Information processing method, computer program, and information processing device

Abstract

The invention provides an information processing method, a computer program, and an information processing device which are expected to assist analysis of monitoring data obtained from a target device. In an information processing method according to an embodiment, an information processing apparatus acquires monitoring data including a plurality of items related to a state of each process performed by a target apparatus, generates a dimension compression model for compressing a dimension number of the monitoring data, inputs the monitoring data to the dimension compression model, and generates a dimension compression model for compressing the dimension number of the monitoring data. The monitoring data is converted into a low-dimensional representation, and the monitoring data converted into the low-dimensional representation is output. Furthermore, a plurality of items classified into a plurality of categories may be included in the monitoring data, and the information processing device generates the dimension compression model for each of the categories, inputs the monitoring data into the dimension compression model corresponding to each of the categories, and converts the monitoring data into a low-dimensional representation for each of the categories.

Description

Technical Field

[0001] The present disclosure relates to an information processing method, a computer program, and an information processing device. Background Art

[0002] Non-Patent Document 1 describes analyzing a signal obtained from a sensor in a chamber in a semiconductor manufacturing process using principal component analysis or the like.

[0003] Non-patent literature 1: Ivan Tan, Russell Dover, Madhukar Reddy, "Enabling Machine Learning for Increased Tool Matching and Greater Process Control", AEC / APCSymposium Asia 2021 Summary of the invention

[0004] The present disclosure provides an information processing method, a computer program, and an information processing device that can be expected to assist in the analysis of monitoring data obtained from a target device.

[0005] In an information processing method involved in one embodiment, an information processing device obtains monitoring data including multiple items related to the status of each processing performed by an object device, generates a dimensional compression model for compressing the number of dimensions of the above-mentioned monitoring data, inputs the above-mentioned monitoring data into the above-mentioned dimensional compression model and converts it into a low-dimensional representation, and outputs the above-mentioned monitoring data converted into the low-dimensional representation.

[0006] According to the present disclosure, it can be expected that analysis of monitoring data obtained from a target device can be assisted. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 This is a schematic diagram for explaining an example of the information processing system involved in the first embodiment.

[0008] Figure 2 This is a block diagram showing a configuration example of an information processing device according to Embodiment 1.

[0009] Figure 3 This is a schematic diagram for explaining a configuration example of monitoring data stored in the monitoring data storage unit.

[0010] Figure 4 This is a flowchart showing an example of a procedure of visualization processing for each category performed by the information processing device according to the first embodiment.

[0011] Figure 5 is a diagram showing an example of visualization for each category.

[0012] Figure 6 This is a flowchart showing an example of a procedure for visualizing deviations in fleet units and chamber units performed by the information processing device according to the first embodiment.

[0013] Figure 7 This is a diagram showing an example of visualization of deviations in fleet units and chamber units.

[0014] Figure 8 This is a diagram showing an example of visualization of deviations in fleet units and chamber units.

[0015] Fig. 9 This is a schematic diagram showing an example of visualization of deviations in fleet units and chamber units based on principal component analysis.

[0016] Fig.10 This is a schematic diagram showing an example of visualization of deviations in fleet units and chamber units based on principal component analysis.

[0017] Fig.11 This is a schematic diagram showing an example of the structure of tag information.

[0018] Fig.12 This is a flowchart showing an example of the procedure of visualization processing based on tag information performed by the information processing device according to the first embodiment.

[0019] Fig.13 is a schematic diagram showing an example of visualization processing based on label information.

[0020] Fig.14 This is a flowchart showing an example of the procedure of visualization processing using scale conversion based on label information performed by the information processing device according to the first embodiment.

[0021] Fig.15 is a schematic diagram showing an example of visualization processing using scale conversion based on label information.

[0022] Fig.16 This is a flowchart showing an example of a procedure of visualization processing using an autoencoder performed by the information processing device according to the first embodiment.

[0023] Fig.17 is a schematic diagram showing an example of visualization processing using an autoencoder.

[0024] Fig.18 1 is a flowchart showing the steps of the chamber verification process performed in the information processing system according to the present embodiment.

[0025] Fig.19 is a diagram for explaining an example of data collection plan information.

[0026] Fig. 20 is a diagram for explaining an example of data collection plan information.

[0027] Fig.21 is a schematic diagram showing an example of sensor arrangement.

[0028] Fig. 22 This is a flowchart for explaining an outline of the analysis process performed by the information processing device according to the second embodiment.

[0029] Fig.23 This is a flowchart showing the steps of the monitoring process over time performed by the information processing device involved in the second embodiment.

[0030] Fig.24 It is a schematic diagram for explaining the monitoring process over time. DETAILED DESCRIPTION

[0031] Hereinafter, specific examples of information processing systems according to embodiments of the present disclosure will be described with reference to the accompanying drawings. In addition, the present disclosure is not limited to these examples, but is indicated by the claims, and is intended to include all modifications within the meaning and scope equivalent to the claims.

[0032] <System Structure>

[0033] Figure 1 It is a schematic diagram for explaining an example of an information processing system involved in Embodiment 1. The information processing system involved in this embodiment is configured to include an information processing device 1 and a plurality of chambers (object devices) 3. The chamber 3 is a device for manufacturing semiconductors, for example, a device for performing processes such as CVD (Chemical Vapor Deposition), sputtering, or etching. The information processing system involved in this embodiment is, for example, a system in which a plurality of chambers 3 are arranged in a semiconductor manufacturing plant, and the operation of these plurality of chambers 3 is controlled by the information processing device 1. In addition, in this embodiment, one information processing device 1 controls a plurality of chambers 3, but this is not limited to this, and a plurality of information processing devices 1 may also control a plurality of chambers 3.

[0034] Each chamber 3 performs various processes on the transported wafers to manufacture semiconductors according to the control of the information processing device 1. Each chamber 3 is equipped with multiple sensors such as temperature sensors and pressure sensors, and various information detected by these sensors when performing various processes on the wafers is output to the information processing device 1 as monitoring data related to these processes.

[0035] The information processing device 1 controls the operation of one or more chambers 3 according to the settings (recipes) related to the semiconductor process pre-entered by the user such as the operator performing the semiconductor manufacturing operation or the administrator of the system. The information processing device 1 manufactures semiconductors by performing various processes on one or more wafers through the operation of the chamber 3. The information processing device 1 receives monitoring data including various information collected by sensors and the like accompanying the operation of the chamber 3 from the chamber 3, and records the received monitoring data in a database for accumulation. In addition, the information processing device 1 involved in this embodiment performs various calculations on the monitoring data recorded in the database, and displays the calculation results visually in a chart or the like. As a result, the user can expect to analyze the causes of individual differences or malfunctions related to multiple chambers 3 based on the visualized monitoring data.

[0036] Figure 2 1 is a block diagram showing a configuration example of an information processing device 1 according to Embodiment 1. The information processing device 1 according to this embodiment is configured to include a processing unit 11, a storage unit (storage device) 12, a communication unit (transceiver) 13, a display unit 14, an operation unit 15, and the like. In addition, in this embodiment, the processing is described as being performed by one information processing device 1, but the processing of the information processing device 1 may be performed in a distributed manner by a plurality of devices.

[0037] The processing unit 11 is configured using a CPU (Central Processing Unit), an MPU (Micro-Processing Unit), a GPU (Graphics Processing Unit), or a quantum processor, etc., a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The processing unit 11 reads out and executes a program 12a stored in the storage unit 12, thereby performing various processes such as a process of controlling the operation of the chamber 3 and a process of collecting monitoring data related to the chamber 3.

[0038] The storage unit 12 is configured to use a large-capacity storage device such as a hard disk. The storage unit 12 stores various programs executed by the processing unit 11 and various data required for the processing of the processing unit 11. In the present embodiment, the storage unit 12 stores a program 12a executed by the processing unit 11. In addition, the storage unit 12 is provided with: a setting information storage unit 12b for storing various setting information related to the processing performed by the chamber 3, a monitoring data storage unit 12c for storing monitoring data obtained from the chamber 3, and a dimensional compression model information storage unit 12d for storing information related to the dimensional compression model used for the processing of dimensional compression of the monitoring data.

[0039] In the present embodiment, the program (computer program, program product) 12a is provided in the form of being recorded on a recording medium 99 such as a memory card or an optical disk, and the information processing device 1 reads the program 12a from the recording medium 99 and stores it in the storage unit 12. However, the program 12a may be written to the storage unit 12, for example, during the manufacturing stage of the information processing device 1. In addition, for example, the program 12a may be obtained by the information processing device 1 during communication as a program distributed by a remote server device or the like. For example, the program 12a may be read from the program recorded on the recording medium 99 by a writing device and written to the storage unit 12 of the information processing device 1. The program 12a may be provided in the form of distribution via a network or in the form of being recorded on the recording medium 99.

[0040] The setting information storage unit 12b of the storage unit 12 stores various setting information related to the processing performed in the chamber 3. The setting information includes information that can be called a recipe for a semiconductor process, for example, it can include settings such as the pressure and flow rate of a gas in an etching process, the voltage and frequency of a discharge, and the temperature of a wafer. In addition, the recipe includes information such as the execution time of a semiconductor process and the settings of each step of the semiconductor process. The information processing device 1 receives input of the setting information by the user through the operation unit 15, and stores the received setting information in the setting information storage unit 12b.

[0041] The monitoring data storage unit 12 c is a database that records and stores monitoring data acquired from the plurality of chambers 3 by the information processing device 1 . Figure 3 1 is a schematic diagram for explaining a configuration example of monitoring data stored in the monitoring data storage unit 12c. The monitoring data shown in the figure is data in which information detected by multiple sensors mounted in the chamber 3 is stored in correspondence with information such as "timestamp", "chamber ID", "recipe ID", "recipe step ID", "wafer ID", "statistical value type" and "flag". The information detected by the multiple sensors is classified and stored according to each type of the sensor. In this example, it is classified into categories such as "pressure category" and "temperature category".

[0042] The chamber 3 involved in the present embodiment is equipped with, for example, M pressure sensors and N temperature sensors. The M pressure sensors are arranged at appropriate positions of the chamber 3 to detect the pressure of each location in the chamber 3. In the illustrated monitoring data, the detection value of the first pressure sensor is represented as "pressure sensor 1", the detection value of the second pressure sensor is represented as "pressure sensor 2", ..., and the detection value of the Mth pressure sensor is represented as "pressure sensor M". Similarly, N temperature sensors are arranged at appropriate positions of the chamber 3 to detect the temperature of each location in the chamber 3. In the illustrated monitoring data, the detection value of the first temperature sensor is represented as "temperature sensor 1", the detection value of the second temperature sensor is represented as "temperature sensor 2", ..., and the detection value of the Nth temperature sensor is represented as "temperature sensor N". In addition, in addition to pressure sensors and temperature sensors, the chamber 3 can also be equipped with various types of sensors, and the monitoring data can include the detection values ​​of more types of sensors. The monitoring data may include information such as the flow rate, pressure, temperature or plasma emission spectrometry of gases related to semiconductor process processing, and may also include information such as maintenance information of chamber 3 or parts consumption, and may also include various types of information other than these.

[0043] The “time stamp” of the monitoring data is information such as the date and time when the processing of the wafer by the chamber 3 is started, the date and time when the detection values ​​of each sensor of the chamber 3 are obtained, the date and time when the chamber 3 generates the monitoring data, the date and time when the chamber 3 sends the monitoring data to the information processing device 1, or the date and time when the information processing device 1 obtains the monitoring data. The “chamber ID” is identification information uniquely added to each chamber 3, and can be predetermined by a user such as a designer or manager of the information processing system involved in this embodiment.

[0044] "Recipe ID" is identification information uniquely added to the setting information of the process performed in chamber 3, and is composed of characters, numbers or a combination thereof, and can be predetermined by a user such as an administrator of an information processing system or an operator performing semiconductor manufacturing operations. The recipe contains setting information about one or more processes to be performed on a wafer. In this embodiment, each process set in the recipe is referred to as a recipe step. The "recipe step ID" of the monitoring data is identification information added to one or more steps contained in the recipe specified by the above-mentioned "recipe ID".

