Data processing method, data processing device and storage medium

Abstract

The invention provides a data processing method, a data processing device and a storage medium. The data processing method comprises the steps of acquiring time sequence data, acquiring an evaluation value, classifying and extracting. In the step of acquiring time-series data, a plurality of time-series data acquired by a substrate processing apparatus are acquired. In the step of acquiring evaluation values, evaluation values of each of the plurality of time series data are acquired. In the classification step, each of the plurality of time series data is classified into any one of a plurality of classes on the basis of the evaluation values. In the extraction step, time series data corresponding to any one of the plurality of categories is extracted as extracted time series data.

Description

[0001] This invention is a divisional application of the invention patent application with application number 202110003618.3 filed on January 4, 2021 and invention name “Data processing method, data processing device and storage medium”. Technical Field

[0002] The invention relates to a data processing method, a data processing device and a storage medium. Background Art

[0003] A substrate processing device for processing a substrate is known. Typically, the substrate processing device is preferably used for processing semiconductor substrates. Research is underway to determine abnormalities in the substrate processing device based on data output from the substrate processing device in a time series (for example, see Patent Document 1). In the semiconductor manufacturing device of Patent Document 1, a two-axis coordinate system is created based on relevant data of a specific monitoring object and other associated monitoring objects, and abnormalities in the semiconductor manufacturing device are determined based on whether they are included in an abnormal area.

[0004] [Prior art literature]

[0005] [Patent Document]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2006-228911 Summary of the invention

[0007] [Problems to be solved by the invention]

[0008] However, in the method of Patent Document 1, two monitoring targets having a desired correlation must be identified, but in reality, such identification is difficult, and the state of the semiconductor manufacturing apparatus cannot be easily determined.

[0009] The present invention is made in view of the above problems, and its purpose is to provide a data processing method, a data processing device, and a storage medium for easily extracting specific time series data from a plurality of time series data generated in a substrate processing device. Moreover, another purpose of the present invention is to provide a data processing method, a data processing device, and a storage medium that can reduce the amount of calculation when matching the target time series data obtained from the substrate processing device with the time series data contained in the learning database.

[0010] [Technical means to solve the problem]

[0011] According to one aspect of the present invention, a data processing method includes a step of acquiring time series data, a step of acquiring evaluation values, a classification step, and an extraction step. In the step of acquiring time series data, a plurality of time series data acquired by a substrate processing device are acquired. In the step of acquiring evaluation values, evaluation values ​​of each of the plurality of time series data are acquired. In the classification step, the plurality of time series data are respectively classified into any one of a plurality of categories based on the evaluation values. In the extraction step, the time series data corresponding to any one of the plurality of categories are extracted as extracted time series data.

[0012] In one embodiment, in the step of obtaining the evaluation value, the evaluation value of each of the plurality of time series data is obtained by comparing each of the plurality of time series data with reference data.

[0013] In one embodiment, in the step of acquiring the evaluation value, the evaluation value when the difference between the time series data and the reference data is large is greater than the evaluation value when the difference between the time series data and the reference data is small.

[0014] In one embodiment, the data processing method further comprises the following step: switching display of an evaluation value chart and the extracted time series data, wherein the evaluation value chart represents changes in evaluation values ​​of the extracted time series data.

[0015] In one embodiment, the data processing method further comprises the following step: storing the extracted time series data.

[0016] In one embodiment, the data processing method further comprises the following step: clustering the extracted time series data to classify it into any one of a plurality of clusters.

[0017] In one embodiment, the data processing method further includes the step of switching between displaying an evaluation value graph and the extracted time series data, wherein the evaluation value graph indicates evaluation values ​​of the extracted time series data classified into any one of the plurality of clusters.

[0018] In one embodiment, the data processing method further includes the step of generating a learning database, wherein the learning database is a cluster corresponding to the extracted time series data, and the cause countermeasure information on the substrate processing apparatus corresponding to the time series data is assigned to the cluster.

[0019] Moreover, according to another aspect of the present invention, the data processing method includes a step of acquiring time series data, a step of acquiring evaluation values, a classification step, and a matching step. In the step of acquiring time series data, a plurality of time series data acquired by a substrate processing device are acquired. In the step of acquiring evaluation values, evaluation values ​​of each of the plurality of time series data are acquired. In the classification step, the plurality of time series data are respectively classified into any one of a plurality of categories based on the evaluation values. In the matching step, the object time series data corresponding to any one of the plurality of categories is matched with the time series data contained in the learning database.

[0020] In one embodiment, cause countermeasure information is stored in the learning database corresponding to the time series data. The data processing method further includes the following step in the matching step, that is, when the matching rate between the object time series data and at least one time series data contained in the learning database is higher than a threshold, reading the cause countermeasure information corresponding to the time series data.

[0021] In one embodiment, in the step of obtaining the evaluation value, the evaluation value of each of the plurality of time series data is obtained by comparing each of the plurality of time series data with reference data.

[0022] In one embodiment, in the step of acquiring the evaluation value, the evaluation value when the difference between the time series data and the reference data is large is greater than the evaluation value when the difference between the time series data and the reference data is small.

[0023] In one embodiment, the data processing method further comprises the following step: storing the object time series data.

[0024] According to another aspect of the present invention, a data processing device includes a data acquisition unit, an evaluation value acquisition unit, a classification unit, and an extraction unit. The data acquisition unit acquires a plurality of time series data acquired by a substrate processing device. The evaluation value acquisition unit acquires evaluation values ​​of each of the plurality of time series data. The classification unit classifies the plurality of time series data into any one of a plurality of categories based on the evaluation values. The extraction unit extracts the time series data corresponding to any one of the plurality of categories as extracted time series data.

[0025] According to another aspect of the present invention, a data processing device includes a data acquisition unit, an evaluation value acquisition unit, a classification unit, and a matching unit. The data acquisition unit acquires a plurality of time series data acquired by a substrate processing device. The evaluation value acquisition unit acquires evaluation values ​​of each of the plurality of time series data. The classification unit classifies the plurality of time series data into any one of a plurality of categories based on the evaluation values. The matching unit matches the object time series data corresponding to any one of the plurality of categories with the time series data contained in the learning database.

[0026] According to another aspect of the present invention, a storage medium stores a program that causes a computer to execute the following steps, namely: a step of acquiring time series data, a step of acquiring evaluation values, a classification step, and an extraction step. In the step of acquiring time series data, a plurality of time series data acquired by a substrate processing device are acquired. In the step of acquiring evaluation values, evaluation values ​​of each of the plurality of time series data are acquired. In the classification step, the plurality of time series data are classified into any one of a plurality of categories based on the evaluation values. In the extraction step, time series data corresponding to any one of the plurality of categories are extracted as extracted time series data.

[0027] According to another aspect of the present invention, a storage medium stores a program that causes a computer to execute the following steps, namely: a step of acquiring time series data, a step of acquiring evaluation values, a classification step, and a matching step. In the step of acquiring time series data, a plurality of time series data acquired by a substrate processing device are acquired. In the step of acquiring evaluation values, evaluation values ​​of each of the plurality of time series data are acquired. In the classification step, the plurality of time series data are classified into any one of a plurality of categories based on the evaluation values. In the matching step, the object time series data corresponding to any one of the plurality of categories is matched with the time series data contained in the learning database.

[0028] [Effects of the Invention]

[0029] According to the present invention, specific time series data can be easily extracted from a plurality of time series data generated in a substrate processing device. Furthermore, according to the present invention, the amount of calculation when matching the target time series data obtained from the substrate processing device with the time series data contained in the learning database can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0031] Figure 2 This is a flowchart of the data processing method of this embodiment.

[0032] FIG. 3( a ) is a graph showing time series data used in the data processing method of the present embodiment, and FIG. 3( b ) is a graph showing reference data used in the data processing method of the present embodiment.

[0033] Figure 4(a) is a schematic diagram of the time series data used in the data processing method of the present embodiment, Figure 4(b) is a schematic diagram of the comparison between the time series data in the data processing method of the present embodiment and the benchmark data, and Figure 4(c) is a schematic diagram of the change of the evaluation value in the data processing method of the present embodiment.

[0034] FIG. 5( a ) is a schematic diagram showing classification based on evaluation values ​​in the data processing method of the present embodiment, and FIG. 5( b ) is a schematic diagram showing extraction of time series data in the data processing method of the present embodiment.

[0035] Figure 6 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0036] Figure 7(a) to Figure 7(c) This is a schematic diagram showing an evaluation value graph and time-series data showing changes in evaluation values ​​displayed on a display unit in the data processing device of the present embodiment.

[0037] Figure 8(a) to Figure 8(c) This is a schematic diagram showing time-series data displayed on a display unit in the data processing device of this embodiment.

[0038] Figure 9(a) to Figure 9(c) This is a schematic diagram showing an extracted evaluation value graph and corresponding time-series data displayed on a display unit in the data processing device of the present embodiment.

[0039] Fig.10 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0040] Figure 11(a) to Figure 11(c) This is a schematic diagram showing time series data extracted in the data processing method of this embodiment.

[0041] Fig.12 This is a flowchart of the data processing method of this embodiment.

[0042] Fig.13 This is a flowchart of the data processing method of this embodiment.

[0043] Fig.14 It is a schematic diagram showing a table of a learning database in the data processing device according to the present embodiment.

[0044] Fig.15It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0045] Fig.16 This is a flowchart of the data processing method of this embodiment.

[0046] Fig.17 This is a flowchart of the data processing method of this embodiment.

[0047] FIG. 18( a ) is a diagram showing a waveform of a control signal, and FIG. 18( b ) is a diagram showing a waveform of time-series data.

[0048] Fig.19 It is a schematic diagram showing the values ​​of multiple time series data.

[0049] FIG. 20( a ) is a graph showing the evaluation value distribution, and FIG. 20( b ) is a graph showing the normalization result of the evaluation value distribution.

[0050] Fig.21 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0051] Fig. 22 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0052] Fig.23 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0053] Fig.24 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0054] Fig.25 It is a schematic diagram of a data processing device and a substrate processing device according to this embodiment.

