Information processing device, program, and control method

Abstract

To provide a technique for reducing an influence on operation due to a failure of a plurality of physical sensors installed on a substrate processing device.SOLUTION: An information processing device comprises: an acquisition unit which acquires physical sensor output values output from a plurality of physical sensors installed in a substrate processing device; a prediction unit which switches a physical mode constructed by a statistical model or a physical technique and predicts virtual sensor output values of a virtual sensor corresponding to the physical sensor to be predicted on the basis of the similarity between the physical sensor output values and the data which has been learned by the statistical model constructed by the statistical method; an abnormality determination unit which determines a physical sensor abnormality by comparing the physical sensor output values of the physical sensor to be predicted with the virtual sensor output values of the virtual sensor; and an instruction unit which switches control based on the physical sensor output value of the physical sensor to be predicted to control by the virtual sensor output value of the virtual sensor when the abnormality determination unit determines that the physical sensor abnormality has occurred.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present disclosure relates to an information processing apparatus, a program, and a control method.

Background Art

[0002] For example, a substrate processing apparatus is equipped with a plurality of physical sensors. There are cases where physical sensors mounted on a substrate processing apparatus cause problems in the operation of the substrate processing apparatus due to sudden failures. When detecting abnormalities in the physical sensors mounted on the substrate processing apparatus, the normal range (bandwidth) of the physical sensor output values is set and compared to detect abnormalities. Also, in semiconductor manufacturing apparatuses such as heat treatment film forming apparatuses equipped with a plurality of physical sensors, a technique for performing pre-failure detection by calculating virtual sensor data output from virtual sensors based on a physical model has been conventionally known (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The present disclosure provides a technique for reducing the impact on operation due to failures of a plurality of physical sensors installed in a substrate processing apparatus.

Means for Solving the Problems

[0005] One aspect of the present disclosure is an information processing device comprising: an acquisition unit configured to acquire physical sensor output values ​​output from a plurality of physical sensors installed in a substrate processing device; a prediction unit configured to predict the virtual sensor output value of a virtual sensor corresponding to a physical sensor to be predicted by switching between the statistical model or a physical model constructed by a physical method based on the similarity between the physical sensor output value and data that a statistical model constructed by a statistical method has learned; an abnormality determination unit configured to determine an abnormality of the physical sensor by comparing the physical sensor output value of the physical sensor to be predicted with the virtual sensor output value of the virtual sensor; and an instruction unit configured to switch from control based on the physical sensor output value of the physical sensor to be predicted to control based on the virtual sensor output value of the virtual sensor when it is determined that an abnormality has occurred in the physical sensor. [Effects of the Invention]

[0006] According to this disclosure, a technology can be provided to mitigate the impact on operations caused by failures of multiple physical sensors installed in a substrate processing device. [Brief explanation of the drawing]

[0007] [Figure 1] This is a configuration diagram of an example of a substrate processing system according to this embodiment. [Figure 2] This is a hardware configuration diagram of an example computer. [Figure 3] This is a functional block diagram of an example of a substrate processing apparatus according to this embodiment. [Figure 4] This is a functional block diagram of an example of the prediction unit according to this embodiment. [Figure 5] This flowchart shows an example of the procedure for creating a physical model. [Figure 6] This flowchart shows an example of the procedure for creating a statistical model. [Figure 7] This flowchart shows an example of the procedure for detecting anomalies and ensuring redundancy in the physical sensors targeted for prediction. [Figure 8]This is an illustrative diagram of an example of a process for detecting anomalies in the physical sensors being predicted. [Figure 9] This flowchart shows an example of the procedure for additional training a statistical model. [Figure 10] This is a functional block diagram of an example of a substrate processing device that controls the opening degree of an automatic pressure control device based on the physical sensor output value of a pressure sensor. [Figure 11] This is a functional block diagram of an example of a circuit board processing device that controls the opening degree of a valve based on the physical sensor output value of a flow sensor. [Modes for carrying out the invention]

[0008] The embodiments for carrying out the present invention will be described below with reference to the drawings.

[0009] <System Configuration> Figure 1 is a configuration diagram of an example of a substrate processing system according to this embodiment. The substrate processing system 1 in Figure 1 has one or more substrate processing devices 10, a device controller 12, a server device 14, and an operator terminal 16. The substrate processing device 10 and the device controller 12 are installed in the manufacturing plant 2. The server device 14 and the operator terminal 16 may be installed in the manufacturing plant 2 or elsewhere. The operator terminal 16 is operated by an operator such as a device operator or analysis operator of the substrate processing device 10 installed in the manufacturing plant 2.