[0045] "Wafer ID" is identification information uniquely added to the wafer processed in each chamber 3, and is composed of characters, numbers or a combination thereof. For example, each wafer can be specified by a user such as an administrator of the information processing system involved in this embodiment or an administrator of semiconductor manufacturing.

[0046] Each sensor of the chamber 3 performs multiple detections during the processing of one recipe step. In this embodiment, the value of each sensor stored in the monitoring data is the average value, minimum value or maximum value of the multiple detection results. "The type of statistical value" is information indicating which type of statistical value of multiple types such as the average value, minimum value or maximum value the value of each sensor stored as monitoring data is. In the information processing system involved in this embodiment, the chamber 3 samples and obtains the output value of each sensor at a predetermined sampling period, and stores it in its own storage unit. For example, the chamber 3 calculates the statistical value by performing a calculation using a statistical method specified in advance by the user on the output value of the time series of each sensor stored in each recipe step preset by the user. The chamber 3 sends the calculated statistical value to the information processing device 1, and the information processing device 1 receives the statistical value and stores it in the monitoring data storage unit 12c as monitoring data. In addition, the chamber 3 can also send the output value of the time series of each sensor to the information processing device 1, and the information processing device 1 calculates the statistical value based on the output value of the time series, and stores it in the monitoring data storage unit 12c as monitoring data. Alternatively, the information processing device 1 may store the time-series output value of each sensor as monitoring data instead of storing the statistical value in the monitoring data storage unit 12 c and calculate the statistical value as needed.

[0047] The "flag" is the result of processing the wafer, for example, information indicating whether a label such as a qualified product or a defective product is assigned. In the present embodiment, the "flag" is information used when, for example, storing and performing analysis processing of monitoring data. Therefore, the "flag" may not be included in the monitoring data obtained from the chamber 3, for example, it may be information added by the information processing device 1 as needed, and it may also be information added by a user such as an analyzer who analyzes the monitoring data when analyzing.

[0048] When processing a wafer, each chamber 3 repeatedly detects pressure, temperature, etc. based on a plurality of sensors mounted thereon at an appropriate cycle, such as once every 100 milliseconds or once every second. Each chamber 3 obtains the detection value of the sensor at this cycle, and each time the processing of a recipe step is completed, the statistical value of the plurality of detection values ​​is calculated, and the monitoring data shown in the figure is generated and sent to the information processing device 1. The information processing device 1 receives the monitoring data from the plurality of chambers 3, and stores the received monitoring data in the monitoring data storage unit 12c. In this case, the monitoring data of the number of recipe steps for processing a wafer is stored in the monitoring data storage unit 12c.

[0049] However, the information processing device 1 may store the detection values ​​of each sensor itself instead of storing the multiple detection values ​​of each sensor as statistical values. In this case, monitoring data of the number of times the sensor detects a wafer is stored in the monitoring data storage unit 12c. In addition, when the detection cycles of the multiple sensors provided in the chamber 3 are different, for example, the monitoring data can be created in accordance with the sensor with the shortest cycle. In addition, for example, the information processing device 1 may also obtain the detection value detected by each sensor as a representative value at a specified timing for processing the wafer, for example, 10 seconds after the start of processing, and store monitoring data containing the representative value. In addition, when the information processing device 1 stores statistical values, the entire interval in a recipe step may not be used as the object of statistical value calculation, for example, only the first and last specified intervals excluding the recipe step may be used as the object of statistical value calculation.

[0050] The dimension compression model information storage unit 12d of the storage unit 12 stores information related to the dimension compression model generated by the information processing device 1 based on the monitoring data stored in the monitoring data storage unit 12c. The information related to the dimension compression model may include, for example, structural information indicating how the dimension compression model is structured, and information such as the values ​​of parameters inside the dimension compression model. The information processing device 1 involved in this embodiment uses the dimension compression model stored in the dimension compression model information storage unit 12d to perform dimension compression on the monitoring data, so that the monitoring data having multiple values ​​(i.e., multiple dimensions) can be displayed in a two-dimensional or three-dimensional manner.

[0051] The communication unit 13 of the information processing device 1 is connected to the plurality of chambers 3 via cables such as communication lines or signal lines, and transmits and receives data with the chambers 3 via the cables. In addition, the communication unit 13 may also be a structure that communicates with the chambers 3 by wireless communication. In the present embodiment, the communication unit 13 transmits control data provided from the processing unit 11 to the chamber 3. In addition, the communication unit 13 receives monitoring data transmitted from the chamber 3, and provides the received monitoring data to the processing unit 11.

[0052] The display unit 14 is configured to use a liquid crystal display or the like, and displays various images and characters based on the processing of the processing unit 11. The display unit 14 displays, for example, various information related to the operating state of the chamber 3, and displays a warning message or the like when an abnormal operating state of the chamber 3 is detected. In addition, in the present embodiment, the display unit 14 displays a screen for analyzing individual differences or malfunctions of the chamber 3 based on monitoring data acquired from the plurality of chambers 3.

[0053] The operation unit 15 receives the user's operation and notifies the processing unit 11 of the received operation. For example, the operation unit 15 receives the user's operation through an input device such as a mechanical button or a touch panel provided on the surface of the display unit 14. In addition, for example, the operation unit 15 may be an input device such as a mouse and a keyboard, and these input devices may also be detachable from the information processing device 1.

[0054] In addition, the storage unit 12 may be an external storage device connected to the information processing device 1. In addition, the information processing device 1 may be a multi-computer including a plurality of computers, or a virtual machine constructed by software. In addition, the information processing device 1 is not limited to the above-mentioned structure, and for example, it may not include the display unit 14 and the operation unit 15.

[0055] In addition, in the information processing device 1 involved in this embodiment, the processing unit 11 reads out the program 12a stored in the storage unit 12 and executes it, so that the control processing unit 11a, the monitoring data acquisition unit 11b, the dimensional compression model generation unit 11c, the data conversion unit 11d and the display processing unit 11e are implemented as functional units of the software in the processing unit 11.

[0056] The control processing unit 11a performs processing for controlling the operations of the plurality of chambers 3 based on the setting information stored in the setting information storage unit 12b. For example, when etching a wafer is performed in the chamber 3, the control processing unit 11a acquires information such as the pressure and flow rate, the voltage and frequency of the discharge, and the temperature of the wafer set in the setting information. The control processing unit 11a transmits the acquired setting information to each chamber 3, thereby enabling the chamber 3 to perform the etching process under the set conditions.

[0057] In addition, the information processing device 1 involved in this embodiment is not limited to the chamber 3 as the device to be monitored, and can be various substrate processing devices, and further can be hardware units such as high-frequency units or gas supply units that constitute these devices. In addition, the processing performed by these substrate processing devices is not limited to etching processing, for example, it can be various processing such as process processing of substrates (wafers), pre- and post-processing of substrates, conveying processing of substrates, waferless dry cleaning processes, plasma health inspection processes, and automatic inspection processes of sensors. The information processing device 1 controls the operation of the substrate processing device and stores and accumulates monitoring data obtained from the substrate processing device.

[0058] The monitoring data acquisition unit 11b acquires monitoring data from the chamber 3 that performs processing on the wafer, and stores the acquired monitoring data in the monitoring data storage unit 12c. For example, when the chamber 3 performs processing on the wafer, the multiple sensors mounted thereon repeatedly detect pressure, temperature, etc. at a predetermined period, and the detection values ​​of the multiple sensors are summarized and sent to the information processing device 1 as monitoring data. The monitoring data acquisition unit 11b receives these monitoring data periodically sent by the multiple chambers 3 accompanying the execution of the processing through the communication unit 13, and acquires the acquired monitoring data, and stores the acquired monitoring data in the monitoring data storage unit 12c. At this time, the monitoring data acquisition unit 11b may also perform processing such as adding a time stamp to the monitoring data acquired from the chamber 3 or performing arithmetic processing on the detection values ​​of the sensors included in the monitoring data as needed. For example, the statistical value of the detection value of the sensor is stored as the monitoring data, but in the case where the chamber 3 does not calculate the statistical value but sends the detection value of the sensor, the monitoring data acquisition unit 11b may also calculate the statistical value. In this case, for example, the monitoring data acquisition unit 11b extracts monitoring data common to the chamber ID, recipe ID, recipe step ID, and wafer ID for a plurality of time series data periodically received from the chamber 3. The monitoring data acquisition unit 11b calculates statistical values ​​such as the average value of the detection value of each sensor for the plurality of extracted time series data. The monitoring data acquisition unit 11b can generate a set of monitoring data including the calculated statistical values ​​for a recipe step performed in the chamber 3, and store the generated set of monitoring data in the monitoring data storage unit 12c. In addition, in the case where a plurality of types of statistical values ​​can be stored for a recipe step, monitoring data of a set of the number of types of statistical values ​​can be stored in the monitoring data storage unit 12c for a recipe step.

[0059] The dimension compression model generation unit 11c uses a plurality of monitoring data stored in the monitoring data storage unit 12c to generate a dimension compression model for dimensionally compressing the monitoring data. In the present embodiment, the dimension compression model generation unit 11c processes the monitoring data in units of wafers, and generates a dimension compression model for dimensionally compressing the monitoring data in order to visualize the differences in the processing of each wafer. In addition, in the present embodiment, the dimension compression model generation unit 11c generates a dimension compression model by a machine learning method of so-called unsupervised learning based on PCA (Principal Component Analysis), but is not limited to this. The dimension compression model generation unit 11c can generate a dimension compression model by methods such as ICA (Independent Component Analysis), factor analysis, t-SNE (t-distributed Stochastic Neighbor Embedding), hierarchical clustering or LSA (Latent Semantic Analysis). In addition, the generation of the dimension compression model based on these methods is a prior art, so a detailed description is omitted. The dimensional compression model generation unit 11 c stores information on the generated dimensional compression model in the dimensional compression model information storage unit 12 d.

[0060] The data conversion unit 11d performs a process of converting the monitoring data stored in the monitoring data storage unit 12c into display data after the dimension is compressed using the dimension compression model stored in the dimension compression model information storage unit 12d. The data conversion unit 11d obtains the monitoring data with the number of dimensions reduced, such as two-dimensional or three-dimensional, output by the dimension compression model by inputting each monitoring data of other dimensions stored in the monitoring data storage unit 12c into the dimension compression model, thereby converting the monitoring data into display data.

[0061] The display processing unit 11e performs processing to visualize the display data converted by the data conversion unit 11d, that is, the monitoring data reduced to two dimensions or three dimensions, and displays it on the display unit 14. In the present embodiment, the display processing unit 11e creates a graph such as a histogram, a box plot, a violin plot, a bee swarm plot, a scatter plot, a density contour map, and a heat map based on the two-dimensional display data, and displays the created graph on the display unit 14.

[0062] <Monitoring data display processing>

[0063] The information processing device 1 visualizes the monitoring data stored in the monitoring data storage unit 12c and displays it on the display unit 14. In the present embodiment, the information processing device 1 visualizes the monitoring data on a wafer basis. The information processing device 1 according to the present embodiment displays the monitoring data using, for example, at least one of the following methods (1) to (5).

[0064] (1) Visualization processing of each category

[0065] (2) Visualization of deviations by fleet and by chamber

[0066] (3) Visualization processing based on label information

[0067] (4) Visualization processing using scale conversion based on label information

[0068] (5) Visualization processing using autoencoders

[0069] (1) Visualization processing of each category

[0070] In the information processing system according to the present embodiment, information (such as sensor detection values) included in monitoring data is classified into a plurality of categories. Figure 3 In the example of monitoring data shown, information is classified into multiple categories such as pressure category and temperature category. The information processing device 1 involved in this embodiment generates a dimensional compression model for each category in the unit of wafer of monitoring data, uses the generated dimensional compression model to perform dimensional compression of monitoring data for each category, and visualizes the monitoring data in the unit of wafer for each category.

[0071] Figure 4 1 is a flowchart showing an example of the steps of visualization processing for each category performed by the information processing device 1 according to Embodiment 1. The monitoring data acquisition unit 11b of the processing unit 11 of the information processing device 1 according to this embodiment reads a plurality of monitoring data stored in the monitoring data storage unit 12c of the storage unit 12 (step S1). At this time, the monitoring data acquisition unit 11b can receive conditions such as the period of the monitoring data to be visualized from the user, and in this case, the monitoring data of the period etc. received as the condition is extracted.