[0055] [Explanation of Symbols]

[0056] 10, 110: Processing Department

[0057] 11: Data acquisition unit (acquisition unit)

[0058] 12: Evaluation value acquisition unit (acquisition unit)

[0059] 13, 113: Classification Department

[0060] 14: Extraction Department

[0061] 15: Clustering Processing Department

[0062] 20, 120, 201b: Storage

[0063] 30, 140: Display unit

[0064] 32: Display screen

[0065] 33: Display area

[0066] 34: Operation area

[0067] 34a: Arrow

[0068] 34b: Button

[0069] 40, 130: Input

[0070] 100: Data processing device

[0071] 111: Data Acquisition Department

[0072] 112: Evaluation value acquisition department

[0073] 114: Matching

[0074] 115: Reading unit

[0075] 200, 200A, 200B: Substrate processing device

[0076] 201: Control device

[0077] 201a: Control Department

[0078] 210: Processing unit

[0079] 220: Chamber

[0080] 230, 270: substrate holding portion

[0081] 231: Rotating base

[0082] 232: Chuck component

[0083] 233: Axis

[0084] 234: Electric Motor

[0085] 235: Base

[0086] 240: Liquid supply unit

[0087] 240b: Valve

[0088] 250: Cup Department

[0089] 260: Processing tank

[0090] A: Period (Rising period)

[0091] Ax: Rotation axis

[0092] B: Period (stable period)

[0093] B1~B3: Boundary line

[0094] BG: Button

[0095] C: Period (decline period)

[0096] CR: Center Robot

[0097] CU: Cursor

[0098] EGE: Extract evaluation value chart

[0099] ETD1, ETD2, ETD3: Extracted time series data Ev, Ev1~Ev25: Evaluation values

[0100] GE: Valuation Chart

[0101] IR: Transfer robot

[0102] L: Treatment liquid

[0103] L0: Initial level

[0104] L1: Level 1

[0105] L2: Level 2

[0106] L3: Level 3

[0107] L4: Level 4

[0108] LB: Fluid Box

[0109] LC: Fluid Chamber

[0110] LP: Loading Port

[0111] R1: First Range

[0112] R2: Second range

[0113] RD: Benchmark Data

[0114] S2 to S24, S102 to S114: Steps t0 to t2: time

[0115] TD, TD1~TD100: time series data Th1, Th2: threshold

[0116] Th3: Specified value

[0117] TW: Tower

[0118] W: substrate

[0119] Wa: upper surface (surface) of substrate W

[0120] Wb: back side (lower surface) of substrate W

[0121] μ: average value

[0122] σ: standard deviation DETAILED DESCRIPTION

[0123] Hereinafter, embodiments of the data processing method, data processing device, and program of the present invention will be described with reference to the accompanying drawings. In addition, in the drawings, the same reference numerals are given to the same or corresponding parts, and the description thereof will not be repeated.

[0124] First, refer to Figure 1 An embodiment of the data processing device 100 of the present invention will be described. Figure 1 Schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200 according to the present embodiment.

[0125] The data processing apparatus 100 performs data processing. Specifically, the data processing apparatus 100 processes the time series data TD generated in the substrate processing apparatus 200 .

[0126] Typically, the substrate processing apparatus 200 includes at least one processing unit 210. The processing units 210 each generate time series data TD.

[0127] The time series data TD is data indicating the time change of a physical quantity in the substrate processing apparatus 200. The time series data TD indicates the time change of a physical quantity (value) that changes in a time series over a predetermined period. For example, the time series data TD is data indicating the time change of a physical quantity related to the processing performed on a substrate by the substrate processing apparatus 200. Alternatively, the time series data TD is data indicating the time change of a physical quantity related to the characteristics of a substrate processed by the substrate processing apparatus 200.

[0128] In addition, the value shown in the time series data TD may be a value directly measured by the measuring device. Alternatively, the value shown in the time series data TD may be a value obtained by performing a calculation on the value directly measured by the measuring device. Alternatively, the value shown in the time series data TD may be a value obtained by performing a calculation on the values ​​measured by multiple measuring devices.

[0129] Typically, the time series data TD includes more than 10 values. The time series data TD may include more than 100 values, or may include more than 1000 values.

[0130] The data processing apparatus 100 may be communicatively connected to the substrate processing apparatus 200. Alternatively, the time series data TD generated in the substrate processing apparatus 200 may be transferred to the data processing apparatus 100 via a storage element.

[0131] The data processing device 100 includes a processing unit 10. The processing unit 10 includes a processor. The processing unit 10 includes, for example, a central processing unit (CPU). Alternatively, the processing unit 10 may include a general-purpose processor.

[0132] The data processing device 100 may also include a storage unit 20. The storage unit 20 stores data and computer programs. The storage unit 20 includes a main storage device and an auxiliary storage device. The main storage device is, for example, a semiconductor memory. The auxiliary storage device is, for example, a semiconductor memory and / or a hard disk drive. The storage unit 20 may also include a removable media. The processing unit 10 executes the computer program stored in the storage unit 20.

[0133] Here, the computer program is stored in a non-transitory computer-readable storage medium, which includes a read-only memory (ROM), a random access memory (RAM), a compact disk-read only memory (CD-ROM), a magnetic tape, a magnetic disk, or an optical data storage device.

[0134] The storage unit 20 stores the time series data generated in the substrate processing apparatus 200. The data processing apparatus 100 and the substrate processing apparatus 200 can be connected to each other for communication, and the time series data TD can be transmitted to the data processing apparatus 100 each time it is generated in the substrate processing apparatus 200, and the storage unit 20 can sequentially store the time series data TD. Alternatively, each time a predetermined number of time series data TD is generated in the substrate processing apparatus 200, the time series data TD can be collectively transmitted to the data processing apparatus 100, and the storage unit 20 can collectively store the time series data TD.

[0135] The processing unit 10 includes a data acquisition unit 11, an evaluation value acquisition unit 12, a classification unit 13, and an extraction unit 14. The data acquisition unit 11 acquires a plurality of time series data. Here, the data acquisition unit 11 acquires the time series data TD from the storage unit 20. In addition, in this specification, the data acquisition unit 11 is sometimes simply referred to as the acquisition unit 11.

[0136] The evaluation value acquisition unit 12 acquires evaluation values ​​for each of the plurality of time series data TD. In this specification, the evaluation value acquisition unit 12 may be simply referred to as an acquisition unit 12 .

[0137] The evaluation value acquisition unit 12 acquires evaluation values ​​for a plurality of time series data TD according to the same evaluation criterion. The evaluation value acquisition unit 12 generates one evaluation value corresponding to one time series data TD according to a specific evaluation criterion. For example, the evaluation value acquisition unit 12 acquires evaluation values ​​for the entire time series data TD according to the same evaluation criterion. Alternatively, the evaluation value acquisition unit 12 acquires evaluation values ​​for a specific part of the time series data TD according to the same evaluation criterion. In addition, the evaluation value acquisition unit 12 may also generate evaluation values ​​by synthesizing a plurality of evaluation criterions using the entire time series data TD or a specific part thereof.

[0138] Typically, the evaluation value acquisition unit 12 generates the evaluation value by calculating the evaluation value of the time series data TD according to the evaluation benchmark set in the data processing device 100. For example, the evaluation value acquisition unit 12 generates the evaluation value by digitizing the waveform of the graph representing the time series data TD. However, the evaluation value acquisition unit 12 may cause other components to calculate the evaluation value and only obtain the calculated evaluation value. The evaluation value may also be generated by comparing the time series data TD with the data used as the benchmark. In addition, it is preferable that the data used as the benchmark is stored in the storage unit 20.

[0139] The number of adoptable evaluation values ​​is smaller than the number of time series data TD. Typically, the number of time series data TD is 100 or more, whereas the number of adoptable evaluation values ​​is 20 or more and less than 100.

[0140] In addition, the upper limit of the time series data TD is not particularly limited. As long as the data processing device 100 can calculate or store, the upper limit of the number of time series data TD can be arbitrary. In one example, the upper limit of the time series data TD can be 10,000, and the upper limit of the time series data TD can be 100 million.

[0141] The classification unit 13 classifies the plurality of time series data TD into any one of a plurality of categories based on the evaluation values. The number of the plurality of categories is less than the number of the time series data TD or the number of adoptable evaluation values ​​generated based on the time series data TD. Typically, the number of the time series data TD is greater than 100, the number of adoptable evaluation values ​​is greater than 20 and less than 100, and the number of the plurality of categories is greater than 2 and less than 10.

[0142] The evaluation value of the time series data TD is generated based on a fixed evaluation criterion. Therefore, the category that uses the evaluation value to classify the time series data TD can also represent the level of the evaluation criterion. In the following description of this specification, the category representing the level of the evaluation criterion is sometimes referred to as a level.

[0143] For example, the classifier 13 classifies the time series data TD into any one of level 1 to level 4 based on the evaluation value. Typically, after the acquisition unit 12 acquires the evaluation value corresponding to the time series data TD based on the time series data TD and the reference data, the classifier 13 classifies the time series data TD into any one of level 1 to level 4 based on the evaluation value. Preferably, the reference data is stored in the storage unit 20.

[0144] In one example, level 1 indicates a category whose evaluation value is the benchmark data or relatively close to the benchmark data, and level 2 indicates a category whose evaluation value is close to the benchmark data after level 1. Furthermore, level 3 indicates a category whose evaluation value is close to the benchmark data after level 2, and level 4 indicates a category whose evaluation value is farthest from the benchmark data than levels 1 to 3.

[0145] Alternatively, the classifier 13 may classify the time series data TD into two categories based on the evaluation value. For example, the time series data TD of category 1 represents the reference data or is relatively close to the reference data, and the time series data TD of category 2 represents the time series data farther from the reference data than category 1.

[0146] The extracting unit 14 extracts, as extracted time series data, time series data corresponding to any of the categories classified by the classifying unit 13. Thus, specific time series data indicating evaluation values ​​in a specific range can be collected from a plurality of time series data as extracted time series data.

[0147] For example, it is preferred that the extracting unit 14 extracts time series data corresponding to the category farthest from the reference data among the categories classified by the classifying unit 13. Thus, it is possible to easily extract time series data corresponding to the evaluation value far from the ideal value or the average value. Therefore, it is possible to grasp the change of the substrate processing device 200 before a substantial abnormality occurs in the substrate processing device 200. At this time, it is preferred to change the control of the substrate processing device 200 so as to suppress the change of the substrate processing device 200 before an abnormality occurs in the substrate processing in the substrate processing device 200.

[0148] As described above, it is preferable that the extraction unit 14 extracts time series data corresponding to the category farthest from the reference data among the categories classified by the classification unit 13. Furthermore, when time series data representing evaluation values ​​far from the reference data are extracted as extracted time series data, it is preferable to further perform clustering processing on the extracted time series data. However, the extracted time series data extracted by the extraction unit 14 may not be the time series data corresponding to the category farthest from the reference data.

[0149] According to the data processing apparatus 100 of the present embodiment, specific time-series data can be easily extracted from a plurality of time-series data generated in the substrate processing apparatus.

[0150] Next, refer to Figure 1 and Figure 2 The data processing method of this embodiment is described. Figure 2 A flowchart showing a data processing method according to the present embodiment.

[0151] like Figure 2 As shown, in step S2, time series data TD is acquired. The time series data TD is data generated in the substrate processing apparatus 200. The acquisition unit 11 acquires a plurality of time series data TD. The plurality of time series data TD may be data generated in the same processing unit 210 or may be data generated in different processing units 210. The plurality of time series data TD each represents a temporal change in a physical quantity of a specific unit in the substrate processing apparatus 200.

