form that allows decision tree analysis.

  In the present invention, the accuracy of the derived rule is important, but the accuracy of the rule can be increased by using more information by utilizing the sensor information described above. Further, as a result, the number of explanatory variables increases, but it is possible to cope with this by applying a so-called data mining technique.

  Next, an example of a decision tree analysis will be described with reference to FIG.

  FIG. 6 is a diagram illustrating an example of a decision tree analysis according to the embodiment of this invention.

  FIG. 6 shows an example of a decision tree analysis result. The decision tree has a tree structure, and a node having a branch condition and a leaf that is one end of the decision tree are configured by a line connecting the two. The top node of the decision tree is a root node indicating the entire data set, and a plurality of nodes (child nodes) or leaves exist for the root node.

  In the decision tree analysis, a branch is made based on a branch condition at each child node from the root node, and then branches to child nodes that satisfy the condition, and finally a leaf that is an end point in the decision tree. To reach.

  The condition for extracting the data set included in the leaf from the entire data set (root node) is expressed as an IF THEN type rule in which all branch conditions of nodes in the process from the leaf to the root node are combined with AND. Can do.

  Next, a rule table for storing rule derivation results will be described with reference to FIG.

  FIG. 7 is a diagram illustrating an example of the rule table according to the embodiment of this invention.

  The rule table manages rule derivation results. The ID identifies a node (root node or child node) in the decision tree, the ID identifies its own upper node (parent node), the branch condition, and the node. For example, identification information (OK / NG), identification information indicating whether it is a leaf (leaf flag: Y (present) / N (not present)), and the like. Here, when the leaf flag is “N”, it indicates that it is not a leaf, that is, a node.

  The determination information may constitute information indicating the determination state of the dependent variable and, as necessary, information indicating the determination state of the rule, such as the certainty of the determination result and the number of data items included in the node, as supplementary information. . Such supplementary information can be used as an evaluation index when selecting a rule.

<Rule representative value calculation method>
Next, details of the representative value calculation of each rule in step S103 will be described.

  The purpose of calculating the representative value of each rule is to determine a set value as a reference for the control parameter corresponding to each rule. Since the derived rule represents the relationship between the observation information and the determination information, determining the setting value that serves as a reference for the control parameter corresponding to this rule is a device for obtaining a certain observation state and determination result. This means that the control instruction value is calculated. The control instruction value at this time is called a rule representative value.

  First, records included in the rule (records satisfying the node identification rule of the rule table) are extracted from the reference data. The control parameters included in the extracted M records are targeted below.

  In order to simplify the description, if all the control parameters are continuous variables, the N control parameters can be expressed as coordinates on an N-dimensional vector space (control parameter space).

Control parameter vector S = (s1, s2,..., SN)
When the control parameter is categorical, it can be easily converted into a continuous variable by introducing a dummy variable, which is generally performed in the field of statistical analysis.

  Next, a control parameter value set as a representative value is calculated from a set of control parameters included in the M records of interest. The reason for calculating the representative value is to eliminate as much as possible the influence of the control parameter value set considered to be low in reproducibility due to variations due to various errors (particularly accidental errors).

  Here, as the representative value calculation method, the following three methods are conceivable.

<Method 1: List of combinations>
<Method 2: Averaging>
<Method 3: Clustering>
Method 1 is a method in which all combinations of control parameter values included in the target data are listed, and those having a certain occurrence frequency or more are used as representative values (a set of control parameter value sets selected as representatives).

  Method 2 is a method in which an average value, a median value, or a mode value is used as a representative value for each control parameter value included in the target data.

  Method 3 performs clustering on the control parameter values included in the target data and divides the control parameter values into k similar groups. Then, for each of the k clusters, an average value for each control parameter value, Alternatively, the median value and the mode value are aggregated as representatives, and k representative values are selected.

  The proper use of these three methods will be described.