[0072] The dimensional compression model generation unit 11c of the processing unit 11 selects one category as a processing target from the multiple categories included in the monitoring data (step S2). The dimensional compression model generation unit 11c generates a dimensional compression model that compresses the number of dimensions of the category by, for example, a principal component analysis method, with respect to the monitoring data read out in step S1. The dimensional compression model generation unit 11c stores information related to the dimensional compression model generated in step S3 in the dimensional compression model information storage unit 12d of the storage unit 12 (step S4).

[0073] For example, in Figure 3 In the monitoring data shown, the pressure category includes information on the detection values ​​of M sensors, and the temperature category includes information on the detection values ​​of N sensors. For example, when a wafer is processed with a recipe step of P and Q types of values ​​are stored as statistical values ​​as monitoring data, the monitoring data for the pressure category is (M×P×Q) dimensional data. The dimensional compression model generation unit 11c generates a dimensional compression model that compresses the (M×P×Q) dimensional data into two-dimensional data. Similarly, in Figure 3 In the example, for the temperature category, the monitoring data is (N×P×Q)-dimensional data, and the dimensional compression model generation unit 11 c generates a dimensional compression model that compresses the (N×P×Q)-dimensional data into two-dimensional data.

[0074] In addition, for example, when there are M detection values, a recipe step of a P step, and Q types of statistical values, as described above, the monitoring data can be of (M×P×Q) dimensions, but not all of the information of the M detection values, the recipe steps of the P step, and the statistical values ​​of the Q types may be the object of dimensional compression. For example, the user may select any M′ detection values ​​from the M detection values, select the recipe step of the P′ step in the P step, and select any Q′ type of statistical value from the Q types. The information processing device 1 can generate monitoring data of (M′×P′×Q′) dimensions with the selected combination, and generate a dimensional compression model for dimensional compression of the generated monitoring data.

[0075] Next, the data conversion unit 11d of the processing unit 11 inputs the monitoring data read out in step S1 to the dimensional compression model generated in step S3, obtains the data output by the dimensional compression model, and converts the monitoring data of three or more dimensions into display data compressed to two dimensions (step S5). The data conversion unit 11d generates a scatter plot for visualizing the display data obtained in step S5 (step S6). In addition, in this flowchart, as an example, visualization is performed by displaying a scatter plot, but it is not limited to this, and various charts other than scatter plots can be displayed.

[0076] The processing unit 11 determines whether the dimensional compression and scatter plot creation processes are completed for all the categories included in the monitoring data (step S7). If the process is not completed for all the categories (S7: No), the processing unit 11 returns the process to step S2, selects another category and repeats the same process. If the process is completed for all the categories (S7: Yes), the display processing unit 11e of the processing unit 11 displays the scatter plot for each category generated in step S6 on the display unit 14 (step S8), and the process ends.

[0077] Figure 5 is a schematic diagram showing an example of visualization for each category. The illustrated example is an example of displaying a scatter plot for each category, and the scatter plot is formed by compressing three-dimensional or higher monitoring data into two dimensions and plotting one monitoring data as a point on a two-dimensional plane. The information processing device 1 generates a scatter plot for two categories, category 1 and category 2, included in the monitoring data, and displays the two scatter plots in a vertical arrangement on the display unit 14. In the present embodiment, a point displayed in the scatter plot by the information processing device 1 is a point whose coordinates are determined based on two-dimensional information obtained by dimensional compression of multiple information obtained for a wafer processed in a semiconductor manufacturing process.

[0078] The portion surrounded by the dotted ellipse in the scatter plot of category 1 is a collection of monitoring data about a specific chamber 3, and the monitoring data of the chamber 3 deviates from the monitoring data of other chambers 3. The monitoring data of the specific chamber 3 exists in the portion surrounded by the dotted ellipse in the scatter plot of category 2, and no deviation from the monitoring data of other chambers 3 is found in the scatter plot of category 2. From these circumstances, it is inferred that the specific chamber 3 in category 1 shows different behavior or characteristics from other chambers 3, and there is an abnormality or individual difference with respect to category 1.

[0079] (2) Visualization of deviations by fleet and by chamber

[0080] The information processing system involved in the present embodiment has a plurality of chambers 3, and the collection of the plurality of chambers 3 can be called a fleet. In the information processing system involved in the present embodiment, for a wafer processed in a chamber 3, one or more monitoring data corresponding to the sampling period of the sensor and the like are stored and accumulated. The information processing device 1 involved in the present embodiment calculates an index of deviation in chamber units and an index of deviation in fleet units for each of these monitoring data. Based on the calculated index, the information processing device 1 visualizes each monitoring data as two-dimensional data of an index of deviation in chamber units and an index of deviation in fleet units.

[0081] Each monitoring data includes the statistical value of the detection values ​​of multiple sensors classified into multiple categories as described above. Each detection value may have a different value range and unit for each sensor. The information processing device 1 performs a scale conversion of the statistical value of the detection value in advance for each sensor, so as to convert all the statistical values ​​included in the monitoring data into, for example, an average of 0 and a standard deviation of 1. Thus, the information processing device 1 can process each monitoring data including the statistical value of the detection value of multiple sensors classified into multiple categories into unified data including numerical information of multiple specified ranges regardless of the distinction of categories. For example, the information processing device 1 obtains the average value and standard deviation of the statistical value of the detection value of a sensor included in the monitoring data, and divides the statistical value by the standard deviation after subtracting the average value from the statistical value of the detection value of the sensor, so as to perform data conversion. In addition, the method of data conversion is an example and is not limited to this. The information processing device 1 can also perform data conversion by any method.

[0082] The information processing device 1 involved in this embodiment uses the Mahalanobis distance as an indicator of the deviation of the monitoring data. The Mahalanobis distance is a scalar value representing the distance of each vector relative to the set of multidimensional vectors. The calculation method of the Mahalanobis distance is a prior art, so the detailed description is omitted. The Mahalanobis distance is calculated based on the mean and covariance matrix of the set of vectors. In addition, as an indicator of the deviation of the monitoring data, in addition to the Mahalanobis distance, other distance indicators such as the Euclidean distance or the Chebyshev distance can also be used.

[0083] Figure 6 1 is a flowchart showing an example of the steps of the visualization processing of the deviation in the fleet unit and the chamber unit by the information processing device 1 according to the first embodiment. The monitoring data acquisition unit 11b of the processing unit 11 of the information processing device 1 according to the present embodiment reads out a plurality of monitoring data stored in the monitoring data storage unit 12c of the storage unit 12 (step S21). The monitoring data read out in step S21 is data in which the monitoring data on the plurality of chambers 3 are mixed.

[0084] The dimensional compression model generation unit 11c of the processing unit 11 uses all the monitoring data read out in step S21 to perform scale conversion on each value (statistical value of the detection value of the sensor) contained in each monitoring data (step S22). The dimensional compression model generation unit 11c calculates the covariance matrix for all the multi-dimensional monitoring data after the scale conversion in step S22 (step S23). The dimensional compression model generation unit 11c uses the covariance matrix calculated in step S23 to calculate the Mahalanobis distance of each monitoring data relative to the set of all monitoring data (step S24). The Mahalanobis distance calculated in step S24 becomes an indicator of the deviation of each monitoring data in fleet units.

[0085] Next, the dimensional compression model generation unit 11c extracts monitoring data about a cavity 3 from the monitoring data read out in step S21 (step S25). The dimensional compression model generation unit 11c uses the monitoring data about a cavity 3 extracted in step S25 to perform scale conversion on each value (sensor detection value) contained in each monitoring data (step S26). The dimensional compression model generation unit 11c calculates the covariance matrix for the monitoring data after the scale conversion in step S26 (step S27). The dimensional compression model generation unit 11c uses the covariance matrix calculated in step S27 to calculate the Mahalanobis distance of each monitoring data relative to the set of monitoring data (step S28). The Mahalanobis distance calculated in step S28 becomes an indicator of the deviation of each monitoring data in units of cavities.

[0086] The dimensional compression model generation unit 11c determines whether the calculation process of the Mahalanobis distance in units of cavities 3 is completed for all cavities 3 (step S29). If the process is not completed for all cavities 3 (S29: No), the dimensional compression model generation unit 11c returns the process to step S25, extracts the monitoring data about other cavities 3, and repeats the same process. If the process is completed for all cavities 3 (S29: Yes), the dimensional compression model generation unit 11c stores the covariance matrix related to all monitoring data calculated in step S23 and the covariance matrix of each cavity 3 calculated in step S27 as information related to the dimensional compression model in the dimensional compression model information storage unit 12d of the storage unit 12 (step S30).

[0087] Through the processing of the above steps S21 to S30, the Mahalanobis distance in fleet units and the Mahalanobis distance in chamber units are calculated for each monitoring data. The display processing unit 11e of the processing unit 11, for example, uses the Mahalanobis distance in fleet units as the horizontal axis and the Mahalanobis distance in chamber units as the vertical axis to generate a scatter plot of each monitoring data (step S31). The display processing unit 11e displays the scatter plot generated in step S31 on the display unit 14 (step S32), and ends the processing.

[0088] Figure 7 This is a schematic diagram showing an example of visualization of deviations in fleet units and chamber units. The example shown in the figure is an example of displaying a scatter plot, which is a scatter plot in which the horizontal axis is the Mahalanobis distance in fleet units and the vertical axis is the Mahalanobis distance in chamber units, and one monitoring data is plotted as a point on a two-dimensional plane. In the scatter plot of this example, monitoring data with small deviations in fleet units and chamber units are plotted in the area close to the origin at the lower left of the scatter plot. The monitoring data plotted in the area at the lower right of the scatter plot (see area A surrounded by dotted lines) has a large deviation in fleet units and a small deviation in chamber units. The monitoring data plotted in the area at the upper left of the scatter plot (see area B surrounded by dotted lines) has a large deviation in chamber units and a small deviation in fleet units. The monitoring data plotted in the area at the upper right of the scatter plot (see area C surrounded by dotted lines) has a large deviation in fleet units and a large deviation in chamber units.

[0089] In addition, the information processing device 1 divides the plurality of monitoring data plotted on the two-dimensional plane into a plurality of monitoring data groups based on the results of the visualization of the deviations in fleet units and in chamber units. Figure 7 In the scatter plot shown, four monitoring data groups can be divided into three monitoring data groups of areas A to C surrounded by dotted lines and monitoring data groups of other monitoring data not surrounded by dotted lines. Division into multiple monitoring data groups can also be performed automatically by the information processing device 1 based on a predetermined division rule, for example, or can be performed by the information processing device 1 receiving a user's selection operation.

[0090] The information processing device 1 reacquires the original monitoring data for each of these monitoring data groups, generates a dimensionality compression model based on principal component analysis, and performs dimensionality compression on the monitoring data. The information processing device 1 uses the dimensionality compressed monitoring data to create, for example, a two-dimensional scatter plot for each monitoring data group, and displays a plurality of scatter plots in a row on the display unit 14.

[0091] Figure 8 This is a diagram showing an example of visualization of deviations in fleet units and chamber units. Figure 8 The example shown is to Figure 7 In the example shown, the monitoring data included in region A and region C are each regarded as a monitoring data group, and the two monitoring data groups are subjected to dimensionality compression based on principal component analysis again. Figure 8In the example shown, the information processing device 1 creates a scatter plot with the first principal component and the second principal component selected by dimensionality compression as the horizontal axis and the vertical axis for each of the two monitoring data groups. The information processing device 1 arranges the scatter plots created for each monitoring data group on Figure 7 In this case, the information processing device 1 can be located at an appropriate position on the scatter plot of Figure 7 The position of each monitoring data group in the scatter plot is displayed, for example, the center position of a plurality of monitoring data included in the monitoring data group is displayed. However, when the display position of each scatter plot is determined by this method, if there is a possibility that part or all of the scatter plots overlap, the information processing device 1 can appropriately adjust the position to eliminate the overlap of the scatter plots.