[0010] The substrate processing apparatus 10, the apparatus controller 12, the server apparatus 14, and the worker terminal 16 are connected to each other via networks 18 and 20, such as the Internet and LAN (Local Area Network).

[0011] The substrate processing apparatus 10 is an apparatus that performs substrate processing such as film deposition, etching, and ashing, and is an apparatus for processing semiconductor wafers or an apparatus for processing glass substrates for flat panel displays. The substrate processing apparatus 10 is, for example, a semiconductor manufacturing apparatus, a heat treatment apparatus, or a film deposition apparatus.

[0012] The substrate processing apparatus 10 receives control commands according to a recipe or a macro from the apparatus controller 12 and executes the processing of the recipe or the macro. The substrate processing apparatus 10 executes processing according to a sequence such as a recipe or a macro. A plurality of physical sensors are installed in the substrate processing apparatus 10.

[0013] Also, as shown in FIG. 1, the substrate processing apparatus 10 may be equipped with the apparatus controller 12, or it does not necessarily have to be equipped as long as it is communicably connected. The apparatus controller 12 has a computer that controls the substrate processing apparatus 10. The apparatus controller 12 outputs a control command for controlling the control components of the substrate processing apparatus 10 according to a recipe or a macro, thereby causing the substrate processing apparatus 10 to execute processing according to the recipe or the macro.

[0014] Note that the apparatus controller 12 has the function of a man-machine interface that receives instructions for the substrate processing apparatus 10 from an operator and provides information about the substrate processing apparatus 10 to the operator.

[0015] The apparatus controller 12 receives the physical sensor output values output from a plurality of physical sensors installed in the substrate processing apparatus 10. The physical sensors are temperature sensors, pressure sensors, flow rate sensors, and the like. The apparatus controller 12 compares the physical sensor output value of the prediction target with the virtual sensor output value predicted by a physical model or a statistical model described later, and determines an abnormality of the physical sensor of the prediction target.

[0016] Also, when the apparatus controller 12 determines that an abnormality has occurred in the physical sensor of the prediction target, it switches from the control of the substrate processing apparatus 10 based on the physical sensor output value of the prediction target to the control by the virtual sensor output value of the virtual sensor corresponding to the physical sensor of the prediction target.

[0017] The server device 14 may receive and store the physical sensor output values output from a plurality of physical sensors installed in the substrate processing device 10. Similar to the device controller 12, the server device 14 may compare the physical sensor output value to be predicted with the virtual sensor output value predicted by the physical model or statistical model described later, and determine an abnormality in the physical sensor to be predicted. When it is determined that an abnormality has occurred in the physical sensor to be predicted, the server device 14 may switch from the control of the substrate processing device 10 based on the physical sensor output value to be predicted to the control by the virtual sensor output value of the virtual sensor corresponding to the physical sensor to be predicted.

[0018] Also, when it is determined that an abnormality has occurred in the physical sensor to be predicted, the device controller 12 or the server device 14 may display the abnormality of the physical sensor on, for example, the display of the device controller 12, the display of the server device 14, or the display of the operator terminal 16, so as to notify the operator of the abnormality of the physical sensor. When it is determined that an abnormality has occurred in the physical sensor to be predicted, the device controller 12 or the server device 14 may notify the operator of the abnormality of the physical sensor using an e-mail or the like. As will be described later, the substrate processing system 1 according to the present embodiment realizes the abnormality detection of the physical sensor to be predicted and the redundancy of the physical sensor to be predicted.

[0019] The operator terminal 16 is a PC (Personal Computer) or a smartphone operated by an operator such as a device person in charge or an analysis person in charge of the substrate processing device 10 installed in the manufacturing factory 2.

[0020] Note that the substrate processing system 1 in FIG. 1 is an example, and it is needless to say that there are various system configuration examples according to the application and purpose. For example, the substrate processing system 1 can have various configurations such as a configuration in which the device controllers 12 of the respective substrate processing devices 10 are integrated into a device controller for a plurality of substrate processing devices 10, or a further divided configuration.

[0021] <Hardware Configuration> The device controller 12, server device 14, and worker terminal 16 of the substrate processing system 1 shown in Figure 1 are implemented by a computer (information processing device) with the hardware configuration shown in Figure 2, for example. Figure 2 is a hardware configuration diagram of an example computer.