[0152] The acquisition unit 11 acquires a plurality of time series data TD from the storage unit 20. Alternatively, the acquisition unit 11 may directly acquire the time series data TD from the substrate processing apparatus 200. For example, the time series data is sequentially generated along with the substrate processing in the substrate processing apparatus 200 or the control of the substrate processing apparatus 200. Next, the process proceeds to step S4.

[0153] In step S4, an evaluation value (score) is obtained. The acquisition unit 12 acquires evaluation values ​​Ev for each of the plurality of time series data TD. For example, the acquisition unit 12 generates evaluation values ​​Ev for each of the plurality of time series data TD according to a specific evaluation criterion. The evaluation criterion may be stored in the data processing device 100 or outside the data processing device 100. One evaluation value Ev is obtained for one time series data TD. However, the evaluation value Ev may represent a plurality of values ​​corresponding to the time series data TD.

[0154] In addition, the range of the maximum value and the minimum value of the evaluation value may be specified. For example, the evaluation value may be normalized so that the difference between the maximum value and the minimum value of the evaluation value is 1.

[0155] The evaluation value is generated according to a specific evaluation benchmark. For example, the evaluation value may be specified such that the higher the evaluation value, the higher the evaluation, and the lower the evaluation value, the lower the evaluation. Alternatively, the evaluation value may be specified such that the higher the evaluation value, the lower the evaluation, and the lower the evaluation value, the higher the evaluation. Alternatively, the evaluation value may be specified such that the closer the evaluation value is to a specific value, the higher the evaluation, or the evaluation value may be specified such that the farther the evaluation value is from a specific value, the higher the evaluation. Next, the process proceeds to step S6.

[0156] In step S6, the time series data TD is classified into any one of a plurality of categories using the evaluation value Ev. The number of categories that can be classified is smaller than the number of values ​​that can be represented by the evaluation value Ev. Next, the process proceeds to step S8.

[0157] In step S8, specific time series data TD classified into the target category among the multiple categories is extracted as extracted time series data from the multiple time series data TD. Typically, the extracted time series data is stored in the storage unit 20. However, the extracted time series data may also be stored in an external storage element.

[0158] The data processing method of the present embodiment is performed as described above. According to the data processing method of the present embodiment, specific time series data can be easily extracted from a plurality of time series data generated in the substrate processing apparatus 200. In addition, the extracted time series data among the plurality of time series data TD acquired in step S2 is stored in the storage unit 20, and the unextracted time series data TD can also be erased and not stored in the storage unit 20. Thus, the required information can be effectively stored.

[0159] The evaluation value of the time series data TD may also be generated by comparing with the reference data. For example, the reference data may also be generated based on a plurality of time series data indicating good substrate processing results. In one example, the reference data is generated by averaging a plurality of time series data indicating good substrate processing results. In the case where the reference data is ideal data or time series data indicating good substrate processing results, the closer the time series data is to the reference data, the higher the evaluation is, and the farther the time series data is from the reference data, the lower the evaluation is.

[0160] Next, the time series data TD and the reference data RD are explained with reference to FIG3(a) and FIG3(b). FIG3(a) is a graph representing the time series data TD, and FIG3(b) is a graph representing the reference data RD. In the graphs of FIG3(a) and FIG3(b), the horizontal axis represents time, and the vertical axis represents physical quantity. For example, the time series data TD and the reference data RD each represent the time changes of 100 physical quantities (values).

[0161] The time series data TD and the reference data RD represent the temporal variation of the same physical quantity, and the shape of the graph representing the time series data TD is substantially the same as the shape of the graph representing the reference data RD. However, strictly speaking, the time series data TD and the reference data RD do not coincide.

[0162] As shown in FIG3(a), in the time series data TD, at the start time point, the physical quantity is approximately zero. When a specified period passes, the physical quantity starts to increase from zero. The physical quantity increases rapidly in a short period and reaches a peak. From then on, the physical quantity maintains a peak for a certain period. Throughout the peak period, the physical quantity has a slight change but is approximately constant. Subsequently, the physical quantity decreases rapidly in a short period and returns to approximately zero.

[0163] In FIG3(b), the reference data RD is generated based on the time series data when the substrate processing in the substrate processing apparatus 200 is progressing in an ideal state. For example, the reference data RD may also be the time series data when the substrate processing in the substrate processing apparatus 200 is progressing in the most ideal state. Alternatively, the reference data RD may also represent the average value of the time series data during the period when the substrate processing in the substrate processing apparatus 200 is progressing in an ideal state.

[0164] As shown in FIG3(b), in the baseline data, the physical quantity is approximately zero at the start time. When the specified period passes, the physical quantity starts to increase from zero. The physical quantity increases rapidly in a short period and reaches a peak. From then on, the physical quantity maintains the peak for the specified period. Subsequently, the physical quantity decreases rapidly in a short period and returns to approximately zero.

[0165] Thus, the graph shown by the time series data TD shows the same tendency as the graph shown by the reference data RD, but strictly speaking, the time series data and the reference data RD are not the same. For example, the rising period until the peak is reached in the graph of the time series data TD is different from the rising period until the peak is reached in the graph of the reference data RD. Moreover, the peak value in the graph of the time series data TD is different from the peak value in the graph of the reference data RD. Furthermore, the peak value in the graph of the time series data TD varies more than that in the graph of the reference data RD.

[0166] Next, refer to Figure 1 to Figure 5(b) 4(a) is a schematic diagram showing time series data TD1 to TD100. Time series data TD1 to TD100 each represent a temporal change of the same physical quantity. For example, the acquisition unit 11 acquires the time series data TD1 to TD100 from the storage unit 20.

[0167] The acquisition unit 12 acquires the evaluation value Ev based on the time series data TD1 to the time series data TD100. For example, the acquisition unit 12 acquires the evaluation value Ev1 to the evaluation value Ev100 based on the difference between the time series data TD1 to the time series data TD100 and the reference data RD. For example, the acquisition unit 12 generates the evaluation value Ev by accumulating the square of the difference between the value of each time series data TD and the value of each reference data RD. At this time, the closer the time series data is to the reference data, the smaller the evaluation value is, and the farther the time series data is from the reference data, the larger the evaluation value is.

[0168] FIG4(b) is a graph showing the result of comparing the time series data TD with the reference data RD. The oblique lines in FIG4(b) represent the difference between the time series data TD and the reference data RD. When the evaluation value is obtained by accumulating the square of the difference, the evaluation value Ev is equivalent to the area of ​​the oblique line portion in FIG4(b).

[0169] FIG4(c) is an evaluation value graph GE showing changes in evaluation values ​​Ev1 to Ev25 corresponding to time series data TD1 to TD25. In FIG4(c), in relation to the accompanying drawings, the evaluation value graph GE shows changes in evaluation values ​​Ev1 to Ev25 corresponding to time series data TD1 to TD25 in time series data TD1 to TD100. As shown in FIG4(c), the evaluation value Ev changes in accordance with the order of the time series data TD. Here, the evaluation value Ev6 is considerably higher than the evaluation values ​​Ev1 to Ev5 and the evaluation values ​​Ev7 to Ev10. Moreover, the evaluation value Ev20 is considerably higher than the evaluation values ​​Ev15 to Ev19 and the evaluation values ​​Ev21 to Ev25.

[0170] Fig. 5(a) is a diagram for explaining the classification of the evaluation value graph GE showing the change of the evaluation value Ev1 to the evaluation value Ev25. As shown in Fig. 5(a), the evaluation value Ev is classified according to the size. For example, the evaluation value is classified into any one of a plurality of categories based on the size.

[0171] Here, the classifier 13 classifies the time series data TD1 to TD25 into any one of four levels based on the evaluation values ​​Ev1 to Ev25. The classifier 13 classifies the time series data TD whose evaluation value Ev does not exceed the threshold value Th1 into level 1. The classifier 13 classifies the time series data TD whose evaluation value Ev is greater than the threshold value Th1 and does not exceed the threshold value Th2 into level 2. The classifier 13 classifies the time series data TD whose evaluation value Ev is greater than the threshold value Th2 and does not exceed Th3 into level 3. The classifier 13 classifies the time series data TD whose evaluation value Ev is greater than the prescribed value Th3 into level 4.

[0172] In the graph of FIG5(a), a boundary line B1 is displayed at the boundary between level 1 and level 2. Similarly, a boundary line B2 is displayed at the boundary between level 2 and level 3, and a boundary line B3 is displayed at the boundary between level 3 and level 4.

[0173] In the graph of FIG5(a), the evaluation values ​​Ev3, Ev8, Ev12, Ev13, Ev15, Ev17, Ev21, and Ev23 correspond to level 1. The evaluation values ​​Ev1, Ev2, Ev4, Ev5, Ev7, Ev9, Ev10, Ev14, Ev16, Ev18, Ev19, Ev22, Ev24, and Ev25 correspond to level 2. The evaluation value Ev11 corresponds to level 3. The evaluation values ​​Ev6 and Ev20 correspond to level 4.

[0174] FIG5(b) is a graph for explaining that the extraction unit 14 extracts the time series data TD6 and the time series data TD20 corresponding to the evaluation value Ev6 and the evaluation value Ev20. As shown in FIG5(b), if the evaluation value Ev obtained according to a specific evaluation benchmark is compared, the values ​​of the evaluation value Ev6 and the evaluation value Ev20 are greater than the other evaluation values. At this time, it can be considered that the time series data TD6 and the time series data TD20 corresponding to the evaluation value Ev6 and the evaluation value Ev20 show an abnormal tendency compared to other time series data TD. Therefore, the extraction unit 14 extracts the time series data TD6 and the time series data TD20 corresponding to the evaluation value Ev6 and the evaluation value Ev20.

[0175] According to the present embodiment, the evaluation values ​​corresponding to the plurality of time series data TD1 to TD100 can be used as a reference to preferably extract the time series data TD6 and the time series data TD20 showing an abnormal tendency from the plurality of time series data TD1 to TD100. Therefore, by extracting the time series data TD6 and the time series data TD20 classified as the object category from the plurality of time series data TD1 to TD100, the abnormal state of the substrate processing apparatus 200 can be grasped and studied without excessive calculation.

[0176] In addition, the description (especially with reference to Figure 3(a) to Figure 5(b) In the description of FIG. 1 , when generating the evaluation value Ev, the time series data TD is compared with the reference data RD, but the present embodiment is not limited thereto. The evaluation value Ev may be generated from the time series data TD without using the reference data RD.

[0177] For example, the acquisition unit 12 may generate the evaluation value Ev by comparing a part of the values ​​of the time series data TD with a specific value stored in the storage unit 20. For example, the storage unit 20 may store a value (peak value equivalent to a peak value) corresponding to the peak value of the graph of the time series data TD, and the acquisition unit 12 may generate the evaluation value Ev by comparing the peak value of the time series data TD with the peak value equivalent to the value stored in the storage unit 20.