[0092] The information processing system according to the present embodiment can visualize the deviations in fleet units and chamber units based on principal component analysis separately from the visualization of the deviations in fleet units and chamber units based on the Mahalanobis distance described above. Fig. 9 as well as Fig.10 This is a schematic diagram showing an example of visualization of deviations in fleet units and chamber units based on principal component analysis.

[0093] Fig. 9 The example shown is an example in which the information processing device 1 performs dimensionality compression based on principal component analysis on the monitoring data of all chambers 3 of the fleet, and creates a scatter plot with the first principal component and the second principal component as the horizontal axis and the vertical axis. In the scatter plot of this example, a plurality of monitoring data are divided into seven monitoring data groups A to G surrounded by dotted lines. In each monitoring data group A to G, a plurality of points representing the monitoring data belonging to the data group and a star indicating the center position of the distribution of these plurality of monitoring data are shown. In addition, the star 100 not included in any of the monitoring data groups A to G indicates the center position of the distribution of all monitoring data.

[0094] The information processing device 1 divides all the monitoring data into a plurality of monitoring data groups based on the result of dimensionality compression based on the principal component analysis for the monitoring data of all the chambers 3. The division into a plurality of monitoring data groups may be performed automatically by the information processing device 1 based on a predetermined division rule, or may be performed by the information processing device 1 receiving a user's selection operation, for example. The information processing device 1 further performs dimensionality compression based on the principal component analysis for each of the plurality of monitoring data groups.

[0095] Fig.10 The example shown is in Fig. 9In the example shown, the monitoring data groups A to G are individually dimensionally compressed through principal component analysis, and the result of the dimension compression is re-shown in the original scatter plot. The information processing device 1, for example, obtains the monitoring data of a monitoring data group A and performs dimension compression, thereby creating a scatter plot with the obtained new first principal component and second principal component as the horizontal axis and the vertical axis. The information processing device 1, for example, overlaps the created scatter plot on the original scatter plot so that the center position of the distribution of the created new scatter plot is consistent with the center position of the distribution of the corresponding monitoring data group A in the original scatter plot. In addition, the information processing device 1 determines the size of the new scatter plot and overlaps it on the original scatter plot so that the scale of the created new scatter plot is consistent with the scale of the original scatter plot. The information processing device 1 similarly performs dimension compression and creates scatter plots for all monitoring data groups A to G, and performs Fig.10 Display shown.

[0096] The information processing system according to the present embodiment can visualize the deviations in fleet units and chamber units by at least one of a method using Mahalanobis distance and a method using principal component analysis.

[0097] (3) Visualization processing based on label information

[0098] In the information processing system according to the present embodiment, for example, for each wafer processed in each chamber 3 , the information processing device 1 stores the processing permission (for example, OK / NG or good / bad) of the wafer as label information. Fig.11 : is a schematic diagram showing an example of the structure of label information. The label information shown in the figure is information that establishes a correspondence between "chamber ID", "recipe ID", "wafer ID" and "label". "Chamber ID", "recipe ID" and "wafer ID" are associated with Figure 3 The information included in the monitoring data shown is the same and is used to associate the monitoring data with the label information. For the wafer processed in the chamber 3, the "label" is the judgment result of "OK (good product)" or "NG (bad product)".

[0099] In addition, it is not necessary to assign label information to all of the monitoring data to be visualized, and label information can be assigned to only a portion of the monitoring data. Figure 3 The information processing system according to this embodiment uses the portion of monitoring data to which label information is added to visualize label information that can assist the user in inferring the monitoring data to which label information is not added when label information is added to a portion of monitoring data.

[0100] The label information is generated based on the result of the determination of whether the wafer can be processed after the wafer passes through the chamber 3, for example, in a test process. The label information can be input by a user using the operation unit 15 of the information processing device 1, or can be generated by a test device, for example. In addition, the information processing device 1 generates the monitoring data to which the label information is added. Fig.11 A predetermined value (eg, '1') is set in the "flag" shown. In the present embodiment, label information is managed separately from monitoring data, but for example, a "label" item may be provided in monitoring data to manage monitoring data and label information together.

[0101] The information processing device 1 involved in this embodiment obtains the monitoring data stored in the monitoring data storage unit 12c and the label information created by establishing a correspondence with these monitoring data, and performs visualization processing to separate the monitoring data labeled "OK" and the monitoring data labeled "NG". First, the information processing device 1 generates a dimensional compression model that compresses the dimension of the monitoring data based on the multiple monitoring data read from the monitoring data storage unit 12c by the PCA method. In addition, in the monitoring data used to generate the dimensional compression model, the monitoring data to which the label information is attached and the monitoring information to which the label information is not attached can be mixed together. In the PCA method, for example, N principal components can be generated for N-dimensional data, and the contribution (importance) can be calculated for each principal component. The information processing device 1 can perform data dimensional compression by selecting a specified number (for example, two) of principal components with high contribution from the top of the N principal components.

[0102] In the visualization processing based on the label information, the information processing device 1 selects a predetermined number of principal components (for example, two) from N principal components, and creates a combination of two principal components. For example, when there are 100 principal components and two principal components are selected from these 100 principal components to create a combination, the information processing device 1 can create about 4950 combinations. The information processing device 1 selects a combination of several principal components with high contribution from the combination of principal components created in this way, for example, from the top, and generates a scatter plot of the monitoring data in each selected combination. The information processing device 1 distinguishes, for example, monitoring data labeled "OK", monitoring data labeled "NG", and monitoring data not labeled by color classification, and displays them in a scatter plot. The information processing device 1 prompts the user by, for example, arranging and displaying a plurality of scatter plots corresponding to a plurality of combinations on the display unit 14, and accepts the user's selection of a scatter plot, that is, a combination of principal components. For example, based on the displayed plurality of scatter plots, the user selects a combination of principal components that most separates the monitoring data labeled "OK" and the monitoring data labeled "NG". The information processing device 1 adopts the selected group of principal components as the group of principal components used for visualization, and displays a scattergram that visualizes the monitoring data using the group of principal components.

[0103] In addition, the method for selecting the combination of principal components described above is that the user manually selects it, but the information processing device 1 can also select the combination of principal components by performing a prescribed operation. The information processing device 1 can, for example, use the method of multivariate analysis of variance to select the combination of two principal components that can best separate the labels OK and NG. In addition, when selecting a combination of two principal components, it is appropriate to perform a 2-variable analysis of variance. The information processing device 1, for example, calculates the F value based on the multivariate analysis of variance for each combination of the two principal components, and calculates the F value for all combinations (or several combinations with high contribution). The information processing device 1 can select the combination of two principal components that best separates the labels by selecting the combination of principal components with the largest F value.

[0104] The F value based on multivariate analysis of variance can be calculated using, for example, the Wilks statistic (Wilk's lambda) and the degrees of freedom. The Wilks statistic is obtained by dividing the determinant of the square sum matrix of the average error calculated for each label by the determinant of the square sum matrix of the average error of the data as a whole. In addition to the Wilks statistic, the F value can also be calculated based on statistics such as Pillai trace, Hotelling trace, or Roy's greatest root. In addition, the calculation of the F value based on multivariate analysis of variance is a prior art, so a detailed description is omitted. In addition, the selection of the principal component by the information processing device 1 is not limited to the method using the F value of multivariate analysis of variance, and various existing methods can be used, such as the method using the regression coefficient of logistic regression or the Gini coefficient of random forest.

[0105] Fig.12 1 is a flowchart showing an example of the steps of visualization processing based on label information performed by the information processing device 1 involved in Embodiment 1. The monitoring data acquisition unit 11b of the processing unit 11 of the information processing device 1 involved in this embodiment reads a plurality of monitoring data stored in the monitoring data storage unit 12c of the storage unit 12 (step S41). Next, the dimensionality compression model generation unit 11c of the processing unit 11 generates a dimensionality compression model that compresses the number of dimensions of the monitoring data read out in step S41, for example, by a principal component analysis method (step S42). The dimensionality compression model generation unit 11c stores information related to the dimensionality compression model generated in step S42 in the dimensionality compression model information storage unit 12d of the storage unit 12 (step S43). The data conversion unit 11d of the processing unit 11 inputs the monitoring data read out in step S41 to the dimensionality compression model generated in step S42, acquires the data output by the dimensionality compression model, and thereby performs data conversion of the monitoring data (step S44).

[0106] Next, the processing unit 11 reads the label information created in advance and stored in the storage unit 12 (step S45). The processing unit 11 selects a combination of a predetermined number (for example, two) of principal components that can best separate the monitoring data to which a predetermined label (for example, NG label) is attached from other monitoring data from the plurality of principal components included in the monitoring data after data conversion (step S46). In addition, the selection of the principal component can be performed automatically by the information processing device 1 as described above, for example, or the user can select it, for example.

[0107] The display processing unit 11e of the processing unit 11 generates a scatter plot that visualizes the monitoring data after data conversion in step S44 using the combination of principal components selected in step S46 (step S47). The display processing unit 11e displays the scatter plot generated in step S47 on the display unit 14 (step S48), and ends the processing.

[0108] Fig.13 : is a schematic diagram showing an example of visualization processing based on label information. The illustrated example is an example of an information processing device 1 displaying a scatter plot, in which two appropriate principal components are selected from a plurality of principal components obtained by a principal component analysis method of monitoring data, the horizontal axis is the first principal component, the vertical axis is the second principal component, and the monitoring data is plotted on a two-dimensional plane. In addition, the information processing device 1 plots the monitoring data with the OK mark in the scatter plot using dots of different colors corresponding to the marks attached to each monitoring data. In this example, the monitoring data with the OK mark attached is plotted in the scatter plot using dark dots, and the monitoring data with the NG mark attached is plotted in the scatter plot using light dots. In the illustrated scatter plot, the monitoring data with the NG mark attached is concentrated in the area on the right (refer to the area surrounded by the dotted line), and is separated from the monitoring data with the OK mark attached.

[0109] In addition, in this example, the labels added to the monitoring data are set to two types, "OK" and "NG", but this is not limited to this, and three or more labels may be added to the monitoring data. In addition, the types of labels are not limited to "OK" and "NG", and various labels other than these may be added to the monitoring data. For example, the information processing device 1 can accept the user's operation of selecting a label to determine which label among the multiple labels to focus on to select the main component of the separated monitoring data.

[0110] (4) Visualization processing using scale conversion based on label information

[0111] The information processing device 1 involved in this embodiment can scale the multiple values ​​(sensor detection values) contained in the monitoring data, convert each value into a value in the range of 0 to 1 or -1 to 1, and use the scaled value to generate a dimensional compression model. The information processing device 1, for example, extracts the minimum value and the maximum value from the multiple values, and for a certain value, the value obtained by subtracting the minimum value from the value is divided by the difference between the maximum value and the minimum value. The value obtained can be set as the scaled value. This is a method called the so-called minimum-maximum scale conversion, which is a method of converting each value into a value in the range of 0 to 1. However, the scale conversion of the information processing device 1 is not limited to the minimum-maximum scale conversion. The information processing device 1 can adopt various scale conversion methods, such as scaling so that the average value becomes 0, the standard deviation becomes 1, etc.

[0112] In the visualization process using the scale conversion based on the label information, the information processing device 1 extracts the monitoring data to which the specified label (e.g., the label of OK) is attached, and determines the scale conversion rule (e.g., the minimum value and the maximum value in the minimum-maximum scale conversion) using only the extracted monitoring data with the specified label. In addition, in the monitoring data of the object to be visualized by the information processing device 1, the monitoring data to which the label information is attached and the monitoring data to which the label information is not attached may be mixed. Furthermore, the information processing device 1 performs the scale conversion of all the monitoring data using the scale conversion rule determined based on the monitoring data with the specified label.

[0113] Next, the information processing device 1 generates a dimensional compression model using the scale-converted monitoring data. At this time, the information processing device 1 may, for example, generate the dimensional compression model using only the monitoring data to which a predetermined label is attached, or may, for example, generate the dimensional compression model using all the monitoring data regardless of the presence or absence of a label. The information processing device 1 converts all the monitoring data using the generated dimensional compression model, thereby generating, for example, two-dimensional display data, and displays a chart such as a scatter plot based on the generated display data.