[0022] The computer 500 in Figure 2 includes an input device 501, an output device 502, an external interface 503, RAM (Random Access Memory) 504, ROM (Read Only Memory) 505, a CPU (Central Processing Unit) 506, a communication interface 507, and an HDD (Hard Disk Drive) 508, all of which are interconnected via bus B. The input device 501 and output device 502 may be connected and used only when necessary.

[0023] The input device 501 is a keyboard, mouse, touch panel, etc., used by the operator to input various operation signals. The output device 502 is a display, etc., which displays the processing results from the computer 500. The communication I / F 507 is an interface that connects the computer 500 to the network 18 or 20. The HDD 508 is an example of a non-volatile storage device that stores programs and data.

[0024] External I / F 503 is an interface to external devices. Computer 500 can read from and / or write to recording media 503a such as an SD (Secure Digital) memory card via External I / F 503. ROM 505 is an example of a non-volatile semiconductor memory (storage device) that stores programs and data. RAM 504 is an example of a volatile semiconductor memory (storage device) that temporarily holds programs and data.

[0025] The CPU 506 is a processing unit that reads programs and data from storage devices such as the ROM 505 and HDD 508 onto the RAM 504 and executes processing, thereby realizing the overall control and functions of the computer 500.

[0026] The device controller 12, server device 14, and worker terminal 16 in Figure 1 can perform various functions described later by executing programs on the computer 500 with the hardware configuration shown in Figure 2.

[0027] <Functional Configuration> The substrate processing apparatus 10 of the substrate processing system 1 according to this embodiment is implemented using functional blocks as shown in Figure 3, for example. Figure 3 is a functional block diagram of an example of the substrate processing apparatus according to this embodiment. Note that the functional block diagram in Figure 3 omits the illustration of components that are not necessary for the explanation of this embodiment.

[0028] The device controller 12 of the substrate processing apparatus 10 in Figure 3 implements the acquisition unit 40, prediction unit 42, abnormality determination unit 44, notification unit 46, and instruction unit 48 by executing a program for the device controller 12.

[0029] The physical sensor 30a to be predicted is permanently installed in the substrate processing device 10 and is a physical sensor that is targeted for abnormality detection. Permanently installed means that it is not a trial installation, but is installed, for example, in a product. The other physical sensors 30 are permanently installed in the substrate processing device 10 and are used to predict the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted. The other physical sensors 30 are preferably physical sensors 30 whose correlation with the physical sensor output value of the physical sensor 30a to be predicted is higher than a threshold.

[0030] Furthermore, the physical sensor 30a being predicted may be used as another physical sensor 30 of another physical sensor 30a being predicted. The other physical sensor 30 may be used as another physical sensor 30a being predicted.

[0031] The acquisition unit 40 of the device controller 12 acquires physical sensor output values ​​output from other physical sensors 30 and the physical sensor 30a to be predicted. The prediction unit 42 has the function of predicting the physical sensor output value of the physical sensor 30a to be predicted from the physical sensor output values ​​of the other physical sensors 30. In this embodiment, the physical sensor output value of the physical sensor 30a to be predicted predicted by the prediction unit 42 is called the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted. Details of the processing of the prediction unit 42 will be described later.

[0032] The abnormality determination unit 44 compares the physical sensor output value of the physical sensor 30a to be predicted with the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted, and determines whether or not an abnormality has occurred in the physical sensor 30a to be predicted. For example, if the difference between the physical sensor output value of the physical sensor 30a to be predicted and the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted is greater than or equal to a predetermined value, the abnormality determination unit 44 determines that an abnormality has occurred in the physical sensor 30a to be predicted. The abnormality determination unit 44 transmits information on whether or not an abnormality has occurred in the physical sensor 30a to be predicted to the notification unit 46 and the instruction unit 48.

[0033] When the notification unit 46 receives information from the abnormality determination unit 44 indicating that an abnormality has occurred in the physical sensor 30a to be predicted, it issues a warning or triggers an alarm to notify the operator of the abnormality in the physical sensor 30a to be predicted. Notification to the operator may be done by displaying it on a display, sending an email, turning on a light, emitting a sound, or printing it out.