  One feature of the present invention is to provide a suitable technique even when the number of control parameters is several tens or hundreds. From this point of view, Method 1 has a high possibility of selecting a very large number of representative values when the number of parameters is large, and the estimation processing result in “similar state estimation processing” described later tends to become unstable. It is better to use only for the control parameters.

  Since Method 2 determines a representative value for each control parameter, for example, when a similar relationship (observation state and determination result) can be obtained by a combination of a plurality of parameters, it should be considered. The problem that cannot be done occurs.

  Here, a specific example of this problem will be described with reference to FIG.

  FIG. 8 is a diagram for explaining a specific example in which averaging is used as a representative value calculation method according to the embodiment of the present invention.

  FIG. 8 shows an example in which there are two control parameters (S1 and S2). When two control parameter values that give similar observation states and determination results are greatly different, such as P1 and P2, when the control parameter values when they are averaged are used as representative values, similar results are not shown. Probability is high.

  This problem is more likely to occur as the number of control parameters increases.

  Method 3 is to divide the control parameter values into similar ones in order to reduce the adverse effects caused by the control parameter averaging described above. In the N-dimensional space of control parameters, those that are far apart are divided into different clusters. Therefore, compared with the method 2, the combination factor of the control parameter values is also taken into consideration. It is possible to easily cope with an increase in the number of parameters.

  Note that clustering is used in the fields of statistics, pattern recognition, data mining, and the like, and the subject of dividing the data on the N-dimensional space, which is the subject of the present invention, is k- It can be configured by a generally well-known clustering method such as a means method.

  For clustering, a general Euclidean distance is used for the distance that is a measure for evaluating the degree of similarity.

  Euclidean distance D between two points A (a1, a2,..., AN) and B (b1, b2,..., BN) on the N-dimensional space.

  If the unit of each control parameter is different and there is a large difference in the value range of each parameter, the influence of the absolute value varies greatly, but there is a possibility that important control parameter changes may be missed. .

  In such a case, a standardized Euclidean distance that adjusts the value of each control parameter so that each variance becomes 1 may be used.

  Further, as representative information representing each cluster, the center of gravity (centroid) of the cluster, the median, or the like is used.

  Therefore, in the present embodiment, the method 3 using clustering assuming a large number of control parameters is used as a representative value calculation method.

  In this embodiment, as shown in FIG. 9A, a representative value in the control parameter space is determined for data belonging to a node. Then, the representative value and the control parameter determined for each node are registered in the representative value table (FIG. 9B).

  Next, the outline of the control parameter adjustment process after the environment change executed by the parameter adjustment apparatus 200 will be described with reference to FIG.

  FIG. 10 is a flowchart showing control parameter adjustment processing executed by the parameter adjustment apparatus according to the embodiment of the present invention.

  The adjustment process is executed by the CPU 1201 of the parameter adjustment device 200 under the control of a program that implements the similar state estimation unit 206a and the parameter correction amount calculation unit 206b of the parameter setting unit 206.

  First, in step S201, adjustment process data of the current adjustment process is extracted. Here, the adjustment process data to be extracted is usually one in the latest trial (test run), but when it is performed multiple times under the same conditions for error evaluation for confirmation of reproducibility, Multiple cases may be extracted. However, cases with different setting conditions are not mixed. That is, all the control parameter values are the same, and the other operation conditions are the same. The extracted record is hereinafter referred to as an adjustment target record.

  Next, in step S202, the similar state estimation unit 206a executes a similar state estimation process.

  The similar state estimation process is performed by estimating similar nodes (that is, rules) from the rule information created from the observation information and determination information of the adjustment target record and the reference data. Here, the similar state means a state in which the relationship between the observation information and the determination information is similar.

  In other words, regardless of the set value of the control parameter, the node representative value of the node whose relationship between the observation information and the determination information is most similar between the reference data and the adjustment target record is selected.