[0114] Fig.14 This is a flowchart showing an example of the steps of visualization processing using scale conversion based on label information performed by the information processing device 1 according to Embodiment 1. The monitoring data acquisition unit 11b of the processing unit 11 of the information processing device 1 according to this embodiment reads a plurality of monitoring data stored in the monitoring data storage unit 12c of the storage unit 12 (step S61). In addition, the processing unit 11 reads label information created in advance and stored in the storage unit 12 (step S62).

[0115] The processing unit 11 extracts the monitoring data to which a predetermined label (e.g., an OK label) is added from the monitoring data read in step S61 based on the label information read in step S62 (step S63). The processing unit 11 determines a scale conversion rule based on the monitoring data extracted in step S63 (step S64). The processing unit 11 performs scale conversion on all the monitoring data read in step S61 based on the rule determined in step S64 (step S65).

[0116] Next, the dimensional compression model generation unit 11c of the processing unit 11 generates a dimensional compression model that compresses the number of dimensions of the monitoring data after the scale conversion in step S64 (step S66). In addition, the dimensional compression model generation unit 11c can generate the dimensional compression model using only the monitoring data to which the specified label is added, or can generate the dimensional compression model using all the monitoring data. The processing unit 11 stores the scale conversion rule determined in step S64 and the information related to the dimensional compression model generated in step S66 in the dimensional compression model information storage unit 12d of the storage unit 12 (step S67). The data conversion unit 11d of the processing unit 11 inputs the monitoring data after the scale conversion in step S64 to the dimensional compression model generated in step S66, obtains the data output by the dimensional compression model, and performs data conversion of the monitoring data (step S68). The display processing unit 11e of the processing unit 11 generates a scatter plot that visualizes the monitoring data after the data conversion in step S68 (step S69). The display processing unit 11 e displays the scattergram generated in step S69 on the display unit 14 (step S70 ), and ends the processing.

[0117] Fig.15 : is a schematic diagram showing an example of visualization processing using scale conversion based on label information. The example shown in the figure is an example of an information processing device 1 displaying a scatter plot, which is formed by determining a scale conversion rule based on monitoring data with an OK label, performing scale conversion on all monitoring data, compressing the monitoring data into two dimensions by a principal component analysis method, and plotting the monitoring data on a two-dimensional plane with the horizontal axis as the first principal component and the vertical axis as the second principal component. In addition, the information processing device 1 plots the monitoring data with the OK label in the scatter plot using dots of different colors corresponding to the marks attached to each monitoring data. In this example, the monitoring data with the OK label is plotted in the scatter plot using dark dots, and the monitoring data with the NG label is plotted in the scatter plot using light dots. In the illustrated scatter plot, the monitoring data with the NG label is concentrated in the upper area (refer to the area surrounded by the dotted line), and is separated from the monitoring data with the OK label other than this.

[0118] (5) Visualization processing using autoencoders

[0119] The information processing device 1 involved in the present embodiment, for example, performs machine learning processing using monitoring data to which an OK label is attached, thereby generating an autoencoder. The autoencoder is a learning model composed of an encoder that converts the input data into low-dimensional data and a decoder that converts the low-dimensional data into the original data. In addition, the autoencoder is a prior art, so the detailed description of the process of generating the autoencoder is omitted. The autoencoder generated using the monitoring data of the OK label can convert the monitoring data of the input OK label into low-dimensional data and restore the low-dimensional data to the original monitoring data. However, the autoencoder generated using the monitoring data of the OK label cannot convert the monitoring data of the input NG label into low-dimensional data and restore the low-dimensional data to the original monitoring data.

[0120] The information processing device 1 involved in this embodiment inputs all monitoring data including monitoring data with OK or NG labels and monitoring data without labels to the autoencoder generated by the monitoring data using the OK label, and obtains the data output by the autoencoder. The information processing device 1 calculates the difference between the data input to the autoencoder and the data output by the autoencoder for each monitoring data. The calculated difference is data of the same dimension as the monitoring data. The information processing device 1 generates a dimensional compression model and performs visualization and other processing on the differential data calculated for each monitoring data.

[0121] Fig.16 1 is a flowchart showing an example of the steps of visualization processing using an autoencoder performed by the information processing device 1 according to Embodiment 1. The monitoring data acquisition unit 11b of the processing unit 11 of the information processing device 1 according to this embodiment reads a plurality of monitoring data stored in the monitoring data storage unit 12c of the storage unit 12 (step S81). In addition, the processing unit 11 reads the label information created in advance and stored in the storage unit 12 (step S82).

[0122] The processing unit 11 extracts the monitoring data to which a predetermined label (e.g., an OK label) is attached from the monitoring data read out in step S81 based on the label information read out in step S82 (step S83). The processing unit 11 performs machine learning processing using the monitoring data with the OK label extracted in step S83, thereby generating an autoencoder (step S84). The processing unit 11 inputs the monitoring data to the autoencoder generated in step S84 for all the monitoring data read out in step S81, calculates the difference between the monitoring data and the data output by the autoencoder, and thereby obtains the difference data (step S85).

[0123] Next, the dimensionality compression model generation unit 11c of the processing unit 11 generates a dimensionality compression model that compresses the number of dimensions of the differential data acquired in step S85 (step S86). The processing unit 11 stores the autoencoder generated in step S84 and the information related to the dimensionality compression model generated in step S86 in the dimensionality compression model information storage unit 12d of the storage unit 12 (step S87).

[0124] The data conversion unit 11d of the processing unit 11 inputs the differential data acquired in step S65 to the dimensional compression model generated in step S86, acquires the data output by the dimensional compression model, and performs data conversion of the differential data (step S88). The display processing unit 11e of the processing unit 11 generates a scatter plot that visualizes the differential data after the data conversion in step S88 (step S89). The display processing unit 11e displays the scatter plot generated in step S89 on the display unit 14 (step S90), and the processing ends.

[0125] Fig.17 : is a schematic diagram showing an example of visualization processing using an autoencoder. The illustrated example is an example of an information processing device 1 displaying a scatter plot, which is generated by using machine learning of monitoring data with an OK label, obtaining differential data of the input and output of the autoencoder for all monitoring data, compressing the differential data and plotting it on a two-dimensional plane. In addition, the information processing device 1 plots the differential data corresponding to the monitoring data with an OK label in the scatter plot with dark dots, and plots the differential data corresponding to the monitoring data with an NG label in the scatter plot with light dots, according to the flags attached to the monitoring data corresponding to each differential data. In the illustrated scatter plot, the monitoring data with an NG label are concentrated in the upper right area (refer to the area surrounded by the dotted line), and are separated from the monitoring data with an OK label other than this.

[0126] As described above, the information processing device 1 according to the present embodiment can present a large amount of multi-dimensional monitoring data using a two-dimensional or three-dimensional graph that can be easily understood by the user by performing the visualization processing of the monitoring data (1) to (5). As a result, the user can expect to verify important factors or causes such as malfunctions of the plurality of chambers 3 and individual differences based on the graph such as the scatter plot displayed on the display unit 14 of the information processing device 1.

[0127] In addition, the information processing device 1 involved in this embodiment does not need to perform all of the visualization processes (1) to (5) above, and only needs to perform at least one of the visualization processes (1) to (5). In addition, in this embodiment, an example of displaying a two-dimensional scatter plot is shown as a result of visualization of monitoring data, but it is not limited to this. The information processing device 1 can use charts other than scatter plots, such as histograms, box plots, violin plots, bee swarm plots, scatter plots, density contour maps, and heat maps to visualize monitoring data.

[0128] <Chamber verification process>

[0129] In the information processing system according to the present embodiment, the information processing device 1 stores and accumulates the monitoring data obtained from the plurality of chambers 3 in a database, converts the accumulated monitoring data into low-dimensional data, and displays the data in a graph such as a scatter plot. For example, in order to utilize the functions such as visualization of monitoring data of the information processing system according to the present embodiment, users such as managers or operators of semiconductor manufacturing plants need to set the information processing device 1 in advance with respect to information related to the fleet (plural chambers 3) that is set as the monitoring object in advance, the type of information collected as monitoring data, and the acquisition timing of the collected information. The information processing device 1 that receives these settings from the user acquires the necessary information from the chamber 3 and accumulates it in the database when performing various processes on the wafer based on the chamber 3.

[0130] After processing the chamber 3 for a predetermined period of time and accumulating a sufficient amount of monitoring data in the database, the user can cause the information processing device 1 to perform the above-mentioned visualization processing, and verify the important factors or causes such as the failure and individual differences of the multiple chambers 3 to be monitored based on the various graphs displayed on the display unit 14. For example, the user can determine the important factors such as the abnormality (which sensor detected the abnormality) for the chamber 3 where the abnormality occurred based on the displayed scatter plot or other graphs. In addition, for example, the user assigns certain labels to each monitoring data based on the displayed scatter plot or other graphs, and uses the monitoring data to which the labels are attached to perform so-called supervised machine learning processing, so that, for example, a learning model can be generated that infers the presence or absence of abnormalities or the degree of deterioration of the chamber 3 based on the input of the monitoring data. In addition, the user can verify, for example, which sensor is highly important for detecting the abnormality of the chamber 3 based on the generated learning model.

[0131] Fig.181 is a flowchart showing the steps of the chamber verification process performed in the information processing system according to the present embodiment. The processing unit 11 of the information processing device 1 according to the present embodiment receives the input of information from the user at the operation unit 15, thereby acquiring fleet composition information (step S101). The fleet composition information includes information such as the fleet ID added to the fleet, the number, type, and chamber ID of the chambers 3 included in the fleet, the number and type of sensors mounted in each chamber 3, and the content and setting value of the process performed in each chamber 3.

[0132] Next, the processing unit 11 receives the user's input of information on the operation unit 15, thereby acquiring data collection plan information (step S102). The data collection plan information is information for the information processing device 1 to set conditions for data to be collected as monitoring data, such as information specifying the recipe ID of the recipe to be collected, the recipe step ID of the recipe step, the type of statistical value, and the type of sensor.

[0133] Fig.19 as well as Fig. 20 is a diagram used to illustrate an example of a data collection plan. Fig.19 The upper part of shows an example of a graph of the change in the value of the output of sensor A mounted in chamber 3 over time when a predetermined process is performed on a wafer. The information processing device 1 can also, for example, cause chamber 3 to actually operate to obtain time series data of the output value of sensor A, and display the illustrated graph on the display unit 14 and prompt the user based on the obtained output value. In the present embodiment, a series of processes performed by chamber 3 on a wafer can be divided into a plurality of recipe steps. In addition, in the present example, for example, a recipe step with a recipe step ID of "1" is recorded by determining each recipe step according to the rule of "recipe step" + "#" + "recipe step ID" as in "recipe step #1". The recipe step is predetermined by a user such as a designer or manager of the system and can be specified in the recipe. In addition, when multiple sensors are mounted in chamber 3, time series data corresponding to the recipe step is obtained for each sensor in the same manner. Based on the time series data of the output value of sensor A obtained from chamber 3, the information processing device 1 calculates statistical values ​​such as the average, standard deviation, maximum and / or minimum of the output value of sensor A for each recipe step (refer to Fig.19 lower part).

[0134] The information processing device 1 acquires output values ​​of a plurality of sensors of a plurality of chambers 3, calculates statistical values ​​in the same manner, creates a histogram of the statistical values, and displays the histogram on the display unit 14 (see Fig. 20). In the illustrated example, for the seven sensors A to G provided in each chamber 3, the statistical values ​​(average values ​​in this example) of the output values ​​in the plurality of recipe steps #1 to #13 are displayed in a matrix in the form of a histogram arranged vertically and horizontally. Based on this display, the user can, for example, select a recipe step and a combination of sensors with large changes as the object of analysis, and can also, for example, exclude steps and sensor combinations that have no changes from the object of analysis. For example, the user performs an operation of selecting the displayed histogram on the operation unit 15, thereby performing an operation of selecting one or more recipe steps for each sensor. Each recipe step selected by the user becomes the object of data collection, and the information processing device 1 stores the statistical values ​​of the time series data of the selected recipe step of each sensor of each chamber 3 as monitoring data. In addition, in this example, the average value is used as the statistical value of each sensor, and the display can be changed to, for example, the standard deviation, the maximum or the minimum statistical value according to the user's selection. The user can select multiple recipe steps and multiple statistical values ​​for one sensor. In this embodiment, the information processing device 1 stores data of different types of recipe steps or statistical values ​​as other monitoring data in the monitoring data storage unit 12c even if it is the same sensor.

