Substrate processing apparatus, data processing method and data processing program

Abstract

This invention reduces the workload associated with learning tasks. The substrate processing apparatus includes: a memory feature generation unit that inputs first sensor data acquired in a time series and outputs memory feature values; a learning unit that, during a learning period, learns the weight parameters by associating prediction result data obtained by calculating the memory feature values ​​based on weight parameters with second sensor data acquired in a time series; a prediction unit that, during a prediction period, calculates the memory feature values ​​to be output by the memory feature generation unit based on the learned weight parameters using the first sensor data of the time series input, and outputs prediction result data; and a determination unit that, during the prediction period, determines the state of the substrate manufacturing process by comparing the prediction result data with the second sensor data acquired in a time series.

Description

[Technical Field]

[0001] This invention relates to a substrate processing apparatus, a data processing method, and a data processing program. [Previous Technology]

[0002] In the field of substrate processing equipment, time-series sensor data obtained from various sensors has been used to determine the status of the substrate manufacturing process (e.g., whether there are any abnormalities). Furthermore, to improve the accuracy of this determination, machine learning models have recently been proposed. [Previous Art Documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2018-77779 [Non-Patent Document]

[0004] [Non-patent document 1] David Sussillo, LFAbbott, "Generating Coherent Patterns of Activity from Chaotic Neural Networks", Neuron. Author manuscript; available in PMC 2010 Aug 27. [Summary of the Invention]

[0005] [The problem the invention aims to solve]

[0006] However, in the case of a typical machine learning model, a large amount of learning data needs to be collected to improve the judgment accuracy. Therefore, in the case of a board processing device that applies a machine learning model, the workload related to the learning operation during startup is relatively high. Furthermore, after startup, if the judgment accuracy decreases over time as the board processing device changes, a large amount of learning data also needs to be collected again. Therefore, in the case of a board processing device that applies a machine learning model, the workload related to the learning operation during maintenance is also relatively high.

[0007] This invention reduces the workload associated with learning operations in a substrate processing apparatus that uses time-series sensor data to determine the status of a substrate manufacturing process. [Technical Means for Solving the Problem]

[0008] One embodiment of the substrate processing apparatus of the present invention has, for example, the following configuration: a memory feature generation unit that inputs first sensor data of a time series acquired during a substrate manufacturing process and outputs memory feature data; a learning unit that, during a learning period, learns the weight parameters by relating prediction result data obtained by calculating the memory feature data output by the memory feature generation unit based on weight parameters with second sensor data of a time series acquired during the substrate manufacturing process; a prediction unit that, during a prediction period, calculates the memory feature data to be output by the memory feature generation unit based on the weight parameters learned by the learning unit using the first sensor data of the time series acquired during the substrate manufacturing process, and outputs prediction result data; and a determination unit that, during the prediction period, determines the state of the substrate manufacturing process by comparing the prediction result data with the second sensor data of the time series acquired during the substrate manufacturing process. [Effects of the Invention]

[0009] In a substrate processing apparatus that uses time-series sensor data to determine the status of a substrate manufacturing process, the workload related to learning operations can be reduced.

Implementation Method

[0011] Hereinafter, each embodiment will be described with reference to the accompanying drawings. Furthermore, in the abstract and drawings of this invention, repeated descriptions of constituent elements that have substantially the same functional configuration are omitted by adding the same symbols.

[0012] [First Embodiment] <System Configuration of the Substrate Processing Apparatus> First, the system configuration of the substrate processing apparatus according to the first embodiment will be described. FIG1 is the first figure showing an example of the system configuration of the substrate processing apparatus, and the arrows in the figure indicate data flow. Among them, FIG1(a) shows the data flow in the substrate processing apparatus 100 during the learning period, and FIG1(b) shows the data flow in the substrate processing apparatus 100 during the prediction period.

[0013] As shown in FIG1(a) and (b), the substrate processing apparatus 100 includes a substrate processing chamber 110, a sensor a120 (an example of a first sensor), a data processing device 130, a sensor b140 (an example of a second sensor), a control device 150, and an actuator 160.

[0014] As shown in Figures 1(a) and (b), during substrate processing in chamber 110, time-series sensor data measured by sensor a120 is sent to the storage calculation unit (hereinafter referred to as RC unit) 131 of data processing device 130. That is, the time-series sensor data sent by RC unit 131 is an example of the first time-series sensor data obtained in the substrate manufacturing process.

[0015] The RC unit 131 performs memory computation with higher expressive power and faster learning speed. Specifically, the RC unit 131 predicts the time series sensor data (input data) measured by the sensor a120 based on the sensor data (input data) of the time series measured by the sensor a120.

[0016] To this end, as shown in FIG1(a), during the learning period, the RC unit 131 acquires the time series sensor data (forward solution data) measured by the sensor b140 via the acquisition unit 132 of the data processing device 130. Then, the RC unit 131 learns the weight parameters by associating the prediction result data obtained by calculating the time series sensor data (input data) measured by the sensor a120 based on the weight parameters with the forward solution data.

[0017] On the other hand, as shown in FIG1(b), during the prediction period, the RC unit 131 calculates the time series sensor data (input data) measured by the sensor a120 based on the learned weight parameters, and outputs the time series sensor data (prediction result data). Furthermore, the time series sensor data (prediction result data) output by the RC unit 131 is notified to the acquisition unit 132 of the data processing device 130.

[0018] The sensor b140 transmits the time-series sensor data measured during the substrate processing in the chamber 110 to the acquisition unit 132 of the data processing device 130. That is, the time-series sensor data transmitted to the acquisition unit 132 is an example of the second time-series sensor data acquired during the substrate manufacturing process.

[0019] As shown in FIG1(a), the acquisition unit 132 of the data processing device 130 notifies the RC unit 131 of the time series sensor data (forward data) measured by the sensor b140 during the learning period.

[0020] Furthermore, as shown in FIG1(b), the acquisition unit 132 of the data processing apparatus 130, during the prediction period, notifies the determination unit 133 of the prediction result data notified by the RC unit 131 and the sensor data (comparison data) of the time series measured by the sensor b140. Moreover, the sensor data of the time series measured by the sensor b140 is called "correct solution data" during the learning period and "comparison data" during the prediction period.