  Then, the selected node, the node representative value, and the similarity (parameter space distance) at that time are registered in the similarity state estimation table (FIG. 11).

  The node representative value and the adjustment target record can be regarded as being in a state where the performance of the apparatus is almost equivalent. Therefore, in step S203, the parameter correction amount calculation unit 206b regards the difference between the two control parameter values as a shift due to an environmental change factor, and executes control parameter correction amount calculation.

  Next, the details of the control parameter adjustment processing of FIG. 10 will be described with reference to FIG.

  FIG. 12 is a flowchart showing details of control parameter adjustment processing executed by the parameter adjustment apparatus according to the embodiment of the present invention.

  In FIG. 12, the same step number is added to the process common to the process of FIG. FIG. 12 particularly shows details of the similar state estimation process in step S202 of FIG. 10, and in particular, the processes of steps S302 to S308 correspond to step S202 of FIG.

  First, in step S201, the adjustment target record is extracted from the adjustment process data as described above.

  Next, in step S302, for each extracted adjustment target record, the leaf number to which each adjustment target record belongs is obtained using the rule table of the reference information. The leaf number can be acquired by satisfying the node identification rule among the leaf flags of “Y”.

  Next, in step S303, the presence / absence of a leaf to which the adjustment target record belongs is determined based on the presence / absence of the leaf number. If there is no leaf (NO in step S303), the process ends. In this case, since it is considered that there is an abnormal state that is not assumed in the reference data, that is, the preliminary adjustment stage, the cause of the abnormal state of the apparatus is investigated.

  On the other hand, if there is a leaf (YES in step S303), in step S304, all the nodes higher than the leaf node to which the adjustment target record belongs (tracing from the leaf to the root node on the tree structure of the decision tree) The standard deviation or the minimum value as the information about the control parameter distance from all the representative values of the corresponding node from the total value indicating the number of records that include the adjustment target records in each node and the representative value table And information on the distribution of control parameter distances such as maximum value and quantile, and the calculation result is stored in the node candidate list (FIG. 13A).

  Here, the total value may be represented by an appearance ratio (0 to 1) obtained by dividing the number of records by the number of records to be adjusted without using the number of records. Hereinafter, the number of records will be described as an aggregate value.

  In step S305, one node (representative value) is selected from the node candidate list. The node can be automatically selected as described below, but the user may be interactively selected.

  Here, an example of the node candidate list will be described with reference to FIG. 13A.

  FIG. 13A is a diagram illustrating an example of a node candidate list according to the embodiment of this invention.

  In the node candidate list, the node ID, the leaf flag value (Y / N), the total value, the distance information (standard deviation of the distance) from the representative value group, and the trial determination that is the actual determination result of the trial of the adjustment target record ( NG / OK), rule determination information extracted from the rule table (NG / OK), reliability, and the like. FIG. 13B shows a specific trial example (four trials) for the node candidate list of FIG. 13A. When a node is automatically selected, the following processing is executed.

<Automatic node selection method>
The search is performed in the following priority order, and if there is one satisfying the condition, it is determined at that time.

Step 1
Leaf attribute = Y
If the maximum aggregate value is 1: Go to step 2. Select the largest aggregate value. If there are multiple: Select the smallest distance information from the representative value group.

Step 2
Leaf attribute = N AND aggregate value> 1
Total value threshold K = 2
For K = 2 to maximum value of total value When there are multiple values: Select the one with the smallest distance information from the representative value group.

Step 3
Leaf attribute = N AND Maximum value of aggregate value = 1
When there are a plurality: Select the smallest distance information with respect to the representative value group Through the above processing, one node is selected from the node candidate list.

  In step S307, one node representative value is selected from the corresponding node representative value group with reference to the representative value table for the node selected by the above processing.

  Next, in step S308, it is determined whether or not the control parameter corresponding to the selected node representative value is valid. This validity determination is performed, for example, by executing threshold processing for the control parameter value. For example, if the control parameter value is greater than or equal to the threshold value, it is determined to be valid, and if not, it is determined not to be valid.