[0135] Next, the processing unit 11 receives the user's information input at the operation unit 15, thereby acquiring the analysis interval information (step S103). The analysis interval information is information related to the period (data range) that is the object of analysis. The processing unit 11 stores the fleet composition information acquired in step S101, the data collection plan information acquired in step S102, and the analysis interval information acquired in step S103 in the storage unit 12 (step S104). The control processing unit 11a of the processing unit 11 operates the chamber 3 and processes the wafer at a preset processing date and time, and the monitoring data acquisition unit 11b of the processing unit 11 acquires and accumulates the detection value of the sensor of the chamber 3 as monitoring data (step S105).

[0136] Next, the processing unit 11 performs a process of visualizing the monitoring data stored in the monitoring data storage unit 12c, for example, based on a request from a user (step S106). In step S106, at least one of the visualization processes based on the above-mentioned methods (1) to (5) is performed. Which visualization process is performed is based on, for example, a user's selection.

[0137] Next, the processing unit 11 receives the user's input of information at the operating unit 15, thereby performing a process of assigning labels to the monitoring data (step S107). For example, the processing unit 11 displays a scatter plot that visualizes the monitoring data on the display unit 14, receives the user's selection of one or more points from a plurality of points included in the scatter plot, and receives input of labels for the selected one or more points. The processing unit 11 stores the labels input by the user in correspondence with the monitoring data corresponding to the points displayed on the scatter plot. In addition, the labels assigned in this step S107 are used in the next step S108, and can also be used for Fig.11 The visualization shown processes the label information.

[0138] Next, the processing unit 11 uses the monitoring data to which the label is added in step S107 to perform so-called supervised machine learning processing, thereby receiving the monitoring data as input and performing processing to generate a prediction model that predicts the label of the monitoring data (step S108). The prediction model can adopt various learning models of various structures such as SVM (Support Vector Machine), random forest, logistic regression, linear discriminant analysis, gradient boosting or neural network. Supervised machine learning processing is a prior art, so detailed description is omitted.

[0139] Next, the processing unit 11 displays the sensor ranking related to the multiple sensors mounted in the chamber 3 on the display unit 14 based on the prediction model generated in step S108 (step S109), and ends the processing. The sensor ranking is information that arranges which sensors have high importance (contribution) to the prediction of the prediction model in order of importance for the multiple sensors mounted in the chamber 3.

[0140] Fig.21 : is a schematic diagram showing an example of sensor sorting. The illustrated sensor sorting is a horizontal bar graph showing the sorting of the importance of sensors included in the monitoring data, and is displayed in order from top to bottom starting from the sensor with the highest importance. In this example, each sensor that is the object of importance calculation is distinguished by a combination of sensor, period, and type of statistical value. Even if it is the same sensor, the importance is calculated as a different sensor when the period or statistical value is different. However, this is an example, and in the case where multiple combinations are included in the monitoring data for one sensor, for example, the maximum value or average value of the importance calculated for each sensor according to each combination of sensor, period, and type of statistical value can be obtained as a representative value, and the sensor sorting can be displayed based on the representative value so that one sensor represents one importance.

[0141] The user selects one or more sensors from the displayed sensor ranking, and the information processing device 1 accepts the user's selection. In the illustrated example, the four dark-colored bar graphs "Sensor A, period #8, average", "Sensor D, period #8, minimum", "Sensor C, period #8, average", and "Sensor C, period #8, minimum" are selected by the user. The information processing device 1 can display a trend graph of the detection values ​​of the time series for the sensor selected by the user, for example. In addition, in order to perform such a graph display, the information processing device 1 can store the data of the detection values ​​of the sensor's time series separately from the statistical values ​​of the monitoring data. In addition, the information processing device 1 can also display a graph such as a scatter plot after dimensional compression of the monitoring data for the sensor selected by the user.

[0142] In the case of using a learning model constructed by a decision tree such as a random forest or gradient boosting as a prediction model, the information processing device 1 can calculate the importance of the input information by investigating which input information is used by the branch conditions of the decision tree and the frequency of the branch conditions of the input information, etc. for the prediction model constructed by the decision tree. However, the structure of the prediction model is not limited to the decision tree, for example, it can also be a structure such as SVM (Support Vector Machine) or a neural network. In this case, for example, XAI (eXplainable Artificial Intelligence) technology such as LIME (Local Interpretable Model-agnostic Explanations) or SHAP (SHapley Additive exPlanations) can be used to calculate the importance of the input information. In addition, for example, the prediction model can also be a structure such as Transformer. In this case, for example, the attention mechanism can be used to calculate the importance of the input information. The calculation of the importance of the input information to the prediction model is a prior art, so the detailed description is omitted.

[0143] Summary

[0144] In the information processing system according to the present embodiment having the above structure, the information processing device 1 acquires monitoring data including a plurality of items related to the status of each process performed by the chamber (object device) 3. The information processing device 1 generates a dimensionality compression model for compressing the number of dimensions of the monitoring data, inputs the monitoring data into the dimensionality compression model, converts the monitoring data into a low-dimensional representation, and displays (outputs) the monitoring data converted into the low-dimensional representation on the display unit 14. Thus, the information processing system according to the present embodiment can convert multi-dimensional monitoring data into a low-dimensional representation and present it to the user, thereby assisting in the analysis of the monitoring data obtained from the chamber 3.

[0145] In addition, in the information processing system involved in the present embodiment, the monitoring data includes a plurality of items classified into a plurality of categories such as a pressure category and a temperature category. The information processing device 1 generates a dimensional compression model for each category included in the monitoring data, converts the monitoring data into a low-dimensional representation for each category, and displays it. Thus, the information processing system involved in the present embodiment can visualize the monitoring data for each category and prompt the user. Therefore, the information processing system can assist in verifying which category the important factors or causes such as the malfunction or individual difference of the chamber 3 are based on.

[0146] In addition, in the information processing system involved in the present embodiment, the information processing device 1 obtains monitoring data for multiple chambers 3. The information processing device 1 calculates the first distance of each monitoring data relative to the monitoring data group of multiple chambers 3 (fleet), and calculates the second distance of each monitoring data relative to the monitoring data group of each chamber 3. The information processing device 1 converts each monitoring data into a low-dimensional representation using the calculated first distance and second distance as dimensions and displays it. The first distance and the second distance can be, for example, Mahalanobis distance. Therefore, in the information processing system involved in the present embodiment, it is possible to assist the user in verifying whether the deviation of the monitoring data is in units of fleets or chambers.

[0147] In addition, in the information processing system involved in the present embodiment, a part of the monitoring data stored in the information processing device 1 is preliminarily assigned a label related to the class classification. The information processing device 1 obtains the label information attached to the monitoring data, selects a predetermined number of dimensions based on the label information for the monitoring data converted into a low-dimensional representation using the dimensionality compression model, and displays the monitoring data in the selected dimensions. Thus, in the information processing system involved in the present embodiment, it is possible to assist the user in verifying which dimension is effective for the class classification of the monitoring data.

[0148] In addition, in the information processing system involved in the present embodiment, the information processing device 1 obtains label information attached to a part of the monitoring data. The information processing device 1 determines the basis for scale conversion based on the monitoring data classified into the specified class using the label information, and scales the monitoring data of the specified class and the class other than the specified class based on the determined basis. The information processing device 1 generates a dimension compression model based on the scaled monitoring data, and displays the dimensionally compressed monitoring data. As a result, in the information processing system involved in the present embodiment, it is possible to display a scatter plot or the like that can separate the monitoring data of the specified class from the monitoring data other than the specified class.

[0149] In addition, in the information processing system involved in the present embodiment, the information processing device 1 obtains label information attached to a part of the monitoring data, and generates an autoencoder based on the monitoring data classified into a specified class using the label information. The information processing device 1 inputs the monitoring data of the specified class, the monitoring data of the class other than the specified class, and the monitoring data with the label attached into the autoencoder, and calculates the differential data of the input and output of the autoencoder. The information processing device 1 generates a dimensional compression model that compresses the number of dimensions of the differential data, and displays the differential data after dimensional compression. As a result, in the information processing system involved in the present embodiment, it is possible to display a scatter plot or the like that can separate the monitoring data of the specified class from the monitoring data outside the specified class.

[0150] In addition, in the information processing system according to the present embodiment, labels are assigned to a portion of the monitoring data converted into a low-dimensional representation, and based on the monitoring data to which the labels are assigned, the information processing device 1 generates a prediction model for predicting the input labels of the monitoring data. Thus, in the information processing system, for example, the monitoring data that can be acquired along with the subsequent processing in the chamber 3 and the generated prediction model can be used to predict the result of the processing performed by the chamber 3.

[0151] In addition, in the information processing system according to the present embodiment, the information processing device 1 calculates the importance of multiple items included in the monitoring data based on the generated prediction model, and displays the ranking of the multiple items based on the calculated importance. The information processing device 1 selects a predetermined number of items from the multiple items included in the monitoring data by accepting the user's selection based on the displayed importance, and visualizes the monitoring data or time series data with the selected items. As a result, the user can analyze the items with high importance first.

[0152] In addition, in the information processing system according to the present embodiment, the information processing device 1 obtains the detection value of the sensor of the chamber 3 as time series data, divides the processing performed by the chamber 3 into a plurality of periods, and calculates the statistical value of the time series data of each period. The information processing device 1 displays a graph such as a histogram of each period based on the calculated statistical value, and accepts the selection of a period as a monitoring target from a plurality of periods from the user. The information processing device 1 regards the received period as the monitoring target period of the chamber 3, calculates the statistical value for each monitoring target period for the value output by the sensor of the chamber 3, and stores the calculated statistical value as monitoring data. In this way, the information processing device 1 can select and store information useful for monitoring from a large amount of information that can be obtained from the sensor of the chamber 3.

[0153] In addition, in the information processing system involved in the present embodiment, it is preferred to perform monitoring and analysis based on the information processing device 1 for processing of substrates (wafers), substrate conveying processing, waferless dry cleaning process, aging process, plasma health check process, and automatic inspection process of sensors performed on the target device such as the chamber 3.

[0154] In the information processing system according to the present embodiment, when dimensional compression and output of monitoring data are performed, it is preferred that output be performed based on, for example, a scatter plot, a histogram, a box plot, a violin plot, a density contour map, and a heat map.

[0155] <Implementation method 2>

[0156] The information processing system according to the first embodiment described above is a system in which the information processing device 1 generates a dimensional compression model based on monitoring data acquired from a plurality of chambers 3, performs visualization of the monitoring data, and analyzes the bad conditions or individual differences of the chambers 3. In contrast, the information processing system according to the second embodiment is a system that uses the analysis results related to the existing chambers 3, for example, to analyze one or more newly added chambers 3. In addition, in the second embodiment, various information related to the analysis results of the existing chambers 3 is referred to as analysis information.

[0157] The analysis information used by the information processing system according to the second embodiment may include, for example, the following information collected or generated by the information processing system according to the first embodiment.

[0158] Monitoring data stored in the monitoring data storage unit 12c

[0159] ·exist Figure 4 The dimension compression model of each category stored in step S4

[0160] ·exist Figure 6The covariance matrix stored in step S30 is

[0161] · Fig.11 Label information shown

[0162] ·exist Fig.12 The dimension compression model stored in step S43

[0163] ·exist Fig.14 The scale conversion rules and dimension compression model stored in step S67

[0164] ·exist Fig.16 The autoencoder and dimension compression model stored in step S87

[0165] ·exist Fig.18 The fleet composition information and data collection plan information stored in step S104

[0166] In addition, the analysis information used by the information processing system according to the second embodiment may include the following information.