[0021] Furthermore, as shown in FIG1(b), during the prediction period, the determination unit 133 calculates the difference between the prediction result data notified by the acquisition unit 132 and the comparison data. Based on the calculated difference data, the status of the substrate manufacturing process is determined. The determination result of the determination unit 133 is notified to the control device 150, and the control device 150 controls the actuator 160 according to the notified determination result.

[0022] Thus, in the first embodiment, the substrate processing apparatus 100 that uses time-series sensor data to determine the state of the substrate manufacturing process includes: ・an RC unit 131 that performs memory calculations with higher expressive power and faster learning speed; ・learning is performed during the learning period by using two types of time-series sensor data as input data and correct answer data; ・the state of the substrate manufacturing process is determined during the prediction period by using two types of time-series sensor data as input data and comparison data.

[0023] Accordingly, the substrate processing apparatus 100 of the first embodiment can efficiently perform learning during startup and maintenance, thereby reducing the workload of the learning operation.

[0024] <Hardware Configuration of the Data Processing Device> Next, the hardware configuration of the data processing device 130 included in the substrate processing device 100 will be described. Figure 2 is a diagram showing an example of the hardware configuration of the data processing device. In particular, Figure 2(a) shows the case where the function of generating memory features is implemented using a memory feature generator as hardware (details are described below). Also, Figure 2(b) shows the case where the function of generating memory features is implemented by executing software (e.g., a memory feature generation program) on a computer.

[0025] As shown in FIG2, the data processing device 130 includes a processor 201, a memory 202, an auxiliary memory device 203, and an external I / F device 206 (in the case of FIG2(a), it further includes an I / F device 204 and a memory feature generator 205). Furthermore, the hardware of the data processing device 130 is interconnected via a bus 207.

[0026] The processor 201 in Figures 2(a) and (b) has various computing devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 201 reads various programs (such as the data processing program described below, which in the case of Figure 2(b) is a data processing program that includes a program for generating memory features, etc.) onto the memory 202 and executes them.

[0027] The memory 202 in Figures 2(a) and (b) has main memory devices such as ROM (Read Only Memory) and RAM (Random Access Memory). The processor 201 and the memory 202 form a so-called computer. By executing various programs read from the memory 202 by the processor 201, the computer can perform various functions (RC unit 131, acquisition unit 132, determination unit 133).

[0028] The auxiliary memory device 203 in Figures 2(a) and (b) stores various programs or various data used when the various programs are executed by the processor 201.

[0029] The I / F device 204 in Figure 2(a) is a connection device connected to the memory feature generator 205. The memory feature generator 205 in Figure 2(a) is implemented, for example, by an FPGA (Field-Programmable Gate Array) board. The memory feature generator 205 has the function of generating memory features, that is, it takes as input the time series sensor data measured by the sensor a120 and outputs the memory features.

[0030] The external I / F device 206 in Figures 2(a) and (b) is a connection device that connects to sensor a120 and sensor b140.

[0031] <Functional Configuration of the RC Unit of the Data Processing Device> Next, the functional configuration of the RC unit 131 of the data processing device 130 will be explained. Figure 3 is the first figure showing an example of the functional configuration of the RC unit of the data processing device.

[0032] As shown in FIG3, the RC unit 131 includes a memory feature generation unit 300, a prediction unit 310, a period determination unit 320, a switching unit 330, and a learning unit 340. The memory feature generation unit 300, as described above, can also be implemented using a memory feature generator 205, which is hardware such as an FPGA board. However, in this embodiment, the memory feature generation unit 300 is described as having the function of having a memory feature generation program executed by a computer.

[0033] Furthermore, the prediction unit 310 to the learning unit 340 have the function of executing a data processing program (a data processing program other than the memory feature generation program) by a computer.

[0034] As shown in FIG3, the memory feature generation unit 300 has the function of generating memory features, that is, it inputs the time series sensor data a (input data) acquired by the sensor a120 and outputs the memory feature.

[0035] The prediction unit 310 includes a weight acquisition unit 311 and a calculation unit 312. The weight acquisition unit 311 acquires the weight parameters learned by the learning unit 340 and stored in the data storage unit 350, and sets them in the calculation unit 312. Furthermore, during the initial learning phase, preset weight parameters can be set as initial values ​​in the calculation unit 312, for example, values ​​generated by random numbers can be set.

[0036] The calculation unit 312 calculates the memory feature quantity output by the memory feature quantity generation unit 300 based on the weight parameters set by the weight acquisition unit 311, and notifies the prediction result data obtained therefrom to the period determination unit 320.

[0037] The period determination unit 320 determines whether the current period is a learning period or a prediction period based on the prediction result data notified by the calculation unit 312 and the sensor data b (positive resolution data) notified by the acquisition unit 132, and notifies the switching unit 330 of the period determination result. Furthermore, the period determination unit 320 notifies the switching unit 330 of the prediction result data regardless of whether the current period is a learning period or a prediction period.

[0038] During the learning period, the switching unit 330 notifies the learning unit 340 of the prediction result data notified by the period determination unit 320. Also, during the prediction period, the switching unit 330 notifies the acquisition unit 132 of the prediction result data notified by the period determination unit 320.

[0039] The learning unit 340 includes a weighting calculation unit 341. The weighting calculation unit 341 calculates the error between sensor data b, which is measured by sensor b140 and notified as positive solution data by acquisition unit 132, and prediction result data notified by switching unit 330. Furthermore, the weighting calculation unit 341 learns weight parameters based on the calculated error. At this time, the weighting calculation unit 341 learns the weight parameters by matrix operation in a manner where the error is close to zero (i.e., in a manner that correlates the prediction result data with the positive solution data).

[0040] Furthermore, the weight calculation unit 341 stores the learned weight parameters in the data storage unit 350. The data storage unit 350 is implemented, for example, in the auxiliary memory device 203 of the data processing device 130.