  If the control parameter is not valid (NO in step S308), the process returns to step S307. On the other hand, if the control parameter is valid (YES in step S308), the process proceeds to step S203, and the control parameter correction amount calculation is executed for the selected valid control parameter. Here, in calculating the control parameter correction amount, the following is defined on the N-dimensional vector space with N control parameters.

V1: Coordinate vector of the selected appropriate control parameter ΔV: Correction vector Vnow: Control parameter value of the adjustment target record Vnext: Recommended value of the next control parameter ΔV = Vnow−V1
Vnext = Vnow + ΔV = 2 × Vnow−V1

  The control parameter correction vector ΔV (control parameter correction amount) is a test run performed with Vnow as the original intended determination result. This means the amount of deviation of the control parameter from the intention. Therefore, the next test run is performed based on the control parameter value Vnext corrected by ΔV, and it is determined whether or not the originally intended determination result can be obtained. Then, the above process (test run) is repeated until the originally intended result can be obtained.

  Here, a specific example of calculation of the control parameter correction vector ΔV will be described with reference to FIG.

  FIG. 14 is a diagram for explaining a specific example of calculation of a control parameter correction vector according to the embodiment of the present invention.

  FIG. 14 shows an example of calculation of the control parameter correction vector in the case of two dimensions.

  In FIG. 14, Vnow is the control parameter value of the record to be adjusted, and a control instruction is given so as to be OK in the area C3 on the control parameter space to which the reference data belongs. However, as a result of the test run, NG This shows the case. The region C2 is a virtual region illustrated in the description, and Vnow is considered to belong to the region C2 in which the determination result is NG.

  On the other hand, V1 is the selected appropriate control parameter value, and is a representative value of the cluster C1 belonging to the node whose determination information is NG. Vnow and V1 have different control parameter values but have similar observation information and determination information, and mean that the apparatus or process is determined to be in the same state.

  A vector representing the difference between Vnow and V1 is a correction vector ΔV. That is, although ΔV aimed at the region C3 that is originally OK, it is considered that a shift of ΔV occurred due to the influence of the environmental change and became V1.

  Therefore, the correction is performed by ΔV, and Vnext is set as the control parameter of the next test run so that it is within the virtual region C4 on the control parameter space to which the adjustment process data that becomes OK belongs.

  As described above, in order to facilitate understanding, the description has been given in the case of the two-dimensional case.

  Note that the reason for selecting the node having the smallest distance information with respect to the representative value group when selecting a node in the above-described step S305 is that the subject of the present invention is an apparatus or process that has been adjusted once. Therefore, the situation where the control parameter must be largely corrected is because there is a deep abnormality of the device or a design problem of the control model (insufficient control function for environmental change factors). For this reason, if the apparatus is normal, a subtle difference in control parameters due to environmental changes is obtained, and it is natural to select a control parameter having a short distance for the purpose of similar state estimation.

  As described above, according to the present embodiment, when adjusting the control parameters of the apparatus and the manufacturing process, the control parameter correction amount for obtaining a normal result is calculated, and this is set as the set value at the next trial. By determining, it is possible to reduce the number of trials and shorten the adjustment time until a set value of the control parameter that makes the determination result normal is obtained.

  By providing a general-purpose adjustment method / apparatus that does not depend on the control method and control model for the device and manufacturing process to be adjusted, the adjustment work can be performed more efficiently than by developing the adjustment method individually. .

  In addition, by using a rule acquired by automatic learning based on adjustment process data that has been performed in advance, it is possible to use it from the second trial in the adjustment process after the environmental change. For this reason, the effect can be exhibited at an early stage as compared with a method based on the premise of accumulating a considerable amount of data as in conventional general statistical analysis and data mining.

  Moreover, the following are mentioned when the application aspect of this invention is examined.