[0178] Alternatively, the storage unit 20 may store values ​​corresponding to the rising period of the graph of the time series data (rising period equivalent values), and the acquisition unit 12 may compare the rising period of the time series data TD with the rising period equivalent values ​​stored in the storage unit 20 to generate the evaluation value Ev. Alternatively, the storage unit 20 may store values ​​corresponding to the change values ​​in the peak period of the graph of the time series data (peak change equivalent values), and the acquisition unit 12 may compare the peak change values ​​of the time series data TD with the peak change equivalent values ​​stored in the storage unit 20 to generate the evaluation value Ev.

[0179] In addition, it is preferable that an instruction from an operator or an administrator can be input in the data processing device 100. Furthermore, it is preferable that information can be displayed to the operator or the administrator in the data processing device 100. For example, it is particularly preferable that an evaluation value or time series data (particularly extracted time series data) can be displayed to the operator or the administrator.

[0180] Next, refer to Figure 6 The data processing device 100 according to this embodiment will be described. Figure 6 Schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200 . Figure 6 The data processing device 100 further includes a display unit 30 and an input unit 40, and in addition, has Figure 1 The above-mentioned data processing device 100 has the same structure. Therefore, in order to avoid redundancy, repeated descriptions are omitted.

[0181] The display unit 30 displays an operation screen or various processing results. In addition, the display unit 30 displays a graph showing time series data or an evaluation value graph GE showing changes in evaluation values.

[0182] The display unit 30 includes a display. For example, the display includes a liquid crystal display or an organic electroluminescence (EL) display.

[0183] In addition, it is preferable that the display unit 30 displays an evaluation value graph GE indicating changes in the evaluation value. Furthermore, it is preferable that when the display unit 30 displays the evaluation value graph GE, when the operator or the manager specifies a value (evaluation value) of the evaluation value graph GE via the input unit 40, the display unit 30 displays time series data corresponding to the specified value.

[0184] The input unit 40 includes, for example, various keys for indicating the type of job and the content of the job. The input unit 40 includes a keyboard and a mouse. Alternatively, the input unit 40 may include a touch sensor. In addition, the display unit 30 and the input unit 40 may also be a touch screen in which both are integrated.

[0185] Next, refer to Figure 4(a) to Figure 7(c) The switching display of the evaluation value graph GE and the graph representing the time-series data TD by the display unit 30 will be described.

[0186] 7(a) and 7(b) show the evaluation value graph GE displayed by the display unit 30. As shown in Fig. 7(a), the display unit 30 displays the evaluation value graph GE showing the changes in the values ​​of the evaluation values ​​Ev1 to Ev100 corresponding to the time series data TD1 to TD100. In order to confirm the time series data TD corresponding to the evaluation value Ev via the input unit 40, the operator or the manager specifies the time series data TD corresponding to the evaluation value Ev.

[0187] As shown in FIG7(b), the operator or manager moves the cursor CU to the evaluation value Ev set as the target via the input unit 40 to specify a point representing a specific evaluation value Ev in the evaluation value graph GE. Here, the cursor CU specifies the evaluation value Ev6. At this time, the display unit 30 displays the time series data TD6 corresponding to the specified evaluation value Ev6.

[0188] Fig. 7(c) is a graph showing the time series data TD displayed on the display unit 30. Here, the display unit 30 displays the time series data TD6 corresponding to the evaluation value Ev6.

[0189] In addition, it is preferable that the display unit 30 displays a button BG for returning to the evaluation value graph GE when displaying the time series data TD. When the button BG is selected, the display unit 30 switches to a screen that displays the evaluation value graph GE again. It is preferable that the display unit 30 switches between displaying the evaluation value graph GE that shows the change of the evaluation value Ev and the graph that shows the time series data TD.

[0190] In addition, the storage unit 20 typically stores a plurality of time series data. Preferably, the storage unit 20 stores the plurality of time series data in the order in which the time series data are generated. In this case, preferably, the display unit 30 switches and displays the plurality of time series data in response to a predetermined operation.

[0191] Next, refer to Figure 1 to Figure 8(c) The display of time-series data by the display unit 30 will be described. Figure 8(a) to Figure 8(c) Schematic diagram showing display changes of a plurality of time-series data on the display unit 30 . Figure 8(a) to Figure 8(c) The display screen 32 of the display unit 30 is shown.

[0192] The display screen 32 includes a display area 33 and an operation area 34. The display area 33 displays a time series data. Here, the display area 33 extends in the horizontal direction. The horizontal length of the display area 33 is greater than the vertical length of the display area 33.

[0193] The time series data displayed in the display area 33 can be switched through the operation area 34. Here, the operation area 34 is arranged below the display area 33.

[0194] Figure 8(a) to Figure 8(c) In the embodiment, the operation area 34 is a scroll bar. The operation area 34 includes an arrow 34a and a knob 34b. The arrow 34a extends linearly in the horizontal direction. The knob 34b can move along the arrow 34a in a manner overlapping with a portion of the arrow 34a. The time series data displayed in the display area 33 is switched according to the position of the knob 34b in the arrow 34a.

[0195] The button 34b moves horizontally within the arrow 34a in response to input from the input unit 40. For example, when the button 34b is moved leftward or rightward while the button 34b is selected, the time series data TD displayed in the display area 33 is switched.

[0196] 8(a) , when the knob 34b is located at the left end of the arrow 34a, the time series data TD1 generated first is displayed in the display area 33. The knob 34b moves horizontally within the arrow 34a in response to input from the input unit 40.

[0197] As shown in Fig. 8(b), when the button 34b is moved to the right while the button 34b is selected, the time series data displayed in the display area 33 is switched. For example, when the button 34b is moved to the center of the arrow 34a, the time series data displayed in the display area 33 is switched to the time series data TD50.

[0198] 8C , when the tab 34b is further moved to the right while the tab 34b is selected, the time series data displayed in the display area 33 is switched to the time series data TD100. As described above, the display unit 30 can switch and display a plurality of time series data in response to a predetermined operation.

[0199] Next, refer to Figure 9(a) to Figure 9(c) The switching display of the graph indicating the change of the evaluation value and the time series data by the display unit 30 is described below. FIG9(a) and FIG9(b) show the extracted evaluation value graph EGE displayed by the display unit 30. The extracted evaluation value graph EGE shows the time change of the evaluation value corresponding to the extracted time series data extracted from the plurality of time series data TD. At this time, the time change of the evaluation value of the extracted time series data can be displayed.

[0200] 9( a ), the display unit 30 displays an extracted evaluation value graph EGE showing changes in the evaluation value Ev corresponding to the extracted time series data. To confirm the extracted time series data corresponding to the evaluation value Ev via the input unit 40 , the operator or manager specifies the time series data corresponding to the evaluation value Ev.

[0201] As shown in Fig. 9(b), the operator or manager moves the cursor CU to the evaluation value Ev set as the target to designate a specific evaluation value Ev. Here, the cursor CU designates the evaluation value Ev. At this time, the display unit 30 displays the extracted time series data corresponding to the designated evaluation value Ev.

[0202] 9(c) is a graph showing the extracted time series data displayed by the display unit 30. Preferably, the display unit 30 switches between displaying the graph showing the change of the evaluation value Ev and the extracted evaluation value graph EGE showing the extracted time series data. Thus, the time change of the evaluation value of the extracted time series data can be displayed.

[0203] In addition, in reference Figure 1 to Figure 9(c)In the above description, the extracted time series data is extracted, but it is preferable to process the extracted time series data. For example, it is preferable to perform clustering processing on the extracted time series data.

[0204] Next, refer to Fig.10 The data processing device 100 according to this embodiment will be described. Fig.10 Schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200 . Fig.10 The data processing device 100 further includes a clustering processing unit 15, and in addition, has Figure 6 The above-mentioned data processing device 100 has the same structure. Therefore, in order to avoid redundancy, repeated descriptions are omitted.

[0205] The data processing device 100 further includes a clustering processing unit 15. The clustering processing unit 15 is included in the processing unit 10.

[0206] The clustering processing unit 15 performs clustering processing on the extracted time series data. For example, the clustering processing unit 15 classifies the extracted time series data into any one of a plurality of clusters based on the characteristics of the extracted time series data. In one example, the clustering processing unit 15 classifies the extracted time series data into any one of a plurality of clusters based on the characteristics of a graph shown in the extracted time series data.

[0207] Preferably, the clustering processing unit 15 performs clustering processing on the extracted time series data by so-called unsupervised learning. For example, the clustering processing unit 15 may also perform clustering processing using all values ​​of the extracted time series data. For example, the clustering processing unit 15 may also perform clustering processing by any one of hierarchical clustering, k-means clustering, Gaussian mixture model clustering, self-organizing map clustering, and hidden Markov model clustering.

[0208] For example, the plurality of clusters include a rising delay cluster of the rising delay of the graph, a peak change cluster of the graph with large peak changes, and a low peak cluster of the graph with low peak values. At this time, the clustering processing unit 15 performs clustering processing to classify the extracted time series data into at least one of the rising delay cluster, the peak change cluster, and the low peak cluster. In addition, the plurality of clusters may further include other clusters. For example, the plurality of clusters may also include a falling delay cluster of the falling delay of the graph.

[0209] In addition, the clustering processing unit 15 may also perform clustering processing using all values ​​of the extracted time series data. For example, the clustering processing unit 15 may perform clustering processing using vector analysis on all values ​​of the extracted time series data. Alternatively, the clustering processing unit 15 may perform clustering processing using a portion of the values ​​of the extracted time series data.

[0210] Next, refer to Figure 11(a) to Figure 11(c)To illustrate the clusters obtained by clustering the extracted time series data. FIG11(a) is a graph showing the extracted time series data ETD1 classified as the rising delay cluster, FIG11(b) is a graph showing the extracted time series data ETD2 classified as the peak change cluster, and FIG11(c) is a graph showing the extracted time series data ETD3 classified as the low peak cluster. In addition, Figure 11(a) to Figure 11(c) In each of the figures, reference data RD is shown together with the extracted time series data.

[0211] As shown in Fig. 11(a), when the extracted time series data is compared with the reference data, the rise of the graph of the extracted time series data ETD1 is delayed relative to the graph of the reference data RD. Such extracted time series data is classified as a rise delay cluster.

[0212] As shown in Fig. 11(b), when the extracted time series data is compared with the reference data, the peak of the graph of the extracted time series data ETD2 changes greatly compared with the graph of the reference data RD. Such extracted time series data is classified as a peak change cluster.

[0213] As shown in Fig. 11(c), when the extracted time series data is compared with the reference data, the peak of the graph of the extracted time series data ETD3 is lower than that of the reference data RD. Such extracted time series data is classified as a low peak cluster.

[0214] In addition, refer to Figure 11(a) to Figure 11(c) Although the clusters into which the extracted time series data are classified are described as examples, in the present embodiment, the clusters into which the extracted time series data are classified are not limited to these. The extracted time series data may be classified into other clusters.