[0167] · Mahalanobis distance statistics for fleet-based surveillance data

[0168] Statistics of monitoring data of each sensor in fleet

[0169] ·Statistics of the Mahalanobis distance of monitoring data in chamber units

[0170] Statistics of monitoring data of each sensor in the chamber

[0171] However, the information processing device 1 can generate or calculate the dimension compression model and statistics based on the monitoring data. Therefore, the information of the dimension compression model and statistics may not be included in the parsing information, and the information processing device 1 generates or calculates the dimension compression model and statistics based on the monitoring data included in the parsing information. In addition, when the parsing information includes the dimension compression model and statistics, the monitoring data may not be included in the parsing information.

[0172] The monitoring data in fleet units are all the monitoring data obtained from all the chambers 3 included in the fleet (data of statistics collected based on the data collection plan). The information processing device 1 calculates the covariance matrix based on all the monitoring data of all the chambers 3 that are the objects, and calculates the Mahalanobis distance of each monitoring data based on the calculated covariance matrix. The statistic can take values ​​such as the mean, variance (standard deviation), skewness, or kurtosis. In the present embodiment, the average value is used as the statistic, but it is not limited to this, and the statistic can take various values. In addition, the Mahalanobis distance can also be a Mahalanobis distance calculated based on the monitoring data that has been dimensionally compressed by the dimensionality compression model. The information processing device 1 calculates the average value of the Mahalanobis distances calculated for each monitoring data, and stores the average value as the "statistic of the Mahalanobis distance of the monitoring data in fleet units" included in the analysis information.

[0173] The monitoring data of each sensor in fleet units is the monitoring data obtained by each sensor installed in all chambers 3 included in the fleet. For example, when the monitoring data includes detection values ​​of N sensors, the recipe steps are P steps, and Q types of values ​​are stored as statistical values, all monitoring data obtained from all chambers 3 are divided into (N×P×Q) monitoring data groups. The information processing device 1 includes the (N×P×Q) monitoring data groups as "statistics of all monitoring data obtained from all chambers 3" in the analysis information and stores them.

[0174] The monitoring data in chamber units are monitoring data acquired from each chamber 3, and when the fleet includes M chambers 3, there are M monitoring data groups. The information processing device 1 calculates a covariance matrix from the monitoring data for each chamber 3, and calculates the Mahalanobis distance of each monitoring data of the chamber 3 based on the calculated covariance matrix. The information processing device 1 includes the Mahalanobis distance calculated for each chamber 3 as a "statistic of the Mahalanobis distance of the monitoring data in chamber units" in the analysis information and stores it.

[0175] The monitoring data of each sensor in the chamber unit is the monitoring data obtained from each chamber 3 divided according to each sensor mounted on each chamber 3. For example, the fleet includes L chambers 3, the monitoring data includes detection values ​​of N sensors, the recipe step is P steps, and when Q types of values ​​are stored as statistical values, the monitoring data is divided into (L×N×P×Q) monitoring data groups. The information processing device 1 includes the (L×N×P×Q) monitoring data groups as "statistics of monitoring data of each sensor in the chamber unit" in the analysis information and stores them.

[0176] The “statistics of monitoring data of each sensor in a chamber” and the “statistics of all monitoring data acquired from all chambers 3” described above are information of substantially the same content, except for how to distinguish and process information acquired from all chambers 3. The information processing device 1 may store information acquired from all chambers 3 in advance, and use the information appropriately as the “statistics of monitoring data of each sensor in a chamber” or the “statistics of all monitoring data acquired from all chambers 3”.

[0177] In addition, it is preferable that the monitoring data used for calculating these statistics is the monitoring data obtained from the chamber 3 in which no failure occurs. Therefore, these statistics can be used as a reference for determining whether it is a normal chamber 3. In addition, the analysis information used by the information processing system involved in the second embodiment does not need to include all of the above information, and only needs to include at least one of the above multiple information.

[0178] Fig. 22 1 is a flowchart for explaining an overview of the analysis process performed by the information processing device 1 according to the second embodiment. The processing unit 11 of the information processing device 1 according to the second embodiment determines whether the analysis information related to the existing chamber 3 can be used, for example, based on whether the analysis information is stored in the storage unit 12 (step S201). If the existing analysis information cannot be used (S201: No), the processing unit 11 performs a normal analysis process without using the analysis information, for example, the analysis process performed in the first embodiment. Fig.18 The processing shown causes the processing to enter step S210.

[0179] If the existing analytical information can be used (S201: Yes), the processing unit 11 reads and acquires the analytical information from the storage unit 12 (step S203). The processing unit 11 receives input of information related to a newly added chamber 3 from a user, for example, and updates information on a plurality of chambers 3 to be monitored, that is, fleet configuration information (step S204). Then, in the processing unit 11, the processing unit 11a is controlled to operate the chamber 3 to process the wafer, and the monitoring data acquisition unit 11b acquires and stores the detection value of the sensor of the chamber 3 as monitoring data (step S205).

[0180] After accumulating the monitoring data, the processing unit 11 uses the existing dimensional compression model contained in the analysis information obtained in step S203 to perform data conversion, i.e., dimensional compression, on the monitoring data accumulated in step S205 (step S206). In addition, the processing unit 11 reads out the existing label information contained in the analysis information obtained in step S203 (step S207). The processing unit 11 performs visualization processing on the new monitoring data accumulated in step S205 based on the result of the data conversion in step S206 and the label information read out in step S207 (step S208). In addition, in step S208, at least one of the visualization processing described in Implementation Example 1 (1) visualization processing for each category, (2) visualization processing of deviations in fleet units and in chamber units, (3) visualization processing based on label information, (4) visualization processing using scale conversion based on label information, and (5) visualization processing using an autoencoder is performed.

[0181] The processing unit 11 uses the new monitoring data accumulated in step S205 to update the existing dimensional compression model included in the analysis information acquired in step S203 (step S209), and the process proceeds to step S210. For example, the processing unit 11 generates a dimensional compression model using the existing monitoring data related to the analyzed chamber 3 included in the analysis information and the new monitoring data related to the added chamber 3, thereby updating the dimensional compression model.

[0182] The processing unit 11 generates analysis information including monitoring data, dimensionality compression model, label information, covariance matrix, statistics, etc. based on the results of the analysis processing of step S202 or the results of the analysis processing of steps S203 to S209 (step S210). The processing unit 11 stores the analysis information generated in step S210 in the storage unit 12 (step S211) and ends the processing.

[0183] In addition, the information processing device 1 according to the second embodiment performs monitoring processing of the time-dependent changes of the plurality of chambers 3 based on various statistics stored as analytical information. The information processing device 1 calculates the above-mentioned statistics (statistics of Mahalanobis distances of monitoring data in fleet units, statistics of monitoring data of each sensor in fleet units, statistics of Mahalanobis distances of monitoring data in chamber units, and statistics of monitoring data of each sensor in chamber units) based on the monitoring data acquired from the plurality of chambers 3 at a predetermined period such as once a day or once every 10 hours of operation. The information processing device 1 stores these repeatedly calculated statistics as analytical information, and compares the latest statistics with the previous statistics every time the latest statistics are calculated to determine the presence or absence of abnormality related to the time-dependent changes of the chambers 3. The information processing device 1 can determine the presence or absence of abnormality based on, for example, whether the deviation of the latest statistics from the previous statistics is within a predetermined range (within 3σ, etc.), or whether the difference between the latest statistics and the previous statistics exceeds a threshold value.

[0184] Fig.23 The flowchart shows the steps of the monitoring process of the time change performed by the information processing device 1 involved in the second embodiment. The processing unit 11 of the information processing device 1 involved in the second embodiment obtains by reading the analysis information stored in the storage unit 12 (step S221). The processing unit 11 obtains the monitoring data stored in the monitoring data storage unit 12c (step S222). The processing unit 11 calculates at least one of the statistics of the Mahalanobis distance of the monitoring data in fleet units, the statistics of the monitoring data of each sensor in fleet units, the statistics of the Mahalanobis distance of the monitoring data in chamber units, or the statistics of the monitoring data of each sensor in chamber units based on the monitoring data read in step S222 (step S223).

[0185] The processing unit 11 compares the past statistics included in the analysis information acquired in step S221 with the latest statistics calculated in step S223 to determine whether the deviation of the latest statistics relative to the past statistics is within a normal range (step S224). If it is within the normal range (S224: Yes), the processing unit 11 includes the statistics calculated in step S223 in the analysis information and stores it (step S225), and ends the processing.

[0186] When the deviation of the latest statistic from the past statistic is not within the normal range (S224: No), the processing unit 11 performs, for example, Fig. 22 The analysis process shown (step S226).

[0187] Based on the analytical processing, the processing unit 11 performs various processes for correcting the bad situation etc. generated in the chamber 3 (step S227). The processes for correcting the bad situation etc. may include, for example, detailed processing of the input and output data of the sensor determined as an important factor of the bad situation confirmed by the analytical processing, calibration processing of the sensor, replacement processing of the sensor or operation confirmation processing of the replaced sensor, etc. The processing unit 11 determines whether the state of the chamber 3 has returned to normal through the processing of step S227 (step S228). In addition, in order to perform the determination of step S228, in the information processing device 1, for example, it is possible to perform processing such as processing the wafer based on the chamber 3 and collecting monitoring data.

[0188] If the state of the chamber 3 has not returned to normal (S228: No), the processing unit 11 performs appropriate individual responses such as stopping the operation of the abnormal chamber 3 or requesting maintenance of the abnormal chamber 3 (step S229), and ends the processing. In addition, the various processes shown in the above-mentioned steps S227 to S229 may not necessarily be performed by the information processing device 1 alone, and may be performed in cooperation with a user such as an administrator of the information processing system involved in this embodiment. In this case, the processing unit 11 receives the user's operation and appropriately performs processing for correcting the malfunction, determining whether the chamber 3 has returned to normal, or performing individual corresponding processing for the abnormal chamber 3.

[0189] When the state of the chamber 3 returns to normal ( S228 : Yes), the processing unit 11 ends the processing.

[0190] Fig.24 Schematic diagram for explaining monitoring processing that changes over time. This figure is a graph showing the change over time of a statistic related to a sensor of the chamber 3, with the horizontal axis being the date and time and the vertical axis being the statistic. The information processing device 1 can calculate the statistic of the monitoring data obtained from the chamber 3 at a predetermined period, such as once a day or once every 10 hours of operation, store the calculated statistic as analysis information, and display the graph as shown in the figure on the display unit 14. Each point depicted in the graph of this example represents a statistic obtained from the sensor as a result of processing a wafer.

[0191] The output value of the sensor in this example tends to increase as the processing of chamber 3 is carried out, but returns to the initial state through maintenance. Fig.24In the graph shown, in the area surrounded by the dotted line, there is a chamber 3 where the output value of the sensor increases greatly with time. By periodically calculating the statistics of the monitoring data and storing them as analytical information, and regularly checking the changes or deviations of the statistics, abnormalities of the chamber 3 can be discovered. In addition, in this example, repairs and maintenance are carried out with the discovery of abnormalities. Through these measures, the output value of the sensor returns to the initial state, and then the figure shows a situation where no abnormality occurs.

[0192] In the information processing system involved in the second embodiment of the above structure, the information processing device 1 obtains monitoring data including multiple items related to the status of each process performed by the chamber (object device) 3. The information processing device 1 obtains the monitoring data related to the chamber 3 that has been analyzed or the dimensionally compressed analysis information generated based on the monitoring data. The information processing device 1 inputs the monitoring data of the chamber 3 of the object into the dimension compression model generated based on the monitoring data related to the analyzed device, and converts it into a low-dimensional representation. The information processing device 1 displays (outputs) the monitoring data converted into the low-dimensional representation on the display unit 14. Thus, the information processing system involved in the present embodiment can utilize the dimensional compression model generated based on the monitoring data of the chamber 3 that has been analyzed, for example, to convert the monitoring data of the newly established chamber 3 of the object into a low-dimensional representation and prompt it to the user. The information processing system can assist in the analysis of the monitoring data obtained from the chamber 3.

[0193] In addition, in the information processing system according to the second embodiment, the information processing device 1 determines whether there is analytical information, and when there is analytical information, the dimensionality of the monitoring data is compressed using the analytical information. When there is no analytical information, the information processing device 1 generates a dimensionality compression model based on the acquired monitoring data of the target chamber 3, and performs dimensionality compression of the monitoring data. Thus, the information processing system according to the present embodiment can assist in the analysis of the monitoring data obtained from the target chamber 3 regardless of the presence or absence of analytical information related to the analyzed chamber 3.