[0041] <Memory Calculation Processing Flow> Next, the memory calculation processing flow of RC section 131 will be explained. Figure 4 is an example of a flowchart showing the memory calculation processing flow.

[0042] In step S401, when the RC unit 131 begins learning, the weight acquisition unit 311 of the prediction unit 310 initializes the weight parameters and sets the preset weight parameters in the calculation unit 312. Furthermore, the memory feature generation unit 300 begins to input sensor data of the time series measured by the sensor a120 and outputs memory feature values. Thereby, the calculation unit 312 of the prediction unit 310 calculates the output memory feature values ​​based on the weight parameters and outputs prediction result data.

[0043] In step S402, the period determination unit 320 determines whether it is a learning period or a prediction period. If it is determined to be a learning period in step S402, the switching unit 330 switches the notification target of the prediction result data to the learning unit 340 and proceeds to step S403. Furthermore, in this embodiment, the RC unit 131 causes the period to start from the learning period.

[0044] In step S403, the learning unit 340 performs learning processing, that is, it learns weight parameters based on the prediction result data and the correct solution data notified by the switching unit 330.

[0045] In step S404, the period determination unit 320 determines whether the learning termination condition is met (whether the state of no error between the correct solution data and the prediction result data has passed the specified time). If it is determined in step S403 that the learning termination condition is not met, it is determined in step S404 that the process will not proceed to the prediction period (it is determined to be no in step S404), and proceeds to step S409.

[0046] On the other hand, if it is determined in step S403 that the learning termination condition is met, it is determined in step S404 that the process should transition to the prediction period (it is determined to be YES in step S404), and proceeds to step S405. Furthermore, in this case, the weight acquisition unit 311 of the prediction unit 310 reads the weight parameters at the end of the learning process from the data storage unit 350 and sets them in the calculation unit 312.

[0047] In step S405, the period determination unit 320 moves to the prediction period and proceeds to step S409. Thereby, the switching unit 330 switches the notification target of the prediction result data to the acquisition unit 132.

[0048] On the other hand, in step S402, if it is determined to be a prediction period, the process proceeds to step S406.

[0049] In step S406, the calculation unit 312 of the prediction unit 310 performs prediction processing, that is, it calculates the memory feature quantity output by the memory feature quantity generation unit 300 based on the weight parameters at the end of learning, and outputs the prediction result data.

[0050] In step S407, the determination unit 320 determines whether relearning is required. In step S407, if it is determined that relearning is not required (if it is NO in step S407), the process proceeds to step S409. On the other hand, in step S407, if it is determined that relearning is required (if it is NO in step S407), the process proceeds to step S408.

[0051] In step S408, the period determination unit 320 moves to the learning period and proceeds to step S409. Thereby, the switching unit 330 switches the notification target of the prediction result data to the learning unit 340.

[0052] In step S409, the RC unit 131 determines whether to end the memory calculation process.

[0053] In step S409, if it is determined that the memory calculation process should continue (if it is not determined in step S409), the process returns to step S402. On the other hand, in step S409, if it is determined that the memory calculation process should end (if it is determined in step S409), the memory calculation process ends.

[0054] <Functional Configuration of the Acquisition and Determination Units of the Data Processing Apparatus> Next, the functional configuration of the acquisition unit 132 and the determination unit 133 of the data processing apparatus 130 will be explained. Figure 5 is a diagram showing an example of the functional configuration of the acquisition and determination units of the data processing apparatus. As shown in Figure 5, the acquisition unit 132 includes a sensor data acquisition unit 501 and a prediction result data acquisition unit 503. Furthermore, the determination unit 133 includes a difference calculation unit 511 and a state determination unit 512.

[0055] The sensor data acquisition unit 501 notifies the RC unit 131 of the time series sensor data b obtained by the sensor b140 as the correct solution data and notifies the determination unit 133 as the comparison data.

[0056] During the prediction period, the prediction result data acquisition unit 503 acquires the prediction result data sent by the RC unit 131 and notifies the judgment unit 133.

[0057] The differential calculation unit 511 calculates the difference between the comparison data notified by the sensor data acquisition unit 501 during the prediction period and the prediction result data sent by the RC unit 131 during the prediction period, and generates differential data. Furthermore, the differential calculation unit 511 notifies the state determination unit 512 of the generated differential data.

[0058] The state determination unit 512 determines the state of the substrate manufacturing process. Specifically, the state determination unit 512 determines whether the differential data notified by the differential calculation unit 511 exceeds a specific threshold. When the state determination unit 512 determines that the specific threshold is exceeded, it determines that an abnormality has occurred in the processed substrate. Furthermore, the state determination unit 512 outputs an abnormality detection result indicating that an abnormality has occurred.

[0059] Furthermore, the status determination unit 512 determines, based on the differential data notified by the differential calculation unit 511, whether the change pattern within a specific time range is similar to a specific change pattern. When the status determination unit 512 determines that the change pattern is similar to the specific change pattern, it determines that an abnormality or malfunction has occurred in the substrate processing apparatus 100. The status determination unit 512 then outputs a fault diagnosis result indicating that an abnormality or malfunction has occurred in the substrate processing apparatus 100 to the control device 150. In this way, the control device 150 can notify the actuator 160 of a stop command when an abnormality or malfunction occurs in the substrate processing apparatus 100.

[0060] <Acquisition and Judgment Processing Flow> Next, the acquisition and judgment processing flow of the acquisition unit 132 and judgment unit 133 of the data processing device 130 will be explained, specifically the acquisition and judgment processing flow when outputting anomaly detection results. Figure 6 is the first flowchart showing the acquisition and judgment processing flow.

[0061] In step S601, the acquisition unit 132 of the data processing device 130 acquires the time series sensor data b measured by the sensor b140.

[0062] In step S602, the acquisition unit 132 of the data processing device 130 determines whether the current period is a learning period or a prediction period. In step S602, if it is determined to be a learning period, the process proceeds to step S603.

[0063] In step S603, the acquisition unit 132 of the data processing device 130 sends the acquired sensor data b as correct solution data to the RC unit 131.

[0064] On the other hand, if the situation is determined to be during the prediction period in step S602, the process proceeds to step S604.