[0034] The instruction unit 48 receives information from the abnormality determination unit 44 regarding whether or not an abnormality has occurred in the physical sensor 30a to be predicted. If no information indicating an abnormality has occurred in the physical sensor 30a to be predicted is received, the instruction unit 48 controls the control target 32 ​​according to a recipe or macro based on the physical sensor output value of the physical sensor 30a to be predicted.

[0035] When the instruction unit 48 receives information indicating that an abnormality has occurred in the physical sensor 30a being predicted, it switches the control of the controlled object 32 from control based on the physical sensor output value of the physical sensor 30a being predicted to control control based on the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a being predicted. When the instruction unit 48 receives information indicating that an abnormality has occurred in the physical sensor 30a being predicted, it controls the controlled object 32 according to a recipe or macro based on the virtual sensor output value of the virtual sensor.

[0036] Thus, if there is an abnormality in the physical sensor 30a to be predicted, the substrate processing apparatus 10 according to this embodiment can perform redundant operation by replacing it with the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted.

[0037] Therefore, according to the substrate processing apparatus 10 of this embodiment, even if the physical sensor 30a to be predicted fails during the process, the control target 32 ​​can be controlled to prevent, for example, the semiconductor wafer from being lost by substituting it with a virtual sensor corresponding to the physical sensor 30a to be predicted. Furthermore, according to the substrate processing apparatus 10 of this embodiment, even if the physical sensor 30a to be predicted fails during the process, it can be substituted with a virtual sensor corresponding to the physical sensor 30a to be predicted until the physical sensor 30a to be predicted can be replaced, thereby improving the operating rate of the substrate processing apparatus 10.

[0038] The prediction unit 42 is implemented using a functional block as shown in Figure 4, for example. Figure 4 is a functional block diagram of an example of the prediction unit according to this embodiment. The functional block diagram in Figure 4 omits the illustration of components that are not necessary for the explanation of this embodiment.

[0039] The prediction unit 42 in Figure 4 has a configuration that includes a known determination unit 60 and a prediction processing unit 64. The prediction processing unit 64 includes a physical model 70 and a statistical model 72. Although Figure 4 shows an example in which the learning unit 66 and the statistical model trained data storage unit 68 are provided separately from the prediction unit 42, the learning unit 66 and the statistical model trained data storage unit 68 may also be provided inside the prediction unit 42.

[0040] Physical model 70 is a predictive model constructed using physical methods. Because physical model 70 is based on physical and scientific laws, it can predict the behavior of the target physical sensor 30a according to these principles. Therefore, physical model 70 can maintain a certain level of predictive accuracy for known data that has been analyzed.

[0041] However, the physical model 70 has difficulty making highly accurate predictions for known data, equivalent to the prediction target physical sensor 30a, and tuning the model to unknown phenomena that have not yet been built or analyzed takes a considerable amount of time.

[0042] The statistical model 72 is a predictive model constructed using statistical methods. The statistical model 72 is constructed using general estimation algorithms such as a Gaussian process regression model or a neural network, based on log data from the substrate processing device 10. For example, other physical sensors 30 that are highly correlated with the physical sensor 30a to be predicted can be calculated based on the log data from the substrate processing device 10. The statistical model 72 has already learned the log data of the physical sensor output values ​​of other physical sensors 30 that are highly correlated with the physical sensor 30a to be predicted. The statistical model 72 can make more accurate predictions than the physical model 70 for known data. However, the statistical model 72 has difficulty making predictions for unknown data.

[0043] Therefore, the prediction unit 42 of this embodiment combines a statistical model 72 capable of making highly accurate predictions for known behaviors with a physical model 70 capable of handling unknown behaviors that are within expectations, thereby realizing a virtual sensor function that can flexibly respond to many conditions.

[0044] The statistical model trained data storage unit 68 stores known data that the statistical model 72 has learned. The known determination unit 60 determines whether the physical sensor output value of the physical sensor 30a to be predicted is known data or not, based on the similarity between the physical sensor output value of the physical sensor 30a to be predicted and the known data that the statistical model 72 has learned, and notifies the prediction processing unit 64 of the result of the known determination.

[0045] If the prediction processing unit 64 determines that the physical sensor output value of the physical sensor 30a to be predicted is known data, it uses the statistical model 72 to predict the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted. If the prediction processing unit 64 determines that the physical sensor output value of the physical sensor 30a to be predicted is not known data, it uses the physical model 70 to predict the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted.