[0215] Next, refer to Figures 1 to 12 The data processing method of this embodiment is described. Fig.12 This is a flowchart of the data processing method of this embodiment. Fig.12 The flowchart of FIG. 1 adds the clustering process of step S10, and other than that, the flowchart of FIG. 1 is the same as that of FIG. Figure 2 The above flowchart is the same. Therefore, in order to avoid redundancy, repeated records are omitted.

[0216] like Fig.12 As shown in step S8, the extraction time series data is extracted. The extraction unit 14 extracts the extraction time series data from the plurality of time series data based on the classification result of the evaluation value. Next, the process proceeds to step S10.

[0217] In step S10 , clustering processing is performed on the extracted time series data. The clustering processing unit 15 classifies the extracted time series data into any one of a plurality of clusters.

[0218] The data processing method of the present embodiment is performed as described above. According to the data processing method of the present embodiment, it is possible to classify the extracted time series data showing abnormal evaluation values ​​from a plurality of time series data into each cluster.

[0219] Generally speaking, if all time series data are clustered, since the time series data also includes normal time series data, clusters that are not abnormal will be formed unnecessarily, and the abnormal state of the substrate processing device may not be fully understood. In contrast, according to the data processing method of this embodiment, the extracted time series data showing abnormal evaluation values ​​are clustered from a plurality of time series data, so that the abnormal state of the substrate processing device 200 can be more accurately classified into clusters. Furthermore, according to the data processing method of this embodiment, clustering is not performed on all time series data, so unnecessary calculations can be avoided.

[0220] Preferably, the display unit 30 displays an evaluation value graph showing changes in the corresponding evaluation value for each cluster. At this time, preferably, when the operator or manager specifies an evaluation value in the graph, the display unit 30 displays time series data corresponding to the specified evaluation value.

[0221] For example, in reference Figure 9(a) to Figure 9(c) In the above description, the display unit 30 displays the extracted evaluation value graph EGE showing the time change of the evaluation value for the extracted time series data extracted by the extraction unit 14, but the present embodiment is not limited thereto. The display unit 30 may also display a graph showing the time change of the evaluation value for the time series data classified into clusters by the clustering processing unit 15. In this case, the time change of the evaluation value within a specific cluster can be displayed.

[0222] In addition, it is preferred that, as referenced Figures 10 to 12 As described above, the extracted time series data is clustered. It goes without saying that the extracted time series data reflects the abnormal state of the substrate processing device 200, and the result of the clustering process represents the state of the substrate processing device 200 corresponding to the type of the abnormal state of the substrate processing device 200. Therefore, it is preferred that the result of the clustering process is used to learn the state of the substrate processing device 200.

[0223] Next, refer to Figures 1 to 13 The data processing method of this embodiment is described. Fig.13 This is a flowchart of the data processing method of this embodiment. Fig.13The flowchart of FIG. 8 determines whether the extracted time series data has been extracted in step S8A, and adds steps S2B to S6B, step S9, and steps S22 to S24. In addition, the flowchart of FIG. 8 is the same as that of FIG. Fig.12 The above flowchart is the same. Therefore, in order to avoid redundancy, repeated records are omitted. Fig.13 The data processing method shown can be preferably used to create a learning database in the substrate processing apparatus 200 .

[0224] In step S2, time series data is acquired. In addition, here, the time series data is used to learn the state of the substrate processing apparatus 200. Therefore, in this specification, the time series data may be referred to as learning time series data.

[0225] Here, the acquisition unit 11 acquires a learning time series data generated in the substrate processing apparatus 200. Next, the process proceeds to step S4. In addition, step S4 and step S6 are similar to the reference Fig.12 The above flowchart is the same. After step 6, the process proceeds to step S8A.

[0226] In step S8A, it is determined whether to extract the learning time series data as the extracted time series data. When the classifier 13 classifies the learning time series data into the object class based on the evaluation value Ev, the extractor 14 extracts the learning time series data as the extracted time series data. On the other hand, when the classifier 13 classifies the learning time series data into the non-object class based on the evaluation value Ev, the extractor 14 does not extract the learning time series data as the extracted time series data.

[0227] For example, when the classifier 13 classifies the learning time series data into level 4 based on the evaluation value Ev of the learning time series data, the extractor 14 extracts the learning time series data as the extracted time series data. Furthermore, when the classifier 13 classifies the learning time series data into any one of levels 1 to 3, the extractor 14 does not extract the learning time series data as the extracted time series data.

[0228] If the extracted time series data is not extracted (No in step S8A), the process returns to step S2. If the extracted time series data is extracted (Yes in step S8A), the process proceeds to step S10. In addition, if clustering processing is not performed in advance, the process may proceed to step S10 after extracting the amount of data that can perform unsupervised learning.

[0229] In step S10, the extracted time series data is subjected to clustering processing. Thus, the learning time series data is classified into any one of a plurality of clusters. Next, the process proceeds to step S22.

[0230] In step S22, the control of the substrate processing apparatus 200 is changed. Specifically, the control of the substrate processing apparatus 200 or the processing unit 210 for which the learning time series data is generated is changed. The control change of the substrate processing apparatus 200 may also be performed according to a program for controlling the drive of the substrate processing apparatus 200. Alternatively, the control change of the substrate processing apparatus 200 may also be manually input into the substrate processing apparatus 200.

[0231] Typically, the control of the substrate processing apparatus 200 is performed in a state where the substrate processing is temporarily stopped in the substrate processing apparatus 200. For example, the time series data TD indicates a change in the supply amount of the processing liquid used to process the substrate. If the supply amount of the processing liquid is in an abnormal state, the substrate processing is temporarily stopped and the valve is adjusted so that the supply amount returns to normal. In addition, the control change of the substrate processing apparatus 200 can also be performed while the substrate processing of the substrate processing apparatus 200 is continued. Next, the process proceeds to step S2B.

[0232] In step S2B, time series data is acquired from the substrate processing apparatus 200 that has undergone a control change. In this specification, time series data acquired from the substrate processing apparatus 200 that has undergone a control change is sometimes referred to as change time series data. In addition, here, the acquisition unit 11 acquires one time series data generated in the substrate processing apparatus 200 as the change time series data. Next, the process proceeds to step S4B.

[0233] In step S4B, an evaluation value (score) is acquired. The acquisition unit 12 acquires an evaluation value for the change time series data for the substrate processing apparatus 200 that has undergone a control change. For example, the acquisition unit 12 generates an evaluation value for the change time series data based on a specific evaluation criterion in the same manner as the learning time series data. Next, the process proceeds to step S6B.

[0234] In step S6B, the change time series data is classified into any one of a plurality of categories using the evaluation value. Next, the process proceeds to step S9.

[0235] In step S9, it is determined whether the change time series data is extracted as the extracted time series data. When the classifier 13 classifies the change time series data as the target class, the extractor 14 extracts the change time series data as the extracted time series data. On the other hand, when the classifier 13 classifies the change time series data as the non-target class, the extractor 14 does not extract the target time series data as the extracted time series data.

[0236] For example, when the classifier 13 classifies the change time series data into level 4 based on the evaluation value Ev of the change time series data, the extractor 14 extracts the change time series data as the extracted time series data. Furthermore, when the classifier 13 classifies the change time series data into any one of levels 1 to 3, the extractor 14 does not extract the change time series data as the extracted time series data.

[0237] If the extraction time series data is extracted (Yes in step S9), the process returns to step S22. At this time, the substrate processing apparatus 200 is controlled again to further change the control of the substrate processing apparatus 200. On the other hand, if the extraction time series data is not extracted (No in step S9), the process proceeds to step S24.

[0238] In step S24, information related to the content of the change in the control of the substrate processing apparatus 200 performed in step S22 is stored as cause countermeasure information in the storage unit 20. The cause countermeasure information is information related to the content of the change in the control of the substrate processing apparatus 200 performed in step S22. For example, the cause countermeasure information may be information indicating the content of the change in the control of the substrate processing apparatus 200 itself. Alternatively, the cause countermeasure information may be information considered to be the reason why the learning time series data presents an abnormal state.

[0239] For example, an operator or manager of the data processing apparatus 100 and / or the substrate processing apparatus 200 inputs the cause countermeasure information via the input unit 40, and the storage unit 20 stores the cause countermeasure information. At this time, it is preferable that the cause countermeasure information is stored together with the learning time series data for each cluster subjected to the clustering process in step S10 in the storage unit 20. In addition, the clusters, the learning time series data, and the cause countermeasure information stored in the storage unit 20 can be preferably used as a learning database.

[0240] The data processing method of this embodiment is performed as described above. According to the data processing method of this embodiment, the acquisition unit 11, the acquisition unit 12, the classification unit 13, and the extraction unit 14 of the data processing device 100 can be used to determine the abnormal state of the substrate processing device 200, and after the control of the substrate processing device 200 is changed, it can be determined whether the abnormal state of the substrate processing device 200 has been eliminated.

[0241] Furthermore, according to the data processing method of this embodiment, after the time series data of a specific state is extracted and clustered in the substrate processing apparatus 200, the cause countermeasure information is stored. Therefore, the cause countermeasure information that is considered to be effective in restoring the time series data of the substrate processing apparatus 200 from an abnormal state to an original state can be stored together.

[0242] In addition, it is preferable that the storage unit 20 stores the clusters, the extracted time series data, and the cause and countermeasure information as a learning database.

[0243] Next, refer to Fig.14 The learning database of the storage unit 20 will be described. Fig.14 It is a schematic table for explaining the learning database of the storage unit 20.

[0244] like Fig.14 As shown, the learning database includes clusters classified by the clustering processing unit 15, time series data included in the clusters, and cause countermeasure information corresponding to the clusters. Fig.14 In the table shown, the clusters include a rise delay cluster, a peak variation cluster, and a low peak cluster.

[0245] Specifically, the rise delay cluster includes time series data TD6 and time series data TD40 as time series data, and includes process liquid concentration drop / valve re-tightening as cause countermeasure information. The peak change cluster includes time series data TD20 and time series data TD41 as time series data, and includes substrate holding force drop / overcurrent suppression as cause countermeasure information. The low peak cluster includes time series data TD80 and time series data TD95 as time series data, and includes environmental unevenness / sensor restart as cause countermeasure information.

[0246] In addition, in reference Fig.14 In the description of, in order to avoid the invention becoming overly complicated, a set of cause countermeasure information is represented corresponding to one cluster, but one cluster may correspond to more than two cause countermeasure information. In this case, it is preferable to use so-called machine learning to appropriately learn the generation of clusters and the corresponding relationship between the cause countermeasure information.