[0065] In step S604, the acquisition unit 132 of the data processing device 130 acquires the prediction result data sent by the RC unit 131.

[0066] In step S605, the determination unit 133 of the data processing device 130 compares the sensor data b acquired by the sensor b140 with the prediction result data acquired by the RC unit 131 as comparison data, and calculates the difference data.

[0067] In step S606, the determination unit 133 of the data processing device 130 determines whether the differential data exceeds a specific threshold. If it is determined in step S606 that the data exceeds the specific threshold (if it is in step S606), the process proceeds to step S607.

[0068] In step S607, the determination unit 133 of the data processing device 130 determines that an abnormality has occurred in the processed substrate, outputs the abnormality detection result indicating the abnormality to the control device 150, and then proceeds to step S608.

[0069] On the other hand, if it is determined in step S606 that the specific threshold has not been exceeded (if it is no in step S606), the process proceeds directly to step S608.

[0070] In step S608, the acquisition unit 132 of the data processing device 130 determines whether to end the acquisition and determination process. If it is determined in step S608 that the acquisition and determination process should not be ended (if it is not in step S608), the process returns to step S602. On the other hand, if it is determined in step S608 that the acquisition and determination process should be ended (if it is in step S608), the acquisition and determination process ends.

[0071] Next, the acquisition and determination process of the acquisition unit 132 and determination unit 133 of the data processing device 130 will be described, and the acquisition and determination process when outputting fault diagnosis results will be explained. Figure 7 is a second flowchart showing the acquisition and determination process.

[0072] The difference between this flowchart and the first flowchart shown in Figure 6 is in steps S701 and S702. Therefore, steps S701 and S702 will be explained here.

[0073] In step S701, the determination unit 133 of the data processing device 130 determines whether the change pattern of a specific time range of the differential data is similar to a specific change pattern. If it is determined in step S701 that the change pattern is similar to the specific change pattern (if it is in step S701), the process proceeds to step S702.

[0074] In step S702, the determination unit 133 of the data processing device 130 determines that an abnormality or malfunction has occurred in the substrate processing device 100, outputs the fault diagnosis result to the control device 150, and then proceeds to step S608.

[0075] On the other hand, if it is determined in step S702 that the change pattern is not similar to a specific change pattern (when it is NO in step S702), the process proceeds directly to step S608.

[0076] <Specific Example of Acquisition and Judgment Processing> Next, a specific example of acquisition and judgment processing by the acquisition unit 132 and judgment unit 133 of the data processing device 130 will be described. Figure 8 is a diagram showing a specific example of acquisition and judgment processing. In Figure 8, the horizontal axis represents time, and the vertical axis represents data value.

[0077] Furthermore, in Figure 8, line graph 801 represents the prediction result data sent by RC unit 131, and line graph 802 represents the time series sensor data b (comparison data) measured by sensor b140. Furthermore, bar graph 803 represents the difference data calculated based on the difference between the prediction result data and the comparison data.

[0078] In the case of the example in Figure 8, it is shown that at the time point indicated by symbol 810, the difference between the predicted data and the comparison data exceeds a certain threshold, so an anomaly is determined to have occurred, and the anomaly detection result is output.

[0079] Furthermore, in the case of the example in Figure 8, it is shown that the change pattern of the differential data over the time range indicated by symbol 820 is determined to be similar to a specific change pattern.

[0080] On the other hand, in the case of the example in Figure 8, it is shown that the change pattern of the differential data of the time range indicated by symbol 830 is similar to a specific change pattern, so it is determined that a fault has occurred in the substrate processing device 100 and a fault diagnosis result is output.

[0081] <Summary> As can be clearly understood from the above description, regarding the substrate processing apparatus 100 of the first embodiment: • It has a memory feature generation unit that inputs time-series sensor data a (time-series sensor data a measured by sensor a) acquired during the substrate manufacturing process and outputs memory feature quantities. • During the learning period, it learns weight parameters by relating the prediction result data obtained by calculating the memory feature quantities output by the memory feature generation unit based on weight parameters with the time-series sensor data b measured by sensor b. • During the prediction period, it calculates the memory feature quantities output by the memory feature generation unit based on the learned weight parameters using the time-series sensor data a acquired during the substrate manufacturing process and outputs prediction result data. • It compares the prediction result data with the time-series sensor data b acquired during the substrate manufacturing process to determine the state of the substrate manufacturing process.

[0082] Accordingly, the substrate processing apparatus 100 according to the first embodiment can efficiently perform learning during startup and maintenance. As a result, according to the first embodiment, in the substrate processing apparatus that uses time-series sensor data to determine the state of the substrate manufacturing process, the workload related to learning operations can be reduced.

[0083] [Second Embodiment] In the first embodiment described above, the case where the RC unit 131 has only one set of memory feature quantity generating units will be described. However, the number of memory feature quantity generating units in the RC unit 131 is not limited to one set, and for example, it may have multiple sets. Hereinafter, the second embodiment will be described focusing on its differences from the first embodiment described above.

[0084] <Functional Configuration of the RC Section of the Data Processing Apparatus> First, the functional configuration of the RC section 900 in the data processing apparatus 130 of the substrate processing apparatus 100 of the second embodiment will be described. Figure 9 is the second figure showing an example of the functional configuration of the RC section of the data processing apparatus.

[0085] The difference between the functional configuration shown in Figure 1 of FIG3 and the configuration shown in FIG9 is that, in the case of FIG9, • the RC unit 900 has a plurality of memory feature generation units 300_1 to 300_L (L is any integer); • the time series sensor data a obtained by the sensor a120 is input in parallel to the plurality of memory feature generation units 300_1 to 300_L; • the memory feature values ​​output from the plurality of memory feature generation units 300_1 to 300_L are input in parallel to the calculation unit 312 of the prediction unit 310.

[0086] Accordingly, in the calculation unit 312, the prediction result data obtained by weighting and summing the memory feature quantities output from the plurality of memory feature quantity generation units 300_1 to 300_L respectively by using weight parameters is notified to the period determination unit 320.