(First application phase)
By using the adjustment process in the manufacturing stage of the apparatus as reference data and applying it to the adjustment work after the apparatus is installed on the production line, the efficiency of the adjustment work can be improved and the start-up period can be shortened.

(Second application phase)
For control parameter adjustment due to equipment aging factors, which is one of environmental changes, the adjustment process at the manufacturing stage of the equipment is used as reference data, and the data at the operation stage is used as adjustment process data, thereby improving the efficiency of the adjustment work. be able to.

Further, when there is a sufficient amount of data at the operation stage, it is possible to deal with a more accurate cause of change with time by using the adjustment process data during a period of stable operation as reference data.
(Third application phase)

  The adjustment process data in the experimental equipment or process development line is used as the reference data for the production of equipment for the mass production line or the adjustment work at the start-up of the mass production process line, thereby improving the efficiency of the adjustment work. be able to.

  Although the embodiments have been described in detail above, the present invention can take an embodiment as, for example, a system, an apparatus, a method, a program, or a storage medium, and specifically includes a plurality of devices. The present invention may be applied to a system that is configured, or may be applied to an apparatus that includes a single device.

  In the present invention, a software program (in the embodiment, a program corresponding to the flowchart shown in the figure) that realizes the functions of the above-described embodiment is directly or remotely supplied to the system or apparatus, and the computer of the system or apparatus Is also achieved by reading and executing the supplied program code.

  Accordingly, since the functions of the present invention are implemented by computer, the program code installed in the computer also implements the present invention. In other words, the present invention includes a computer program itself for realizing the functional processing of the present invention.

  In that case, as long as it has the function of a program, it may be in the form of object code, a program executed by an interpreter, script data supplied to the OS, or the like.

  As a recording medium for supplying the program, for example, floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, MO, CD-ROM, CD-R, CD-RW, magnetic tape, nonvolatile memory card ROM, DVD (DVD-ROM, DVD-R) and the like.

  As another program supply method, a client computer browser is used to connect to an Internet homepage, and the computer program of the present invention itself or a compressed file including an automatic installation function is downloaded from the homepage to a recording medium such as a hard disk. Can also be supplied. It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer is also included in the present invention.

  In addition, the program of the present invention is encrypted, stored in a storage medium such as a CD-ROM, distributed to users, and key information for decryption is downloaded from a homepage via the Internet to users who have cleared predetermined conditions. It is also possible to execute the encrypted program by using the key information and install the program on a computer.

  In addition to the functions of the above-described embodiments being realized by the computer executing the read program, the OS running on the computer based on the instruction of the program is a part of the actual processing. Alternatively, the functions of the above-described embodiment can be realized by performing all of them and performing the processing.

  Furthermore, after the program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion board or The CPU or the like provided in the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

It is a figure which shows the structural example of the data analysis system of embodiment of this invention. It is a figure which shows another structural example of the data analysis system of embodiment of this invention. It is a figure which shows the hardware constitutions of the terminal of embodiment of this invention. It is a figure for demonstrating the concept of the adjustment object apparatus of the parameter adjustment apparatus of embodiment of this invention. It is a figure which shows the structural example of the reference | standard data and adjustment process data of embodiment of this invention. It is a figure for demonstrating the concept of the reference data and adjustment process data of embodiment of this invention. It is a flowchart which shows the reference | standard information generation process which the parameter adjustment apparatus of embodiment of this invention performs. It is a figure which shows an example of the decision tree analysis of embodiment of this invention. It is a figure which shows an example of the rule table of embodiment of this invention. It is a figure explaining the specific example at the time of using average value as a calculation method of the representative value of embodiment of this invention. It is a figure for setting the determination of the representative value of the control parameter of the embodiment of the present invention. It is a figure which shows an example of the representative value table of embodiment of this invention. It is a flowchart which shows the adjustment process of the control parameter which the parameter adjustment apparatus of embodiment of this invention performs. It is a figure which shows an example of the similar state estimation table of embodiment of this invention. It is a flowchart which shows the detail of the adjustment process of the control parameter which the parameter adjustment apparatus of embodiment of this invention performs. It is a figure which shows an example of the node candidate list | wrist of embodiment of this invention. It is a figure for demonstrating the production | generation of the node candidate list | wrist of embodiment of this invention. It is a figure for demonstrating the specific example of calculation of the correction vector of the control parameter of embodiment of this invention.