[0046] However, even with the physical model 70, prediction is difficult for completely unforeseen and unknown data. Therefore, the learning unit 66 additionally trains the statistical model 72 with the unknown data. Through this additional training, the unknown data becomes known data that the statistical model 72 has already learned, enabling highly accurate predictions by the statistical model 72.

[0047] The timing of additional training of the statistical model 72 by the learning unit 66 can vary. For example, additional training of the statistical model 72 may be performed at predetermined intervals, whenever unknown data is generated, after processing (RUN) according to a recipe or macro, or in the middle of processing according to a recipe or macro. The learning unit 66 may prepare a statistical model 72 for additional training, and after training the statistical model 72 for additional training, update the statistical model 72 currently in operation in the prediction processing unit 64.

[0048] <Processing> The substrate processing apparatus 10 of this embodiment creates a physical model 70 using a procedure such as that shown in Figure 5. Figure 5 is a flowchart showing an example of the procedure for creating a physical model. The creation of the physical model 70 may be performed, for example, by the server device 14 and registered with the substrate processing apparatus 10, or it may be performed by the device controller 12. Here, we will describe an example where the process is performed by the server device 14.

[0049] In step S10, the server device 14 reads the design information of the substrate processing device 10 from the storage unit where the design information is stored. The design information of the substrate processing device 10 may be stored inside the server device 14 or stored elsewhere.

[0050] In step S12, the server device 14 constructs a physical model using physical methods in accordance with physical and scientific laws. In step S14, the server device 14 evaluates the physical model constructed in step S12 using an existing evaluation method. If the evaluation result in step S14 is not satisfactory, the server device 14 returns to step S10 and continues processing. If the evaluation result in step S14 is satisfactory, the server device 14 proceeds to step S18 and registers the physical model that passed the evaluation in step S14 as an operational physical model in the prediction unit 42 of the device controller 12.

[0051] The substrate processing apparatus 10 of this embodiment creates a statistical model 72 using a procedure such as that shown in Figure 6. Figure 6 is a flowchart showing an example of the procedure for creating a statistical model. The statistical model 72 may be created, for example, by a server device 14 and registered in the substrate processing apparatus 10, or it may be created by a device controller 12. Here, we will describe an example where the process is performed by the server device 14.

[0052] In step S20, the server device 14 reads the log data of the substrate processing device 10 from the storage unit where the log data of the substrate processing device 10 is stored. The log data of the substrate processing device 10 may be stored inside the substrate processing device 10, inside the server device 14, or elsewhere. The log data includes the physical sensor output values ​​output from the physical sensor 30a to be predicted and other physical sensors 30, and the behavior of those physical sensor output values.

[0053] In step S22, the server device 14 performs a correlation analysis between the physical sensor 30a to be predicted and other physical sensors 30, and calculates other physical sensors 30 whose correlation with the physical sensor 30a to be predicted is higher than a threshold. In step S24, the server device 14 reads the physical sensor output values ​​of the other physical sensors 30 whose correlation with the physical sensor 30a to be predicted is higher than a threshold, and the behavior of those physical sensor output values, from the log data, and performs the necessary preprocessing.

[0054] In step S26, the server device 14 constructs a statistical model using statistical methods based on the physical sensor output values ​​of other physical sensors 30 whose correlation with the physical sensor 30a to be predicted is higher than a threshold, and the behavior of those physical sensor output values. In step S28, the server device 14 evaluates the statistical model constructed in step S26 using an existing evaluation method. If the evaluation result in step S28 is not satisfactory, the server device 14 returns to step S20 and continues processing. On the other hand, if the evaluation result in step S28 is satisfactory, the server device 14 proceeds to step S32 and registers the statistical model that passed the evaluation in step S28 as an operational statistical model in the prediction unit 42 of the device controller 12.

[0055] The substrate processing apparatus 10 of this embodiment performs, for example, abnormality detection of the physical sensor 30a to be predicted and redundancy of the physical sensor 30a to be predicted, as shown in Figure 7. Figure 7 is a flowchart showing an example of the procedure for abnormality detection and redundancy of the physical sensor to be predicted.

[0056] In step S50, the acquisition unit 40 of the device controller 12 acquires physical sensor output values ​​output from other physical sensors 30 and the physical sensor 30a to be predicted, and transmits them to the known determination unit 60 of the prediction unit 42. In step S52, the known determination unit 60 of the prediction unit 42 acquires the trained data of the statistical model 72 from the statistical model trained data storage unit 68.