[0087] <Summary> As can be clearly understood from the above description, regarding the substrate processing apparatus 100 of the second embodiment: ・It has a plurality of memory feature generation units that input in parallel the time-series sensor data a (the time-series sensor data a measured by sensor a) acquired during the substrate manufacturing process and output memory features respectively. ・During the learning period, it learns weight parameters by relating the prediction result data obtained by weighted summing of the memory features output by the plurality of memory feature generation units with time-series sensor data b using weight parameters. ・During the prediction period, it outputs prediction result data by weighted summing of the memory features to be output by the plurality of memory feature generation units using the learned weight parameters, based on the time-series sensor data a acquired in parallel during the substrate manufacturing process. ・It compares the prediction result data with the time-series sensor data b acquired during the substrate manufacturing process to determine the state of the substrate manufacturing process.

[0088] Accordingly, the substrate processing apparatus 100 according to the second embodiment can output prediction result data with higher accuracy while enjoying the same effect as the first embodiment.

[0089] [Third Embodiment] In the first and second embodiments described above, the case where the data processing apparatus 130 has only one set of RC units 131 will be described. However, the number of RC units in the data processing apparatus is not limited to one set, and for example, multiple sets may be provided. Hereinafter, the third embodiment will be described focusing on its differences from the first and second embodiments described above.

[0090] <Functional Configuration of the RC Section of the Data Processing Apparatus> First, the functional configuration of the RC section of the data processing apparatus 1000 included in the substrate processing apparatus 100 of the third embodiment will be described. Figures 10A and 10B are the third and fourth figures showing an example of the functional configuration of the RC section of the data processing apparatus.

[0091] The difference between the functional configuration shown in Figure 1 of FIG3 and that in FIG10A and FIG10B, ・a plurality of RC units 1010_1 to 1010_L are connected in series; ・the time series sensor data obtained by sensor a120 is input to the starting RC unit 1010_1; ・during the learning period, the prediction result data output from the prediction unit 310 of each RC unit 1010_1 to 1010_L-1 is input to the memory feature generation unit 300_2 to 300_L of the next RC unit; ・during the learning period, the switching unit 1020 of each RC unit 1010_1 to 1010_L-1 is in the ON (connected) state, and the prediction result data output from the corresponding prediction unit 310 is input to the corresponding learning unit 340 for learning. • During the learning period, the switching unit 1021 of the RC unit 1010_L of the terminal switches to the learning unit 340 side, and the prediction result data output from the prediction unit 310 is input to the learning unit 340, thereby performing learning in the learning unit 340; • During the prediction period, the switching units 1020 of each RC unit 1010_1 to 1010_L-1 are in the OFF (disconnected) state, and the prediction result data output from the corresponding prediction unit 310 is input to the memory feature generation units 300_2 to 300_L of the next RC unit; • During the prediction period, the switching unit 1021 of the RC unit 1010_L of the terminal switches to the acquisition unit 132 side, and the final prediction result data output from the corresponding prediction unit 310 is output to the acquisition unit 132.

[0092] <Summary> As can be clearly understood from the above description, regarding the substrate processing apparatus 100 of the third embodiment: ・It has a plurality of RC units connected in series. During the prediction period, the time-series sensor data a (the time-series sensor data a measured by sensor a) acquired in the substrate manufacturing process is input to the starting RC unit, and the final prediction result data is output from the ending RC unit. ・During the learning and prediction periods, the prediction result data output by the previous RC unit is input to the memory feature quantity generation unit of each RC unit after the starting point. ・During the learning period, each RC unit learns each weight parameter in a manner that correlates the prediction result data obtained by calculation based on weight parameters with the time-series sensor data b measured by sensor b. ・During the prediction period, each RC unit calculates the memory feature quantity to be output by the memory feature quantity generation unit based on the learned weight parameters, and outputs the prediction result data. • It compares the final prediction results with the time-series sensor data b obtained during the substrate manufacturing process to determine the status of the substrate manufacturing process.

[0093] Accordingly, the substrate processing apparatus 100 of the third embodiment can output prediction result data with higher accuracy while enjoying the same effect as the first embodiment.

[0094] [Fourth Embodiment] In the third embodiment described above, the case where the complex array RC units of the data processing apparatus 1000 are connected in series was described. In contrast, in the fourth embodiment, the case where the complex array RC units are connected in parallel will be described. Hereinafter, the fourth embodiment will be described focusing on its differences from the third embodiment described above.

[0095] <Functional Configuration of the RC Section and Integration Section of the Data Processing Device> First, the functional configuration of the RC section and integration section of the data processing device 1100 included in the substrate processing device 100 of the fourth embodiment will be explained. Figure 11 is a diagram showing an example of the functional configuration of the RC section and integration section of the data processing device.