[0247] In this embodiment, the learning database uses the time series data subsequently generated by the substrate processing apparatus 200 and / or the processing unit 210 to change the control of the substrate processing apparatus 200 and / or the processing unit 210. The data processing apparatus 100 may also change the control of the substrate processing apparatus 200 according to the time series data generated in the substrate processing apparatus 200 based on the learning content stored in the learning database of the storage unit.

[0248] Next, refer to Figures 1 to 15 The data processing device 100 according to this embodiment will be described. Fig.15 1 is a schematic diagram of a data processing device 100 of the present embodiment. The data processing device 100 includes a processing unit 110, a storage unit 120, a display unit 130, and an input unit 140. Here, the storage unit 120 stores a learning database. The display unit 130 and the input unit 140 correspond to the reference Figure 6The display unit 30 and the input unit 40 mentioned above.

[0249] The processing unit 110 includes a data acquisition unit 111, an evaluation value acquisition unit 112, a classification unit 113, a matching unit 114, and a reading unit 115. The data acquisition unit 111, the evaluation value acquisition unit 112, and the classification unit 113 correspond to the reference Figure 6 The above-mentioned data acquisition unit 11, evaluation value acquisition unit 12 and classification unit 13.

[0250] The matching unit 114 matches the time series data classified as the object class by the classification unit 113 with the extracted time series data of the learning database stored in the storage unit 120. In addition, in this specification, the time series data classified as the object class by the classification unit 113 is sometimes referred to as the object time series data. In addition, the time series data similar to the object time series data in the time series data of the learning database is sometimes referred to as similar time series data.

[0251] The matching unit 114 determines the time series data similar to the object time series data in the time series data of the learning database as the similar time series data. Typically, the time series data closest to the object time series data in the time series data of the learning database is the similar time series data. However, there is sometimes no time series data in the learning database that is similar time series data.

[0252] Typically, the matching unit 114 matches the object time series data with the time series data stored in the learning database based on a specific evaluation criterion. For example, the matching unit 114 matches the object time series data with the time series data in the learning database for the entirety of the time series data. In one example, the matching unit 114 may also process the object time series data with the time series data in the learning database by vector analysis. Alternatively, the matching unit 114 may also match the object time series data with the time series data stored in the learning database for a specific portion of the time series data. The matching unit 114 determines similar time series data that is close to the object time series data in the time series data stored in the learning database of the storage unit 120. In addition, the matching of the object time series data with the similar time series data may be performed based on the same evaluation criterion as that of the evaluation value acquisition unit 112, or may be performed based on an evaluation criterion different from that of the evaluation value acquisition unit 112.

[0253] Then, the reading unit 115 identifies the cluster corresponding to the similar time series data in the learning database of the storage unit 120, and reads out the cause and countermeasure information corresponding to the identified cluster. Typically, the display unit 130 displays the cause and countermeasure information.

[0254] According to the data processing device 100 of this embodiment, similar time series data close to the target time series data and corresponding cause countermeasure information are read from the learning database. Therefore, the data processing device 100 can obtain appropriate cause countermeasure information for the target time series data classified into a specific category.

[0255] Next, refer to Figures 1 to 16 The data processing method of this embodiment is described. Fig.16 A flowchart showing a data processing method according to the present embodiment. Fig.16 Steps S102 to S106 correspond to Fig.12 Steps S2 to S6.

[0256] like Fig.16 As shown, in step S102, time series data is acquired. The time series data is data generated in the substrate processing apparatus 200. Next, the process proceeds to step S104.

[0257] In step S104, the evaluation value (score) of the time series data is acquired. The acquisition unit 12 acquires the evaluation value of the time series data. For example, the acquisition unit 12 acquires the evaluation value of the time series data according to the same evaluation criteria as when the learning database is created. Next, the process proceeds to step S106.

[0258] In step S106, the time series data is classified into any one of a plurality of categories using the evaluation value of the time series data. For example, the classification unit 13 classifies the time series data into any one of a plurality of categories using the evaluation value in the same manner as when creating a learning database. Next, the process proceeds to step S106a.

[0259] In step S106a, it is determined whether the time series data has been classified into the target category. For example, the target category is level 4.

[0260] If the time series data has not been classified into the target category (No in step S106a), the process returns to step S102. If the time series data has been classified into the target category (Yes in step S106a), the process proceeds to step S108.

[0261] In step S108, the target time series data is matched with the time series data stored in the learning database of the storage unit 120. The matching unit 114 matches the target time series data with the time series data stored in the learning database of the storage unit 120. The matching unit 114 determines similar time series data in the extracted time series data stored in the learning database of the storage unit 120. Next, the process proceeds to step S110.

[0262] In step S110, the cause countermeasure information corresponding to the similar time series data is read out from the learning database. The reading unit 115 determines the cluster corresponding to the similar time series data in the learning database of the storage unit 120, and reads out the cause countermeasure information corresponding to the determined cluster. Typically, the display unit 130 then displays the cause countermeasure information. In addition, as needed, the operator or manager of the data processing apparatus 100 and / or the substrate processing apparatus 200 can also change the control of the substrate processing apparatus 200 based on the displayed cause countermeasure information.

[0263] The data processing method of this embodiment is performed as described above. According to the data processing method of this embodiment, matching is performed on target time series data that is identified as abnormal via evaluation values. Therefore, the amount of calculation related to the matching process can be reduced.

[0264] Furthermore, according to the present embodiment, the cause countermeasure information corresponding to the cluster of the object time series data can be obtained from the learning database. Therefore, even if the time series data is abnormal, the cause countermeasure information corresponding to the time series data can be obtained from the learning database. Therefore, the abnormal state of the substrate processing device 200 can be effectively grasped.

[0265] Typically, an experienced operator of a substrate processing apparatus can infer the condition of the substrate processing apparatus based on the time series data obtained from the substrate processing apparatus, but an unskilled operator cannot infer the condition of the substrate processing apparatus based on the time series data obtained from the substrate processing apparatus. However, according to the data processing method of this embodiment, if the target time series data of the substrate processing apparatus 200 is abnormal, the cause countermeasure information of the substrate processing apparatus 200 can be obtained using the learning database prepared in the past. Therefore, even if the operator of the substrate processing apparatus 200 is unskilled, the bad condition of the substrate processing apparatus 200 can be eliminated.

[0266] In addition, in reference Fig.16 In the description of FIG. 1 , in step S110, the cause countermeasure information is read, and then the operator or the manager changes the control of the substrate processing apparatus 200 based on the displayed cause countermeasure information, but the present embodiment is not limited thereto. The control of the substrate processing apparatus 200 may also be changed based on the read cause countermeasure information without the operator or the manager.

[0267] Moreover, in reference Fig.16 In the above description, in the matching of step S108, similar time series data is determined from the learning database, but if the time series data obtained in step S102 is extremely abnormal, it may be considered not to determine similar time series data from the learning database. In this case, the time series data obtained in step S102 may also be stored in the storage unit 120 separately.

[0268] Next, refer to Figures 1 to 17 The data processing method of this embodiment is described. Fig.17 A flowchart showing a data processing method according to the present embodiment. Fig.17 The flowchart of the data processing method of the invention adds step S108a, step S112 to step S114, and other steps are the same as Fig.16 The flowchart is the same, so in order to avoid redundancy, repeated records are omitted.

[0269] As described above, in step S108, the target time series data is matched with the time series data stored in the learning database of the storage unit 120. The matching unit 114 matches the target time series data with the time series data stored in the learning database of the storage unit 120. For example, the matching is performed by obtaining the matching ratio between the target time series data and the time series data of the learning database. Next, the process proceeds to step 108a.

[0270] In step S108a, it is determined whether similar time series data similar to the object time series data exists in the learning database. For example, the matching unit 114 may also determine whether similar time series data similar to the object time series data exists in the learning database based on whether the matching rate exceeds a threshold. For example, if the matching rate is higher than the threshold, the matching unit 114 determines that similar time series data exists. On the other hand, if the matching rate is lower than the threshold, the matching unit 114 determines that there is no similar time series data.

[0271] If it is determined that similar time series data exists (Yes in step S108a), the process proceeds to step S110. If it is determined that similar time series data does not exist (No in step S108a), the process proceeds to step S114.

[0272] In step S110, cause countermeasure information corresponding to the similar time series data is read out from the learning database. The reading unit 115 reads cause countermeasure information corresponding to the similar time series data from the learning database of the storage unit 120. The process proceeds to step S112.

[0273] In step S112, control of the substrate processing apparatus 200 is changed based on the cause countermeasure information. For example, the processing unit 110 notifies the operator or the manager of the cause countermeasure information. In one example, the display unit 130 displays the cause countermeasure information.

[0274] In step S114, the target time series data is stored in the learning database of the storage unit 120. At this time, preferably, the storage unit 120 stores information indicating a new cluster corresponding to the target time series data together with the target time series data. In addition, at this time, the display unit 130 may also display that the matching rate is below the threshold.

[0275] According to the data processing method of this embodiment, the cause countermeasure information corresponding to the cluster of the target time series data is read out from the learning database and displayed, so that the operator or manager of the data processing apparatus 100 can obtain the cause countermeasure information about the substrate processing apparatus 200.

[0276] Next, refer to Figure 18(a) to Figure 20(b) The time series data TD will be described in detail. Typically, the time series data TD can be decomposed into a rising portion, a stable period portion, and a falling portion according to the shape of the graph. Furthermore, the time series data TD typically represents a temporal change in a physical quantity in response to a predetermined control signal.

[0277] First, refer to Figure 1 The time series data TD is described with reference to Fig. 18(a) and Fig. 18(b). Fig. 18(a) is a graph showing the time variation of the control signal CS in the substrate processing apparatus 200. Fig. 18(b) is a graph showing the time series data TD, which represents the time variation of the physical quantity controlled according to the control signal CS.

[0278] As shown in FIG18( a ), the control signal CS is at a low level in the initial state (time t0 ), the control signal CS changes from a low level to a high level at time t1 , and changes from a high level to a low level at time t2 .

[0279] As shown in FIG. 18( b ), the time series data TD changes between an initial level L0 and a target level L1 (where L0 < L1 ) according to the control signal CS.

[0280] The time series data TD has an initial level L0 in the initial state (time t0). When the control signal CS changes from a low level to a high level at time t1, the time series data TD starts to rise from the initial level L0 toward the target level L1. Here, the time series data TD falls after rising above the target level L1. The time series data TD repeatedly rises and falls near the target level L1, and finally stabilizes near the target level L1.

[0281] Subsequently, when the control signal CS changes from a high level to a low level at time t2, the time series data TD starts to decrease from near the target level L1 toward the initial level L0. The time series data TD increases after decreasing to or near the initial level L0. The time series data TD repeatedly increases and decreases near the initial level L0, and finally stabilizes at the initial level L0.

[0282] For example, in the time series data TD, a first range R1 including the target level L1 and a second range R2 including the initial level L0 are set. For example, a range of 90% to 110% of the target level L1 is set as the first range R1, and a range of -10% to 10% of the target level L1 is set as the second range R2. The upper limit and lower limit of the first range R1 and the upper limit and lower limit of the second range R2 are arbitrarily determined by the user.