[0057] In step S54, the known determination unit 60 determines the similarity between the physical sensor output value of the physical sensor 30a to be predicted, acquired in step S50, and the trained data of the statistical model 72, acquired in step S52. The similarity between the physical sensor output value of the physical sensor 30a to be predicted and the trained data of the statistical model 72 is the similarity between the data of the explanatory variables that the statistical model 72 has previously learned and the data of the explanatory variables that the prediction processing unit 64 intends to use for the current prediction. The known determination unit 60 determines the similarity by comparing whether the data of the explanatory variables that the prediction processing unit 64 intends to use for the current prediction is included in the data of the explanatory variables that the statistical model 72 has previously learned, or by comparing whether the behavior of the data of the explanatory variables that the prediction processing unit 64 intends to use for the current prediction is included in the behavior of the data of the explanatory variables that the statistical model 72 has previously learned. In addition, the similarity determination may be made using cosine similarity from a predetermined interval before the prediction point.

[0058] In step S56, the known determination unit 60 determines whether the physical sensor output value of the physical sensor 30a to be predicted is known data, based on the similarity between the physical sensor output value of the physical sensor 30a to be predicted and the trained data of the statistical model 72, and notifies the prediction processing unit 64 of the result of the known determination.

[0059] If the physical sensor output value of the physical sensor 30a to be predicted is known data, the prediction processing unit 64 proceeds to step S58 and predicts the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a using the statistical model 72. If the physical sensor output value of the physical sensor 30a to be predicted is not known data, the prediction processing unit 64 proceeds to step S60 and predicts the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a using the physical model 70.

[0060] The process proceeds from step S58 or S60 to step S62, where the abnormality determination unit 44 compares the physical sensor output value of the physical sensor 30a to be predicted with the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted, and determines whether or not an abnormality has occurred in the physical sensor 30a to be predicted.

[0061] For example, the abnormality determination unit 44 uses the physical sensor output value of the physical sensor 30a to be predicted, as shown in Figure 8, and the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted, to determine whether or not an abnormality has occurred in the physical sensor 30a to be predicted.

[0062] Figure 8 is an illustrative diagram of an example of a process for detecting anomalies in the physical sensor being predicted. In Figure 8, the virtual sensor output value of the virtual sensor is shown as "○", and the physical sensor output value of the physical sensor 30a being predicted is shown as "●". The anomaly determination unit 44 sets a normal range (bandwidth) from the virtual sensor output value of the virtual sensor, and if each point "●" of the physical sensor output value falls outside the bandwidth, it determines that an anomaly has occurred in the physical sensor 30a being predicted.

[0063] Thus, the abnormality determination unit 44 determines that an abnormality has occurred in the physical sensor 30a if the difference between the physical sensor output value of the physical sensor 30a to be predicted and the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted is greater than or equal to a predetermined value. The virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted is treated as the ideal physical sensor output value of the physical sensor 30a to be predicted.

[0064] When the abnormality of the physical sensor 30a to be predicted is detected, the abnormality determination unit 44 notifies the instruction unit 48 that an abnormality has occurred in the physical sensor 30a to be predicted. In step S66, the instruction unit 48 switches the control of the controlled object 32 from control based on the physical sensor output value of the physical sensor 30a to control based on the virtual sensor output value of the virtual sensor corresponding to the physical sensor 30a to be predicted.

[0065] Thus, when the instruction unit 48 receives information indicating that an abnormality has occurred in the physical sensor 30a being predicted, it can perform redundant operation by controlling the target 32 ​​based on the virtual sensor output value of the virtual sensor.

[0066] Furthermore, unless the instruction unit 48 receives information indicating that an abnormality has occurred in the physical sensor 30a being predicted, it controls the control target 32 ​​based on the physical sensor output value of the physical sensor 30a being predicted.

[0067] The prediction unit 42 further trains the statistical model 72 with unknown data using a procedure such as that shown in Figure 9. Figure 9 is a flowchart showing an example of the procedure for further training the statistical model.

[0068] In step S80, the acquisition unit 40 of the device controller 12 acquires physical sensor output values ​​output from other physical sensors 30 and the physical sensor 30a to be predicted, and transmits them to the known determination unit 60 of the prediction unit 42. In step S82, the known determination unit 60 of the prediction unit 42 acquires the trained data of the statistical model 72 from the statistical model trained data storage unit 68.