[0096] The difference between the functional configuration shown in Figures 10A and 10B is that, in the case of Figure 11, • a plurality of RC sections 1110_1 to 1010_L are connected in parallel; • the integration section 1130 is connected in series with respect to the plurality of RC sections 1110_1 to 1010_L. Furthermore, in the case of Figure 11, the integration unit 1130 includes a prediction unit 1140 (an example of an integrated prediction unit), a period determination unit 1150, a switching unit 1160, and a learning unit 1170 (an example of an integrated learning unit); the prediction unit 1140 further includes a weight acquisition unit 1141 and a calculation unit 1142. The weight acquisition unit 1141 learns from the learning unit 1170, acquires the weight parameters stored in the data storage unit 1180, and sets them in the calculation unit 1142; the calculation unit 1142 uses the weight parameters set by the weight acquisition unit 1141 to weight and sum the prediction result data output by the prediction units 1120 of each RC unit 1110_1 to 1110_L, and notifies the period determination unit 1150 of the prediction result data. • During the learning period, the switching unit 1160 notifies the learning unit 1170 and the prediction units 1120 of each RC unit 1110_1 to 1110_L of the prediction result data notified by the period determination unit 1150; • During the prediction period, the switching unit 1160 notifies the acquisition unit 132 and the prediction units 1120 of each RC unit 1110_1 to 1110_L of the prediction result data notified by the period determination unit 1150 as the final prediction result data; • The learning unit 1170 has a weight calculation unit 1171; • During the learning period, the weight calculation unit 1171 calculates the error between the sensor data b measured by the sensor b140 and notified by the acquisition unit 132 as the correct solution data and the prediction result data notified by the switching unit 1160; • During the learning period, the weight calculation unit 1171 learns weight parameters based on the calculated error and stores the learned weight parameters in the data storage unit 1180.Furthermore, in the case of Figure 11: • The time-series sensor data measured by sensor a120 is input in parallel to the memory feature generation units 300_1 to 300_L of each RC unit 1110_1 to 1110_L, and the memory features are output respectively; • During the learning period, the corresponding prediction unit 1120 of each RC unit 1110_1 to 1110_L calculates the memory features based on weight parameters, and outputs the calculated prediction results after correction based on the prediction results notified by the integration unit 1130; • During the prediction period, the corresponding prediction unit 1120 of each RC unit 1110_1 to 1110_L calculates the memory features based on weight parameters, and outputs the calculated prediction results after correction based on the prediction results notified by the integration unit 1130. • During the learning and prediction periods, the prediction result data output from the corresponding prediction units 1120 of each RC unit 1110_1 to 1110_L is input to the integration unit 1130; • During the learning period, the corresponding switching units 1020 of each RC unit 1110_1 to 1110_L are in the ON state, and the prediction result data output from the corresponding prediction unit 1120 is input to the corresponding learning unit 340 for learning.

[0097] <Summary> As can be clearly understood from the above description, regarding the substrate processing apparatus 100 of the fourth embodiment: ・It has a plurality of RC units connected in parallel. Each RC unit inputs in parallel the time-series sensor data a (the time-series sensor data a measured by sensor a) obtained in the substrate manufacturing process, and outputs its own prediction result data to the integration unit. ・During the prediction period, the prediction result data output from the plurality of RC units is integrated in the integration unit based on weight parameters, and output as the final prediction result data. ・During the learning period, each RC unit learns each weight parameter by associating each prediction result data obtained by calculation based on weight parameters with the time-series sensor data b measured by sensor b. ・During the learning period and the prediction period, the prediction result data integrated in the integration unit is input to the prediction unit of each RC unit, and the stored characteristic values ​​are calculated based on the weight parameters to correct the prediction result data obtained therefrom. • During the learning period, the integration unit learns weight parameters by associating the integrated prediction results with the time-series sensor data b measured by sensor b. • During the prediction period, the final prediction results are compared with the time-series sensor data b obtained in the substrate manufacturing process to determine the status of the substrate manufacturing process.

[0098] Accordingly, the substrate processing apparatus 100 of the third embodiment can output prediction result data with higher accuracy while enjoying the same effect as the first embodiment.

[0099] [Fifth Embodiment] In the first to fourth embodiments described above, the case of predicting sensor data of a time series measured by a single sensor b140 based on sensor data of a time series measured by a single sensor a120 was explained. In contrast, in the fifth embodiment, the sensor data of a time series measured by a single sensor b140 is predicted based on sensor data of a time series measured by a plurality of sensors. Hereinafter, the fifth embodiment will be described focusing on its differences from the first embodiment described above.

[0100] <System Configuration of the Substrate Processing Apparatus> First, the system configuration of the substrate processing apparatus according to the fifth embodiment will be described. FIG12 is the second figure showing an example of the system configuration of the substrate processing apparatus. The difference between it and the system configuration shown in FIG1 is that, in the case of FIG12, sensors a1120_1 to a3120_3 (another example of the first sensor) are arranged instead of sensor a120, and the functions of RC unit 1210 and RC unit 131 are different.

[0101] Furthermore, similar to FIG1, FIG12(a) shows the data flow within the substrate processing apparatus 1200 during the learning period, and FIG12(b) shows the data flow within the substrate processing apparatus 1200 during the prediction period.

[0102] As shown in Figures 12(a) and (b), during substrate processing in chamber 110, time-series sensor data obtained by sensors a1120_1 to a3120_3 (another example of the first sensor data) is sent to RC unit 1210.

[0103] The RC unit 1210 performs memory calculations with higher expressive power and faster learning speed. Specifically, the RC unit 1210 predicts the sensor data of the time series measured by the sensor b140 based on the sensor data (input data) of the time series measured by the sensor a1120_1 to the sensor a3120_3.

[0104] Therefore, as shown in FIG12(a), during the learning period, the RC unit 1210 acquires the time series sensor data (forward solution data) measured by the sensor b140 via the acquisition unit 132 of the data processing device 130. Then, the RC unit 1210 learns the weight parameters by associating the prediction result data obtained by calculating the time series sensor data (input data) measured by the sensor a1120_1 to the sensor a3120_3 based on the weight parameters with the forward solution data.

[0105] On the other hand, as shown in FIG12(b), during the prediction period, the RC unit 1210 calculates the time series sensor data (input data) measured by sensors a1120_1 to a3120_3 based on the learned weight parameters. Thereby, the RC unit 1210 outputs the time series sensor data (prediction result data). Furthermore, the time series sensor data (prediction result data) output by the RC unit 1210 is notified to the acquisition unit 132 of the data processing device 130.

[0106] <Functional Configuration of the RC Section of the Data Processing Apparatus> Next, the functional configuration of the RC section 1210 in the data processing apparatus 130 of the substrate processing apparatus 1200 of the fifth embodiment will be described. Figure 13 is the fifth figure showing an example of the functional configuration of the RC section of the data processing apparatus.

[0107] The difference between its functional configuration and that of Figure 1 in Figure 3 is that, in the case of Figure 13, the time series sensor data a1 to a3 obtained by sensors a1120_1 to sensors a3120_3 are input to the storage feature quantity generation unit 300.

[0108] Accordingly, the arithmetic unit 312 can input the stored feature values ​​of the time series sensor data a1 to sensor data a3 obtained by the plurality of sensors a1120_1 to sensors a3120_3 respectively, and output the prediction result data.