[0283] The evaluation value acquisition unit 12 may also acquire the evaluation value from the time series data TD as follows. First, the evaluation value acquisition unit 12 obtains the period from when the control signal CS changes from a low level to a high level until the time series data TD converges within the first range R1 as "period A (rising period)", obtains the period from when the control signal CS changes from a high level to a low level until the time series data TD converges within the second range R2 as "period C (falling period)", and obtains the period between the rising period and the falling period as "period B (stable period)". In addition, the so-called "time series data converges within a certain range" means that the time series data will not take a value outside the range after this time point.

[0284] The evaluation value acquisition unit 12 may acquire the evaluation value from any one of "period A", "period B" and "period C". For example, the evaluation value acquisition unit 12 may acquire the evaluation value from "period A". Alternatively, the evaluation value acquisition unit 12 may acquire the evaluation value from "period B" or "period C".

[0285] For example, the evaluation value acquisition unit 12 obtains the score of the rising period, the score of the stable period, and the score of the falling period by a predetermined method. For example, the evaluation value acquisition unit 12 may obtain the length of the rising period as the score of the rising period, and obtain the length of the falling period as the score of the falling period. In this way, the evaluation value acquisition unit 12 may use the lengths of "period A", "period B", and "period C" themselves as evaluation values. Alternatively, the evaluation value acquisition unit 12 may use the average value or composite value of the lengths of "period A", "period B", and "period C" as the evaluation value.

[0286] Alternatively, the evaluation value acquisition unit 12 may acquire any one value (score) from “period A”, “period B”, and “period C”, respectively, and determine, compare, and / or synthesize these scores to acquire the evaluation value.

[0287] The evaluation value acquisition unit 12 may obtain statistical values ​​of the time series data TD in “Period A”, “Period B”, and “Period C” as evaluation values ​​or scores of “Period A”, “Period B”, and “Period C”.

[0288] For example, the evaluation value acquisition unit 12 may use a plurality of time series data to obtain a burst value in a stable period as a score in a stable period. The evaluation value acquisition unit 12 obtains an average value, a median value, or a variance of the time series data TD in a stable period as a score in a stable period.

[0289] Fig.19 A graph showing values ​​of multiple time series data. Fig.19 The n time series data TD1, TD2, ..., TDn shown in the figure each contain m values. Here, when i is an integer greater than 1 and less than n, and j is an integer greater than 1 and less than m, the jth data contained in the time series data TDi is referred to as x. ij At this time, the evaluation value acquisition unit 12 obtains the score Sp of the time series data TDp in accordance with the following equations (1) to (4) in order.

[0290] [Number 1]

[0291]

[0292] In addition, in formula (1), the value μ pj represents the average value of the jth data included in the (n-1) time series data other than the target time series data TDp. In formula (2), the value μ p represents the average value of all data contained in the (n-1) time series data other than the time series data TDp. In formula (3), the value σ p 2 Indicates the variance of (n-1) time series data other than the time series data TDp.

[0293] Furthermore, the evaluation value acquisition unit 12 may also obtain the overshoot amount of the time series data TD as the score of the rising period. When the target level of the time series data TD is set to L1 and the maximum value of the time series data TD is set to M, the evaluation value acquisition unit 12 may also obtain the overshoot amount V of the time series data according to formula (5).

[0294] V=(M-L1) / L1×100…(5)

[0295] Alternatively, the evaluation value acquisition unit 12 may also calculate the overshoot amount V of the time series data according to equation (6).

[0296] V=M-L1…(6)

[0297] In addition, the methods for obtaining the evaluation values ​​are merely examples and are not limited to these methods. Figure 4(a) to Figure 4(c) As described above, the evaluation value is acquired by accumulating the square of the difference between the value of the time series data TD and the value of the reference data RD.

[0298] In addition, the classification unit 13 ( Figure 1 etc.) can also be classified into any one of a plurality of categories by processing the evaluation values ​​Ev of each of a plurality of time series data TD.

[0299] Next, an example of the classification process performed by the classification unit 13 is described with reference to FIG. 20(a) and FIG. 20(b). FIG. 20(a) is a graph showing the distribution of evaluation values. The distribution of evaluation values ​​represents the frequency (number of times) of the evaluation values ​​Ev for each of the plurality of time series data TD. In the graph of FIG. 20(a), the horizontal axis represents the magnitude of the evaluation value Ev, and the vertical axis represents the frequency (number of times or number of occurrences) of the evaluation value Ev.

[0300] In Fig. 20(a), μ represents the average value of a plurality of evaluation values ​​Ev, and σ represents the standard deviation. As shown in Fig. 20(a), the evaluation value distribution has a peak near the average value μ. The evaluation value distribution has a shape close to a normal distribution.

[0301] The classifier 13 normalizes the evaluation value distribution. Specifically, the classifier 13 generates a normalized distribution by normalizing the evaluation value distribution according to the equation (7).

[0302] [Number 2]

[0303]

[0304] Here, Sold represents the evaluation value distribution, and Snew represents the standardized distribution.

[0305] 20( b ) is a graph showing a normalized distribution. The normalized distribution is generated by performing a normalization process so that the average μ of the evaluation value distribution becomes 0 and the standard deviation becomes 1.

[0306] For example, the standardized distribution is divided into L1 to L4. In the standardized distribution, if it is less than 1, it is set to level 1 (L1), if it is greater than 1 and less than 2, it is set to level 2 (L2), if it is greater than 2 and less than 3, it is set to level 3 (L3), and if it is greater than 3, it is set to level 4 (L4). Here, the level indicates the degree of abnormality.

[0307] Here, the classification is performed using the standard deviation, so the operator or manager does not need to input threshold values ​​for abnormality determination individually.

[0308] In addition, refer to Figure 18(a) to Figure 20(b) The description is mainly about the evaluation value acquisition unit 12 and the classification unit 13, but it can also be applied to the evaluation value acquisition unit 112 and the classification unit 113. Moreover, it can also be applied to the clustering processing unit 15 and / or the matching unit 114.

[0309] in addition, Fig.10 In the data processing device 100, the processing unit 10 stores a learning database indicating the state of the substrate processing device 200 in the storage unit 20, and on the other hand, Fig.15 In the data processing device 100, the processing unit 110 uses the learning database stored in the storage unit 120, but the processing unit of the data processing device 100 can also store the learning database representing the state of the substrate processing device 200 in the storage unit and use the learning database stored in the same storage unit.

[0310] Next, refer to Fig.21 The data processing apparatus 100 and the substrate processing apparatus 200 according to the present embodiment will be described. Fig.21 Schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200 .

[0311] like Fig.21 As shown, the data processing device 100 includes a processing unit 10 and a storage unit 20. The processing unit 10 includes a data acquisition unit 111, an evaluation value acquisition unit 112, a classification unit 113, a matching unit 114, and a reading unit 115 in addition to the data acquisition unit 11, the evaluation value acquisition unit 12, the classification unit 13, the extraction unit 14, and the clustering processing unit 15. The storage unit 20 stores a learning database.

[0312] The processing unit 10 can store the learning database in the storage unit 20 through the data acquisition unit 11, the evaluation value acquisition unit 12, the classification unit 13, the extraction unit 14, and the clustering processing unit 15. In addition, the processing unit 10 can effectively use the learning database stored in the storage unit 20 through the data acquisition unit 111, the evaluation value acquisition unit 112, the classification unit 113, the matching unit 114, and the reading unit 115.

[0313] In addition, generally speaking, the substrate processing apparatus 200 is roughly classified into a single-wafer type or a batch type. The substrate processing apparatus 200 may be either a single-wafer type or a batch type.

[0314] Next, refer to Fig. 22 The data processing apparatus 100 and the substrate processing apparatus 200 according to the present embodiment will be described. Fig. 22 Schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200 .

[0315] The data processing apparatus 100 processes the time series data TD generated in the substrate processing apparatus 200. Here, the substrate processing apparatus 200 is a single-chip type.

[0316] The substrate processing apparatus 200 processes the substrate W. The substrate processing apparatus 200 processes the substrate W by performing at least one of etching, surface processing, imparting characteristics, forming a processing film, removing at least a portion of a film, and cleaning the substrate W.

[0317] The substrate W includes, for example, a semiconductor wafer, a substrate for a liquid crystal display device, a substrate for a plasma display, a substrate for a field emission display (FED), a substrate for an optical disk, a substrate for a magnetic disk, a substrate for an optical magneto-optical disk, a substrate for a photomask, a ceramic substrate, and a substrate for a solar cell. For example, the substrate W is roughly disk-shaped. The substrate processing apparatus 200 processes the substrates W piece by piece.

[0318] The substrate processing apparatus 200 includes a chamber 220 , a substrate holding portion 230 , and a liquid supply portion 240 as a processing unit 210 . The chamber 220 accommodates a substrate W. The substrate holding portion 230 holds the substrate W.

[0319] The chamber 220 is substantially box-shaped and has an internal space. The chamber 220 accommodates the substrate W. Typically, the chamber 220 is an air environment. However, in the chamber 220, an air flow may be formed by downflow or the like.

[0320] Here, the substrate processing apparatus 200 is a single-wafer type that processes substrates W one by one, and the substrates W are accommodated one by one in the chamber 220. The substrates W are accommodated in the chamber 220 and processed in the chamber 220. The chamber 220 accommodates at least a portion of each of the substrate holding unit 230 and the liquid supply unit 240.

[0321] The substrate holding part 230 holds the substrate W. For example, the substrate holding part 230 clamps the end of the substrate W. The substrate holding part 230 holds the substrate W horizontally in such a manner that the upper surface (front surface) Wa of the substrate W faces upward and the back surface (lower surface) Wb of the substrate W faces vertically downward. Moreover, the substrate holding part 230 rotates the substrate W while holding the substrate W. For example, the substrate W rotates counterclockwise when viewed from vertically above.

[0322] For example, the substrate holding part 230 includes a spin base 231, a chuck member 232, a shaft 233, an electric motor 234, and a base 235. For example, the spin base 231 is in the shape of a plate extending along the XY plane. Here, the spin base 231 is in the shape of a disk (thin circular plate shape). The spin base 231 faces the substrate W.

[0323] The chuck member 232 is provided on the rotating base 231 . The chuck member 232 fixes (chucks) the substrate W. Typically, a plurality of chuck members 232 are provided on the rotating base 231 .

[0324] The shaft 233 may be a hollow shaft. The shaft 233 extends in the vertical direction along the rotation axis Ax. The rotating base 231 is coupled to the upper end of the shaft 233. The substrate W is placed on the rotating base 231.