[0069] In step S84, the known determination unit 60 determines the similarity between the physical sensor output value of the physical sensor 30a to be predicted, which was acquired in step S80, and the trained data of the statistical model 72, which was acquired in step S82.

[0070] In step S86, the known determination unit 60 determines whether the physical sensor output value of the physical sensor 30a to be predicted is known data, based on the similarity between the physical sensor output value of the physical sensor 30a to be predicted and the trained data of the statistical model 72, and notifies the prediction processing unit 64 of the result of the known determination.

[0071] If the physical sensor output value of the physical sensor 30a to be predicted is known data, the prediction processing unit 64 returns to step S80. If the physical sensor output value of the physical sensor 30a to be predicted is unknown data, the learning unit 66 proceeds to step S88 and additionally trains the statistical model 72 with the physical sensor output value of the physical sensor 30a to be predicted.

[0072] In step S90, the learning unit 66 evaluates the statistical model 72 that was further trained in step S88 using an existing evaluation method. If the evaluation result in step S90 is not satisfactory, the learning unit 66 returns to step S80 and continues processing. On the other hand, if the evaluation result in step S90 is satisfactory, the learning unit 66 proceeds to step S94 and registers the further trained statistical model 72 as an operational statistical model with the prediction processing unit 64.

[0073] Thus, in this embodiment, if the physical sensor output value of the physical sensor 30a to be predicted is unknown data, the statistical model 72 used in the prediction unit 42 is further trained, thereby increasing the amount of known data and realizing a prediction model that can adapt to changes over time.

[0074] The control method according to this embodiment can be applied to a substrate processing apparatus 10 that controls the opening degree of an automatic pressure control device 100 based on the physical sensor output value of a pressure sensor 102, for example, as shown in Figure 10. Figure 10 is a functional block diagram of an example of a substrate processing apparatus that controls the opening degree of an automatic pressure control device based on the physical sensor output value of a pressure sensor.

[0075] In the substrate processing apparatus 10 shown in Figure 10, the physical sensor output value of the pressure sensor 102 is compared with the virtual sensor output value of the virtual sensor corresponding to the pressure sensor 102 to determine whether or not an abnormality has occurred in the pressure sensor 102. If it is determined that an abnormality has occurred in the pressure sensor 102, the substrate processing apparatus 10 switches the control of the automatic pressure control device 100 from control based on the physical sensor output value of the pressure sensor 102 to control based on the virtual sensor output value of the virtual sensor corresponding to the pressure sensor 102. In this way, when information indicating that an abnormality has occurred in the pressure sensor 102 is received, the instruction unit 48 can control the automatic pressure control device 100 according to a recipe or macro based on the virtual sensor output value of the virtual sensor corresponding to the pressure sensor 102.

[0076] The control method according to this embodiment can be applied to a substrate processing apparatus 10 that controls the opening degree of a valve 116 based on the physical sensor output value of a flow sensor 112 of a mass flow controller 110, as shown in Figure 11. Figure 11 is a functional block diagram of an example of a substrate processing apparatus that controls the opening degree of a valve based on the physical sensor output value of a flow sensor.

[0077] In the substrate processing apparatus 10 shown in Figure 11, if no abnormality occurs in the flow sensor 112, the flow rate setting value is transmitted from the instruction unit 48 of the apparatus controller 12 to the mass flow controller 110. The control circuit 114 of the mass flow controller 110 controls the opening degree of the valve 116 based on the flow rate setting value received from the instruction unit 48 and the physical sensor output value of the flow sensor 112.

[0078] Therefore, if an abnormality occurs in the flow sensor 112, the control circuit 114 cannot control the opening degree of the valve 116 to achieve an appropriate flow rate. In the substrate processing device 10 of Figure 11, the physical sensor output value of the flow sensor 112 is compared with the virtual sensor output value of the virtual sensor corresponding to the flow sensor 112 to determine whether or not an abnormality has occurred in the flow sensor 112. In the substrate processing device 10 of Figure 11, if it is determined that an abnormality has occurred in the flow sensor 112, the opening degree setting value of the valve 116 is transmitted from the instruction unit 48 of the device controller 12 to the mass flow controller 110 to achieve an appropriate flow rate. The relationship between the virtual sensor output value of the virtual sensor corresponding to the flow sensor 112 and the opening degree setting value of the valve 116 is assumed to be registered in advance in the device controller 12.