[0109] <Summary> As can be clearly understood from the above description, regarding the substrate processing apparatus 100 of the fifth embodiment: ・It has a memory feature generation unit that inputs a plurality of time-series sensor data a1 to a3 (a plurality of time-series sensor data a1 to a3 measured by a plurality of sensors a1 to a3) acquired in the substrate manufacturing process, and outputs memory feature quantities. ・During the learning period, it learns weight parameters by associating the prediction result data obtained by calculating the memory feature quantities to be output by the memory feature generation unit based on weight parameters with the time-series sensor data b measured by sensor b. ・During the prediction period, it calculates the memory feature quantities to be output by the memory feature generation unit based on the learned weight parameters using the plurality of time-series sensor data a1 to a3 acquired in the substrate manufacturing process, and outputs prediction result data. • It compares the predicted data with the time-series sensor data b obtained in the substrate manufacturing process to determine the status of the substrate manufacturing process.

[0110] Accordingly, according to the fifth embodiment, the substrate processing apparatus that uses sensor data of a plurality of time series to determine the state of the substrate manufacturing process can enjoy the same effects as the first embodiment described above.

[0111] [Sixth Embodiment] In the above embodiments, the case of predicting the sensor data b of the time series measured by sensor b140 based on the sensor data of the time series measured by sensor a120 (or sensor a1120_1 to sensor a3120_3) has been described.

[0112] In contrast, in the sixth embodiment, the sensor data of the time series measured by the sensor a120 is predicted based on the sensor data of the time series measured by the sensor a120. Hereinafter, the sixth embodiment will be described with a focus on its differences from the embodiments described above.

[0113] <System Configuration of the Substrate Processing Apparatus> First, the system configuration of the substrate processing apparatus according to the sixth embodiment will be described. Figure 14 is the third figure showing an example of the system configuration of the substrate processing apparatus. The difference between it and the system configuration described using Figure 1 in the first embodiment is that the sensor b140 is not included in the case of Figure 14.

[0114] Furthermore, similar to FIG1, FIG14(a) shows the data flow within the substrate processing apparatus 1400 during the learning period, and FIG14(b) shows the data flow within the substrate processing apparatus 1400 during the prediction period.

[0115] As shown in Figures 14(a) and (b), during substrate processing in chamber 110, time-series sensor data measured by sensor a120 is sent to RC unit 131 and acquisition unit 132. That is, the time-series sensor data measured by sensor a120 is one example of the first time-series sensor data acquired in the substrate manufacturing process, and another example of the second sensor data.

[0116] The RC unit 131 performs memory calculations with higher expressive power and faster learning speed. Specifically, the RC unit 131 predicts the sensor data of the time series measured by the sensor a120 based on the sensor data (input data) of the time series measured by the sensor a120.

[0117] Therefore, as shown in FIG14(a), during the learning period, the RC unit 131 acquires the sensor data (forward solution data) of the time series measured by the sensor a120 via the acquisition unit 132 of the data processing device 130. Then, the RC unit 131 learns the weight parameter by associating the prediction result data obtained by calculating the sensor data (input data) of the time series measured by the sensor a120 based on the weight parameter with the forward solution data.

[0118] On the other hand, as shown in FIG14(b), during the prediction period, the RC unit 131 calculates the time series sensor data (input data) measured by the sensor a120 based on the learned weight parameters. Thereby, the RC unit 131 outputs the time series sensor data (prediction result data). Furthermore, the time series sensor data (prediction result data) output by the RC unit 131 is notified to the acquisition unit 132 of the data processing device 130.

[0119] Furthermore, as shown in FIG14(b), during the prediction period, the acquisition unit 132 notifies the determination unit 133 of the time series sensor data (prediction result data) output by the RC unit 131 and the time series sensor data a (comparison data) measured by the sensor a120. Moreover, the time series sensor data a measured by the sensor a120 and sent to the acquisition unit 132 of the data processing device 130 is called "correct solution data" during the learning period and "comparison data" during the prediction period.

[0120] <Summary> As can be clearly understood from the above description, regarding the substrate processing apparatus 100 of the sixth embodiment: ・It has a memory feature quantity generation unit that inputs time-series sensor data a (time-series sensor data a measured by sensor a) acquired during the substrate manufacturing process and outputs memory feature quantities. ・During the learning period, it learns weight parameters by relating the prediction result data obtained by calculating the memory feature quantities output by the memory feature quantity generation unit based on weight parameters to the time-series sensor data a measured by sensor a. ・During the prediction period, it calculates the memory feature quantities to be output by the memory feature quantity generation unit based on the learned weight parameters using the time-series sensor data a acquired during the substrate manufacturing process and outputs prediction result data. ・It compares the prediction result data with the time-series sensor data a acquired during the substrate manufacturing process to determine the state of the substrate manufacturing process.

[0121] Therefore, the substrate processing apparatus 100 according to the sixth embodiment can enjoy the same effects as the first embodiment described above.

[0122] [Other Embodiments] In the above embodiments, the function of inputting time-series sensor data obtained in the substrate manufacturing process and outputting memory feature quantities is described as being realized by the memory feature quantity generation unit.

[0123] However, the function of inputting time-series sensor data obtained in the substrate manufacturing process and outputting memory feature values ​​can also be realized by the memory feature value generator 205.

[0124] Therefore, in this specification, the memory feature generation unit includes the following two functions: a function implemented by causing the processor 201 to execute a data processing program, and a function implemented by executing the memory feature generator 205.

[0125] Furthermore, in the above embodiments, the memory feature generator 205 is described as being implemented by an FPGA board or the like. When implemented by an FPGA board or the like, it may also be configured to include a specific physical memory.

[0126] In the above embodiments, the method of predicting sensor data of a time series measured by other sensors (or sensor data of a time series measured by this sensor) based on sensor data of a time series measured by one or more sensors has been described. Furthermore, in the above embodiments, the method of determining the state of the substrate manufacturing process based on the difference between the predicted result data and the comparison data has been described. However, the method for determining the state of the substrate manufacturing process is not limited to this; for example, it may be configured to directly predict the state of the substrate manufacturing process based on sensor data of a time series measured by one or more sensors. In this case, data representing the state of the substrate manufacturing process is input to the learning unit as correct solution data.