[0194] In addition, in the information processing system according to the second embodiment, the information processing device 1 stores the monitoring data related to the analyzed chamber 3 and the dimensional compression model generated based on the monitoring data as the analysis information. Thus, the information processing device 1 can perform dimensional compression of the monitoring data of the target chamber 3 using the stored dimensional compression model without regenerating the dimensional compression model. In addition, the information processing device 1 can regenerate the dimensional compression model using the stored monitoring data related to the analyzed chamber 3 and the newly acquired monitoring data of the target chamber 3.

[0195] In addition, in the information processing system involved in the second embodiment, the monitoring data includes a plurality of items classified into a plurality of categories such as a pressure category and a temperature category. The information processing device 1 stores a dimensional compression model for each category as analytical information, and uses the stored dimensional compression models to convert the monitoring data into a low-dimensional representation for each category and display it. Thus, the information processing system involved in the second embodiment can visualize the monitoring data for each category and present it to the user. The information processing system can, for example, assist in verifying which category the important factors or causes such as the malfunction or individual differences of the chamber 3 are based on.

[0196] In addition, Figure 3 In the embodiment, "pressure category" and "temperature category" are described as categories, but this is an example, and the categories can be classified in any form. For example, multiple categories such as pressure and temperature can be combined into one category, such as "pressure and temperature category". In addition, for example, the same category can be further subdivided into multiple categories, such as "pressure category 1" and "pressure category 2". This is suitable for both the information processing systems of embodiment 1 and embodiment 2.

[0197] In addition, in the information processing system according to the second embodiment, various statistics of monitoring data related to the analyzed chamber 3 are included in the analysis information. The information processing device 1 calculates the statistics of monitoring data related to the chamber 3 to be analyzed. The information processing device 1 determines the presence or absence of abnormality or the state of change over time of the chamber 3 to be analyzed based on the comparison of the statistics of monitoring data related to the analyzed chamber 3 and the statistics of monitoring data related to the chamber 3 to be analyzed. Thus, the information processing system according to the second embodiment can assist in the analysis of the presence or absence of abnormality or the state of change over time of the chamber 3 to be analyzed based on the statistics related to the analyzed chamber 3.

[0198] In addition, the other structures of the information processing system involved in Embodiment 2 are the same as those of the information processing system involved in Embodiment 1, and therefore the same reference numerals are attached to the same parts, and the detailed description is omitted.

[0199] The embodiments disclosed herein are illustrative in all aspects and should not be construed as restrictive. The scope of the present disclosure is indicated by the claims rather than the above meaning, and is intended to include all modifications within the meaning and scope equivalent to the claims.

[0200] The matters described in each embodiment can be combined with each other. In addition, the independent claims and dependent claims described in the claims can be combined with each other in all combinations regardless of the reference form. Moreover, the claims use a form of recording claims that reference two or more other claims (multiple claim form), but are not limited to this. It is also possible to record multiple claims that reference at least one multiple claim (multiple-reference-multiple claims).

[0201] Description of Reference Numerals

[0202] 1…information processing device (computer); 3…chamber (object device); 11…processing unit; 11a…control processing unit; 11b…monitoring data acquisition unit; 11c…dimensional compression model generation unit; 11d…data conversion unit; 11e…display processing unit; 12…storage unit; 12a…program (computer program); 12b…setting information storage unit; 12c…monitoring data storage unit; 12d…dimensional compression model information storage unit; 13…communication unit; 14…display unit; 15…operation unit.

Claims

1. An information processing method, The information processing device acquires monitoring data including a plurality of items related to the status of each process performed by the target device. Generate a dimensional compression model for compressing the number of dimensions of the monitoring data. The above monitoring data is input into the above dimension compression model and converted into a low-dimensional representation. Output the above monitoring data converted into a low-dimensional representation.

2. The information processing method according to claim 1, wherein: The monitoring data includes a plurality of items classified into a plurality of categories. The information processing device generates the dimensional compression model for each of the categories. The monitoring data is input into a dimension compression model corresponding to each of the categories, and converted into a low-dimensional representation of each of the categories.

3. The information processing method according to claim 1, wherein: The information processing device acquires the monitoring data for a plurality of target devices. calculating a first distance of each monitoring data relative to the monitoring data group of the plurality of target devices, calculating a second distance of each monitoring data relative to the monitoring data group of each target device, The monitoring data is converted into a low-dimensional representation with the first distance and the second distance as dimensions.

4. The information processing method according to claim 3, wherein: The information processing device calculates a first covariance matrix associated with the monitoring data group of the plurality of target devices. Calculating the Mahalanobis distance of each monitoring data based on the first covariance matrix as the first distance, calculating a second covariance matrix associated with the monitoring data set of each target device, The Mahalanobis distance of each monitoring data based on the second covariance matrix is ​​calculated as the second distance.

5. The information processing method according to claim 3 or 4, wherein: Based on the monitoring data converted into a low-dimensional representation, one or more monitoring data groups are selected, For each selected monitoring data group, a dimensional compression model is generated to compress the number of dimensions of the monitoring data included in the monitoring data group. The above monitoring data is input into the above dimension compression model and converted into a low-dimensional representation. The monitoring data converted into a low-dimensional representation is output for each monitoring data group.

6. The information processing method according to claim 1, wherein: The information processing device acquires monitoring data for a plurality of target devices. converting the monitoring data about the plurality of target devices into a low-dimensional representation, Based on the monitoring data converted into a low-dimensional representation, one or more monitoring data groups are selected, Convert the monitoring data into a low-dimensional representation according to each selected monitoring data group, The monitoring data of each of the above monitoring data groups converted into a low-dimensional representation is output.

7. The information processing method according to claim 1, wherein: The information processing device obtains first label information related to the class classification added to the monitoring data, For the monitoring data converted into a low-dimensional representation by the dimensionality compression model, a predetermined number of dimensions are selected based on the first label information. The above monitoring data of the selected dimension is output.

8. The information processing method according to claim 1, wherein: The information processing device obtains first label information related to the class classification added to the monitoring data, determining a criterion for scaling based on the monitoring data classified into a predetermined category using the first label information, Based on the determined criteria, the monitoring data of the predetermined category and the categories other than the predetermined category are converted into a scale. The dimensional compression model is generated based on the scale-converted monitoring data.

9. The information processing method according to claim 1, wherein: The information processing device obtains first label information related to the class classification added to the monitoring data, generating an autoencoder based on the monitoring data classified into a predetermined class using the first label information, calculating a difference between an input and an output of the autoencoder when monitoring data of the predetermined class and monitoring data of a class other than the predetermined class are input to the autoencoder, The above-mentioned dimensional compression model is generated by compressing the number of dimensions of the calculated difference.

10. The information processing method according to claim 1, wherein: The information processing device assigns second label information to the monitoring data represented by the low-dimensional representation of the data output, Based on the monitoring data to which the second label information is assigned, a prediction model for predicting the second label information with respect to an input of the monitoring data is generated.

11. The information processing method according to claim 10, wherein: The information processing device calculates the importance of a plurality of items included in the monitoring data based on the generated prediction model, For the monitoring data, a predetermined number of items are selected based on the importance. The above monitoring data of the selected item is output.

12. The information processing method according to claim 1, wherein: The information processing device acquires time series data from the target device. The processing executed by the target device is divided into a plurality of periods, and a statistical value of each period of the acquired time series data is calculated. Accepts the selection of the period based on the above statistical values ​​calculated, The statistical values ​​during the period in which the selection is received are stored as the monitoring data.

13. The information processing method according to claim 10, wherein: The process performed by the target device includes any one of a substrate process, a substrate conveyance process, a waferless dry cleaning process, an aging process, a plasma health check process, and a sensor automatic check process.

14. The information processing method according to claim 1, wherein: The information processing device outputs a histogram, a box plot, a violin plot, a scatter plot, a density contour map, and a heat map based on the monitoring data converted into a low-dimensional representation.

15. An information processing method, The information processing device acquires monitoring data including a plurality of items related to the status of each process performed by the device to be analyzed, Acquiring analysis information including monitoring data related to the analyzed device or a dimensional compression model generated based on the monitoring data, The acquired monitoring data of the device to be analyzed is input into a dimensional compression model generated based on the acquired monitoring data included in the analysis information or a dimensional compression model included in the analysis information to convert the acquired monitoring data into a low-dimensional representation. Output the above monitoring data converted into a low-dimensional representation.

16. The information processing method according to claim 15, wherein: Determine whether the above analysis information exists. If it is determined that the analysis information does not exist, a dimensional compression model is generated to compress the number of dimensions of the monitoring data. The above monitoring data is input into the above dimension compression model and converted into a low-dimensional representation. Output the above monitoring data converted into a low-dimensional representation.

17. The information processing method according to claim 16, wherein: The monitoring data and the dimensional compression model generated based on the monitoring data are stored as the analysis information.

18. The information processing method according to claim 17, wherein: The monitoring data includes a plurality of items classified into a plurality of categories. Generate the above dimensional compression model according to each of the above categories, The monitoring data is input into a dimension compression model corresponding to each of the categories, and converted into a low-dimensional representation of each of the categories.

19. The information processing method according to claim 15, wherein: The analysis information includes statistics of monitoring data related to the analyzed device. Calculating statistics of monitoring data of the device to be analyzed, Based on the comparison between the statistic of the analyzed device and the statistic of the analyzed device, the state of the analyzed device is determined.

20. The information processing method according to claim 19, wherein: Based on the monitoring data of the plurality of devices, statistics related to the plurality of devices are calculated, Based on the monitoring data of each device, statistics related to each device are calculated.

21. The information processing method according to claim 19 or 20, wherein: The device is provided with a plurality of sensors for acquiring the monitoring data. Based on the monitoring data of the multiple sensors, statistics related to the multiple sensors are calculated. Based on the monitoring data of each sensor, statistics related to each sensor are calculated.

22. The information processing method according to claim 21, wherein: Calculating the distance of each monitoring data relative to the monitoring data group of the plurality of sensors, Calculate the above statistics related to the calculated distances.

23. The information processing method according to claim 15, wherein: The monitoring data of the device to be analyzed converted into a low-dimensional representation and the monitoring data of the device after analysis converted into a low-dimensional representation are simultaneously output.

24. A computer program that causes a computer to perform a process: acquiring monitoring data including a plurality of items related to the status of each process performed by the target device, Generate a dimensional compression model for compressing the number of dimensions of the monitoring data. The above monitoring data is input into the above dimension compression model and converted into a low-dimensional representation. Output the above monitoring data converted into a low-dimensional representation.

25. A computer program that causes a computer to perform the process: Acquiring monitoring data, the monitoring data including a plurality of items related to the status of each process performed by the device to be analyzed, Acquiring analysis information including monitoring data related to the analyzed device or a dimensional compression model generated based on the monitoring data, The acquired monitoring data of the device to be analyzed is input into a dimensional compression model generated based on the acquired monitoring data included in the analysis information or a dimensional compression model included in the analysis information to convert the acquired monitoring data into a low-dimensional representation. Output the above monitoring data converted into a low-dimensional representation.

26. An information processing device comprising: an acquisition unit that acquires monitoring data, the monitoring data including a plurality of items related to states of respective processes performed by the target device; A generating unit, generating a dimensional compression model for compressing the number of dimensions of the monitoring data; a conversion unit, inputting the monitoring data into the dimensionality compression model to convert the monitoring data into a low-dimensional representation; and The output unit outputs the monitoring data converted into a low-dimensional representation.

27. An information processing device comprising: a monitoring data acquisition unit that acquires monitoring data, the monitoring data including a plurality of items related to the status of each process performed by the device to be analyzed; An analysis information acquisition unit that acquires analysis information including monitoring data related to the analyzed device or a dimensional compression model generated based on the monitoring data; a conversion unit, inputting the acquired monitoring data of the device of the analysis object into a dimensional compression model generated based on the acquired monitoring data included in the analysis information or the dimensional compression model included in the analysis information, and converting the acquired monitoring data into a low-dimensional representation; as well as The output unit outputs the monitoring data converted into a low-dimensional representation.

Images