[0325] The shaft 233 extends downward from the center of the rotating base 231. The electric motor 234 applies a rotational force to the shaft 233. The shaft 233 rotates relative to the base 235. The base 235 rotatably supports the shaft 233. The electric motor 234 rotates the shaft 233 in a rotation direction, thereby rotating the substrate W and the rotating base 231 around the rotation axis Ax. The electric motor 234 is an example of a rotating member.

[0326] The liquid supply unit 240 supplies liquid to the substrate W. Typically, the liquid supply unit 240 supplies liquid to the upper surface Wa of the substrate W. For example, the liquid includes a rinse solution or a chemical solution.

[0327] The eluent may also include any one of deionized water (DIW), carbonated water, electrolytic ionized water, ozone water, ammonia water, hydrochloric acid water with a diluted concentration (eg, about 10 ppm to 100 ppm), or reduced water (hydrogen water).

[0328] The chemical solution contains hydrofluoric acid. For example, the hydrofluoric acid may be heated to 40° C. or higher and 70° C. or higher and 60° C. or lower. However, the hydrofluoric acid may not be heated. Furthermore, the chemical solution may contain water or phosphoric acid.

[0329] Furthermore, the chemical solution may further include hydrogen peroxide solution. Furthermore, the chemical solution may include SC1 (a mixture of ammonia and hydrogen peroxide solution), SC2 (a mixture of hydrochloric acid and hydrogen peroxide solution), or aqua regia (a mixture of concentrated hydrochloric acid and concentrated nitric acid).

[0330] The substrate processing apparatus 200 further includes a cup 250. The cup 250 recovers liquid scattered from the substrate W. The cup 250 can rise and fall. For example, the cup 250 rises vertically upward to the side of the substrate W during the period when the liquid supply unit 240 supplies liquid to the substrate W. At this time, the cup 250 recovers liquid scattered from the substrate W due to the rotation of the substrate W. Moreover, when the period when the liquid supply unit 240 supplies liquid to the substrate W ends, the cup 250 descends vertically downward from the side of the substrate W.

[0331] The control device 201 includes a control unit 201a and a storage unit 201b. The control unit 201a controls the substrate holding unit 230, the liquid supply unit 240 and / or the cup unit 250. In one example, the control unit 201a controls the electric motor 234, the valve 240b and / or the cup unit 250.

[0332] The substrate processing apparatus 200 of this embodiment can be preferably used for manufacturing a semiconductor device having a semiconductor. The substrate processing apparatus 200 can be preferably used for cleaning and / or processing (eg etching, characteristic change, etc.) of a semiconductor device during the manufacturing of the semiconductor device.

[0333] As described above, the substrate processing apparatus 200 may also be a single-chip type. In this case, it is preferred that the substrate processing apparatus 200 includes a plurality of processing units 210 .

[0334] Next, refer to Fig.23 The data processing apparatus 100 and the substrate processing apparatus 200 according to the present embodiment will be described. Fig.23 2 is a schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200. Here, the substrate processing apparatus 200 includes a plurality of processing units 210. Each of the processing units 210 processes a substrate W. The plurality of processing units 210 are arranged in a predetermined arrangement.

[0335] like Fig.23 As shown, the substrate processing apparatus 200 includes a plurality of processing units 210, a fluid chamber LC, a fluid box LB, a plurality of load ports LP, an indexer robot IR, a center robot CR, and a control device 201. The control device 201 controls the load port LP, the indexer robot IR, and the center robot CR. The control device 201 includes a control unit 201a and a storage unit 201b.

[0336] The loading ports LP each stack and accommodate multiple substrates W. The transfer robot IR transfers the substrates W between the loading port LP and the central robot CR. The central robot CR transfers the substrates W between the transfer robot IR and the processing unit 210. The processing units 210 each spray liquid onto the substrates W to process the substrates W. For example, the liquid includes a processing liquid, a rinse liquid, and / or a chemical solution. The fluid chamber LC accommodates the liquid. In addition, the fluid chamber LC may also accommodate a gas.

[0337] Specifically, the plurality of processing units 210 are formed of a plurality of towers TW ( Figure 1 Each tower TW includes a plurality of processing units 210 ( Figure 1 There are three processing units 210 in the figure. The fluid boxes LB correspond to the multiple towers TW. The liquid in the fluid chamber LC is supplied to all the processing units 210 included in the tower TW corresponding to the fluid box LB through any one of the fluid boxes LB. In addition, the gas in the fluid chamber LC is supplied to all the processing units 210 included in the tower TW corresponding to the fluid box LB through any one of the fluid boxes LB.

[0338] The substrate processing apparatus 200 further includes a control device 201 . The control device 201 controls various operations of the substrate processing apparatus 200 .

[0339] The control device 201 includes a control unit 201a and a storage unit 201b. The control unit 201a includes a processor. The control unit 201a includes, for example, a central processing unit (CPU). Alternatively, the control unit 201a may include a general-purpose processor.

[0340] The storage unit 201b stores data and computer programs. The data includes recipe data. The recipe data includes information indicating a plurality of recipes. Each of the plurality of recipes specifies the processing content and processing flow of the substrate W.

[0341] The storage unit 201b includes a main storage device and an auxiliary storage device. The main storage device is, for example, a semiconductor memory. The auxiliary storage device is, for example, a semiconductor memory and / or a hard disk drive. The storage unit 201b may also include a removable medium. The control unit 201a executes the computer program stored in the storage unit 201b to perform substrate processing operations.

[0342] In addition, in reference Fig. 22 and Fig.23 In the above description, the substrate processing apparatus 200 and / or the processing unit 210 are of a single-chip type, but the present embodiment is not limited thereto. The substrate processing apparatus 200 and / or the processing unit 210 may also be of a batch type.

[0343] Reference Fig.24 The data processing apparatus 100 and the substrate processing apparatus 200 according to the present embodiment will be described. Fig.24 1 is a schematic diagram of a data processing apparatus 100 and a substrate processing apparatus 200 according to the present embodiment. Here, the substrate processing apparatus 200 is a batch type, and can process a plurality of substrates W at once.

[0344] The substrate processing apparatus 200 includes a processing tank 260, a substrate holding portion 270, and a control device 201. The processing tank 260 stores a processing liquid L for processing a substrate W.

[0345] The substrate holding portion 270 holds the substrate W. The normal direction of the main surface of the substrate W held by the substrate holding portion 270 is parallel to the Y direction. The substrate holding portion 270 moves the substrate W while holding the substrate W. For example, the substrate holding portion 270 moves vertically upward or vertically downward along the vertical direction while holding the substrate W.

[0346] Typically, the substrate holding part 270 collectively holds a plurality of substrates W. Here, the plurality of substrates W are arranged in a row along the Y direction. Alternatively, the substrate holding part 270 may hold only one substrate W.

[0347] Figures 1 to 24 The data processing apparatus 100 shown processes time series data generated in one substrate processing apparatus 200, but the present embodiment is not limited thereto. The data processing apparatus 100 may also process time series data generated in substrate processing apparatuses 200 disposed in different locations.

[0348] Next, refer to Fig.25 The data processing apparatus 100 and the substrate processing apparatus 200 according to the present embodiment will be described. Fig.25 Schematic diagram of the data processing device 100 and the substrate processing device 200 of this embodiment. Fig.25 As shown, the substrate processing apparatus 200 includes a substrate processing apparatus 200A and a substrate processing apparatus 200B. The substrate processing apparatus 200A and the substrate processing apparatus 200B are connected to the data processing apparatus 100 in a communicable manner.

[0349] Typically, the substrate processing apparatus 200A and the substrate processing apparatus 200B are arranged in separate places. For example, the substrate processing apparatus 200A and the substrate processing apparatus 200B may be arranged in different places in the same country. Alternatively, the substrate processing apparatus 200A and the substrate processing apparatus 200B may be arranged in different countries.

[0350] The data processing apparatus 100 is, for example, a server. For example, the substrate processing apparatus 200A can communicate information with the substrate processing apparatus 200B via the data processing apparatus 100 .

[0351] In addition, in reference Fig.25 In the above description, the data processing apparatus 100 can communicate with the substrate processing apparatus 200A and the substrate processing apparatus 200B, but the substrate processing apparatus 200A and the substrate processing apparatus 200B may not exist at the same time.

[0352] Above, the embodiments of the present invention are described with reference to the accompanying drawings. However, the present invention is not limited to the embodiments, and can be implemented in various forms without departing from the scope of its main purpose. Moreover, by appropriately combining the multiple constituent elements disclosed in the embodiments, various inventions can be formed. For example, several constituent elements can also be deleted from all the constituent elements shown in the embodiments. Furthermore, the constituent elements between different embodiments can also be appropriately combined. For ease of understanding, the accompanying drawings schematically show each constituent element in the main body, and in order to facilitate the production of the accompanying drawings, the thickness, length, number, spacing, etc. of each constituent element shown in the drawings are sometimes different from the actual ones. Moreover, the material, shape, size, etc. of each constituent element shown in the embodiments are examples, and are not particularly limited, and various changes can be made within the scope of not substantially departing from the effect of the present invention.

Claims

1. A data processing method comprising the following steps: Acquiring a plurality of time series data obtained by a substrate processing device; Obtaining evaluation values ​​of each of the plurality of time series data; Based on the evaluation values, classify the plurality of time series data into any one of a plurality of categories respectively; as well as The object time series data corresponding to any one of the plurality of categories and the time series data included in the learning database are matched.

2. The data processing method according to claim 1, wherein in the learning database, cause countermeasure information is stored corresponding to the time series data, The matching step further includes the step of reading out cause countermeasure information corresponding to the time series data when a matching rate between the target time series data and at least one time series data included in the learning database is higher than a threshold value.

3. The data processing method according to claim 1 or 2, wherein in the step of obtaining the evaluation value, the evaluation value of each of the multiple time series data is obtained by comparing each of the multiple time series data with benchmark data.

4. The data processing method according to claim 3, wherein in the step of obtaining the evaluation value, the evaluation value when the difference between the value of the time series data and the value of the benchmark data is large is greater than the evaluation value when the difference between the value of the time series data and the value of the benchmark data is small.

5. The data processing method according to claim 1 or 2, further comprising the following steps: The object time series data is stored.

6. A data processing device, comprising: A data acquisition unit that acquires a plurality of time series data obtained by the substrate processing device; An evaluation value acquisition unit, which acquires evaluation values ​​of each of the plurality of time series data; a classification unit that classifies the plurality of time series data into any one of a plurality of categories based on the evaluation value; as well as The matching unit matches the object time series data corresponding to any one of the plurality of categories with the time series data included in the learning database.

7. A storage medium storing a program, wherein the program causes a computer to execute the following steps: acquiring a plurality of time series data obtained by a substrate processing device; Obtaining evaluation values ​​of each of the plurality of time series data; Based on the evaluation values, classify the plurality of time series data into any one of a plurality of categories respectively; as well as The object time series data corresponding to any one of the plurality of categories and the time series data included in the learning database are matched.

Images