Explanation of symbols

100 Host computer 101, 102 Device 200 Parameter adjustment device 201 Data I / F unit 202 Data management unit 202a Reference data 202b Reference information 202c Adjustment process data 202d Data storage unit 203 Control unit 204 User I / F unit 205 Reference information generation unit 205a Rule derivation unit 205b Representative value calculation unit 206 Parameter setting unit 206a Similar state estimation unit 206b Parameter correction amount calculation unit

Claims (12)

A data analysis device for adjusting a setting value of a control parameter for controlling the device to set a more appropriate setting value,
Rule deriving means for deriving a plurality of rules indicating the relationship between the observation information and the determination information regarding the device, based on reference data serving as a reference for adjusting the set value of the control parameter;
A plurality of rules derived by the rule deriving unit , and a reference information generating unit that generates reference information including a set value of a control parameter that satisfies the rule for each of the plurality of rules ;
Among the plurality of rules, the relationship between the observation information and the determination information about the device derived based on the adjustment process data obtained in the process of adjusting the setting value of the control parameter of the device according to the environmental change of the device A similar state estimating means for estimating a second rule similar to the first rule indicating:
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculating means for calculating a set value of the control parameter to be adjusted,
When the setting value of the control parameter is expressed as coordinates in an N-dimensional vector space, the calculation means sets the first vector indicating the control parameter setting value in the first rule and the control parameter setting in the second rule. A data analysis apparatus that calculates a third vector indicating a set value of a control parameter used for a next trial from a second vector indicating a value . The reference information generation means calculates a representative value of the control parameter included in the reference data that satisfies the rule for the plurality of rules, and sets the calculated representative value as a control parameter setting value that satisfies the rule. The data analysis apparatus according to claim 1, wherein the reference information is generated. The reference information generation unit divides the plurality of rules into similar clusters by a clustering method based on the set values of the control parameters, and calculates representative values of the control parameters included in the reference data belonging to each cluster The data analysis apparatus according to claim 2, wherein the reference information is generated using the calculated representative value as a set value of a control parameter that satisfies the rule . When the first vector is Vnow and the second vector is V, the calculation means calculates Vnext, which is the third vector , as follows : Vnext = 2 × Vnow−V
The data analysis apparatus according to claim 1, wherein the data analysis apparatus calculates the data according to claim 1. The rule derivation means derives the plurality of rules by using a decision tree composed of nodes and leaves, and manages the result as a rule table. The data analysis device described in 1. The data analysis apparatus according to any one of claims 1 to 5, wherein the reference data includes at least a set value of the control parameter, the observation information, and the determination information. A data analysis device for adjusting a setting value of a control parameter for controlling the device to set a more appropriate setting value,
A plurality of rules indicating the relationship between the observation information and the determination information regarding the device derived based on reference data serving as a reference for adjusting the set value of the control parameter, and satisfying the rule for each of the plurality of rules Storing means for storing reference information including a set value of the control parameter ;
Observation derived based on adjustment process data obtained in the process of adjusting the set value of the control parameter according to the change in the environment of the device among the plurality of rules in the reference information stored in the storage means Similarity state estimation means for estimating a second rule similar to the first rule indicating the relationship between the information and the determination information;
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculating means for calculating a set value of the control parameter to be adjusted,
When the setting value of the control parameter is expressed as coordinates in an N-dimensional vector space, the calculation means sets the first vector indicating the control parameter setting value in the first rule and the control parameter setting in the second rule. A data analysis apparatus that calculates a third vector indicating a set value of a control parameter used for a next trial from a second vector indicating a value . When the first vector is Vnow and the second vector is V, the calculation means calculates Vnext, which is the third vector , as follows : Vnext = 2 × Vnow−V
The data analysis device according to claim 7 , wherein the data analysis device is calculated by: A data analysis method for adjusting a set value of a control parameter for controlling an apparatus to set a more appropriate set value,