[0079] When the instruction unit 48 receives information indicating that an abnormality has occurred in the flow sensor 112, it can switch from controlling the opening degree of the valve 116 based on the physical sensor output value of the flow sensor 112 to controlling the opening degree of the valve 116 based on the virtual sensor output value of the virtual sensor corresponding to the flow sensor 112.

[0080] According to this embodiment, it is possible to handle sudden failures such as the sudden cessation of output of the physical sensor output value from the physical sensor 30a to be predicted, or failures in which the physical sensor output value deviates from the intended value.

[0081] Although preferred embodiments of the present invention have been described in detail above, the present invention is not limited to the embodiments described above, and various modifications and substitutions can be made to the embodiments described above without departing from the scope of the present invention. [Explanation of Symbols]

[0082] 1. Substrate Processing System 10 Substrate Processing Equipment 12. Device Controller 14 Server Devices 16. Worker terminal 18, 20 Network 30 Other physical sensors 30a Physical sensor to be predicted 32 Controlled object 40 Acquisition Department 42 Prediction Section 44 Abnormality determination section 46 Notification Department 48 Instruction section 60 Known Determination Unit 64 Prediction Processing Unit 66 Learning Department 68 Statistical model trained data storage unit 70 Physical Models 72 Statistical Models

Claims

1. An acquisition unit configured to acquire physical sensor output values ​​output from multiple physical sensors installed in a substrate processing device, A prediction unit is configured to predict the virtual sensor output value of a virtual sensor corresponding to the physical sensor to be predicted, by switching between the statistical model or a physical model constructed by a physical method, based on the similarity between the physical sensor output value and data trained by a statistical model constructed by a statistical method. An abnormality determination unit is configured to determine an abnormality in the physical sensor by comparing the physical sensor output value of the physical sensor to be predicted with the virtual sensor output value of the virtual sensor, An instruction unit is configured to switch from control based on the physical sensor output value of the predicted physical sensor to control based on the virtual sensor output value of the virtual sensor when it is determined that an abnormality has occurred in the physical sensor, An information processing device having

2. The prediction unit is configured to predict the virtual sensor output value of the virtual sensor using the statistical model when the similarity is higher than a threshold, and to predict the virtual sensor output value of the virtual sensor using the physical model when the similarity is not higher than a threshold. The information processing apparatus according to claim 1, characterized in that

3. The prediction unit is configured to predict the virtual sensor output value of the virtual sensor from the physical sensor output values ​​of other physical sensors whose correlation with the physical sensor to be predicted is higher than a threshold. The information processing apparatus according to claim 1, characterized in that

4. The system further includes a notification unit configured to notify the operator of an abnormality in the physical sensor when it is determined that an abnormality has occurred in the physical sensor. The information processing apparatus according to any one of claims 1 to 3.

5. In an information processing device, Acquisition procedure for obtaining physical sensor output values ​​from multiple physical sensors installed in a circuit board processing device. A prediction procedure that predicts the virtual sensor output value of a virtual sensor corresponding to the target physical sensor by switching between the statistical model or the physical model constructed by the physical method, based on the similarity between the physical sensor output value and the data trained by the statistical model constructed by the statistical method, An anomaly determination procedure for determining an anomaly in the physical sensor by comparing the physical sensor output value of the physical sensor to be predicted with the virtual sensor output value of the virtual sensor, An instruction procedure to switch from control based on the physical sensor output value of the predicted physical sensor to control based on the virtual sensor output value of the virtual sensor when it is determined that an abnormality has occurred in the physical sensor, A program to execute.

6. A control method performed by a substrate processing system having one or more substrate processing devices and an information processing device that is communicatively connected to the substrate processing devices, To acquire physical sensor output values ​​from multiple physical sensors installed in the circuit board processing device, Based on the similarity between the aforementioned physical sensor output value and the data trained by the statistical model constructed using statistical methods, the system switches between the statistical model and the physical model constructed using physical methods to predict the virtual sensor output value of the virtual sensor corresponding to the physical sensor to be predicted. The physical sensor output value of the physical sensor to be predicted is compared with the virtual sensor output value of the virtual sensor to determine an abnormality in the physical sensor. When it is determined that an abnormality has occurred in the physical sensor, the control is switched from control based on the physical sensor output value of the predicted physical sensor to control based on the virtual sensor output value of the virtual sensor, A control method having

Images