A rule derivation step for deriving a plurality of rules indicating the relationship between the observation information and the determination information regarding the device, based on reference data serving as a reference for adjusting the set value of the control parameter;
A plurality of rules derived in the rule derivation step , and a reference information generation step for generating reference information including a set value of a control parameter satisfying the rule for each of the plurality of rules ;
Among the plurality of rules, the relationship between the observation information and the determination information about the device derived based on the adjustment process data obtained in the process of adjusting the setting value of the control parameter of the device according to the environmental change of the device A similar state estimation step for estimating a second rule similar to the first rule indicating
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A data analysis method comprising: calculating a third vector indicating a set value of a control parameter used for the next trial from a second vector indicating a value . A data analysis method for adjusting a set value of a control parameter for controlling an apparatus to set a more appropriate set value,
A plurality of rules indicating the relationship between the observation information and the determination information regarding the device derived based on reference data serving as a reference for adjusting the set value of the control parameter, and satisfying the rule for each of the plurality of rules A storage step of storing reference information including a set value of the control parameter in a storage device;
Observation derived based on adjustment process data obtained in the process of adjusting the set value of the control parameter in accordance with the change in the environment of the device among the plurality of rules in the reference information stored in the storage device A similar state estimation step of estimating a second rule similar to the first rule indicating the relationship between the information and the determination information;
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A data analysis method comprising: calculating a third vector indicating a set value of a control parameter used for the next trial from a second vector indicating a value . A program for adjusting a set value of a control parameter for controlling an apparatus and causing a computer to perform data analysis for setting a more appropriate set value,
A rule derivation step for deriving a plurality of rules indicating the relationship between the observation information and the determination information regarding the device, based on reference data serving as a reference for adjusting the set value of the control parameter;
A plurality of rules derived in the rule derivation step , and a reference information generation step for generating reference information including a set value of a control parameter satisfying the rule for each of the plurality of rules ;
Among the plurality of rules, the relationship between the observation information and the determination information about the device derived based on the adjustment process data obtained in the process of adjusting the setting value of the control parameter of the device according to the environmental change of the device A similar state estimation step for estimating a second rule similar to the first rule indicating
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A program for calculating a third vector indicating a set value of a control parameter used for the next trial from a second vector indicating a value . A program for adjusting a set value of a control parameter for controlling an apparatus and causing a computer to perform data analysis for setting a more appropriate set value,
A plurality of rules indicating the relationship between the observation information and the determination information regarding the device derived based on reference data serving as a reference for adjusting the set value of the control parameter, and satisfying the rule for each of the plurality of rules A storage step of storing reference information including a set value of the control parameter in a storage device;
Observation derived based on adjustment process data obtained in the process of adjusting the set value of the control parameter in accordance with the change in the environment of the device among the plurality of rules in the reference information stored in the storage device A similar state estimation step of estimating a second rule similar to the first rule indicating the relationship between the information and the determination information;
Based on the difference between the set value of the control parameter in the second rule of the set value in said reference information control parameter of the first rule, and a calculation step of calculating the set value of the control parameter to be adjusted,
In the calculation step, when the set value of the control parameter is expressed as coordinates in an N-dimensional vector space, the first vector indicating the set value of the control parameter in the first rule and the setting of the control parameter in the second rule A program for calculating a third vector indicating a set value of a control parameter used for the next trial from a second vector indicating a value .

Images