[0127] Furthermore, although the learning method for the learning unit to learn the weight parameters is not mentioned in the above embodiments, the learning method for the learning unit to learn the weight parameters is arbitrary. For example, the learning unit may also learn using learning methods such as iterative least squares method and RLS (Recursive Least Squares method). Alternatively, the learning unit may also learn using learning methods such as Bayesian linear regression, Kalman filter, information filter, and particle filter.

[0128] Furthermore, in the above-described second to fourth embodiments, examples of using a plurality of memory feature quantity generating units 300_1 to 300_L have been described, but the plurality of memory feature quantity generating units may be identical to each other or may be different.

[0129] Furthermore, in the fifth embodiment described above, the case of predicting sensor data b of a time series measured by a single sensor b140 based on sensor data a1 to a3 of a time series measured by three sensors a1120_1 to a3120_3 is explained. However, the number of sensor data used for prediction is not limited to three; for example, it can be n sensor data measured by n sensors (n is any integer greater than 2) belonging to the first sensor. Also, the number of sensor data of the time series to be predicted is not limited to one; for example, it can be m sensor data measured by m sensors (m is any integer greater than 1) belonging to the second sensor.

[0130] Furthermore, the present invention is not limited to the configurations exemplified in the above embodiments, or combinations with other elements shown herein. In such respect, modifications may be made without departing from the spirit of the present invention, depending on the application. [Simplified Explanation of the Diagram]

[0010] Figure 1(a) and (b) are the first figures showing an example of the system configuration of a substrate processing device. Figure 2(a) and (b) are figures showing an example of the hardware configuration of a data processing device. Figure 3 is the first figure showing an example of the functional configuration of the RC section of a data processing device. Figure 4 is a flowchart showing the process of memory calculation and processing. Figure 5 is a figure showing an example of the functional configuration of the acquisition section and the determination section of a data processing device. Figure 6 is the first flowchart showing the process of acquisition and determination processing. Figure 7 is the second flowchart showing the process of acquisition and determination processing. Figure 8 is a specific example showing the acquisition and determination processing. Figure 9 is the second figure showing an example of the functional configuration of the RC section of a data processing device. Figure 10A is the third figure showing an example of the functional configuration of the RC section of a data processing device. Figure 10B is the fourth figure showing an example of the functional configuration of the RC section of a data processing device. Figure 11 is a figure showing an example of the functional configuration of the RC section and the integration section of a data processing device. Figure 12(a) and (b) are the second figures showing an example of the system configuration of a substrate processing device. Figure 13 is the fifth figure showing an example of the functional configuration of the RC section of a data processing device. Figure 14(a) and (b) are the third figures showing an example of the system configuration of a processing device.

Claims

1. A substrate processing apparatus comprising: a memory feature generation unit that inputs first sensor data of a time series acquired during a substrate manufacturing process and outputs memory feature data; a learning unit that, during a learning period, learns the weighting parameters by relating prediction result data obtained by calculating the memory feature data output by the memory feature generation unit based on weighting parameters to second sensor data of a time series acquired during the substrate manufacturing process; a prediction unit that, during a prediction period, calculates the memory feature data to be output by the memory feature generation unit based on the weighting parameters learned by the learning unit using the first sensor data of the time series acquired during the substrate manufacturing process and the data input by the first sensor data of the time series acquired during the substrate manufacturing process, and outputs prediction result data; and a determination unit that, during the prediction period, determines the state of the substrate manufacturing process by comparing the prediction result data with the second sensor data of the time series acquired during the substrate manufacturing process.

2. The substrate processing apparatus of claim 1, comprising a plurality of memory feature generation units that input first sensor data of time series acquired in the substrate manufacturing process in parallel and output memory feature values ​​respectively; the learning unit learning a weighting parameter by associating prediction result data obtained by weighted summation of the memory feature values ​​output by the plurality of memory feature generation units with second sensor data of time series acquired in the substrate manufacturing process using a weighting parameter; the prediction unit outputting prediction result data by weighted summation of the first sensor data of time series acquired in the substrate manufacturing process using the weighting parameter learned by the learning unit, and using the weighted summation of the memory feature values ​​output by the plurality of memory feature generation units.

3. The substrate processing apparatus of claim 1, comprising a plurality of memory feature generation units connected in series, including a starting memory feature generation unit and subsequent memory feature generation units, wherein the starting memory feature generation unit is input with first sensor data of time series acquired in the substrate manufacturing process, and the subsequent memory feature generation units are input with prediction result data obtained by calculating the memory feature data output by the previous memory feature generation unit based on weight parameters; during the learning period, each learning unit corresponding to the plurality of memory feature generation units learns corresponding weight parameters by associating the prediction result data obtained by calculating the memory feature data output by the plurality of memory feature generation units based on weight parameters with second sensor data of time series acquired in the substrate manufacturing process. During the aforementioned prediction period, each prediction unit corresponding to the plurality of memory feature quantity generation units calculates the memory feature quantities output by the plurality of memory feature quantity generation units based on the weight parameters learned by the corresponding learning unit, and outputs its own prediction result data. During the aforementioned prediction period, the determination unit determines the state of the aforementioned substrate manufacturing process by comparing the final prediction result data output by the prediction unit corresponding to the terminal's memory feature quantity generation unit with the second sensor data of the time series obtained in the aforementioned substrate manufacturing process.

4. The substrate processing apparatus of claim 1, comprising: a plurality of memory feature generation units that input first sensor data of time series acquired during the substrate manufacturing process in parallel and output memory features respectively; a plurality of learning units that, during the learning period, learn the corresponding weight parameters respectively, such that each prediction result data obtained by calculating the memory features output by the plurality of memory feature generation units based on corresponding weight parameters is correlated with second sensor data of time series acquired during the substrate manufacturing process; and a plurality of prediction units that calculate the memory features output by the plurality of memory feature generation units based on the weight parameters learned by the corresponding learning units and output prediction result data respectively. The integrated learning unit, during the learning period, learns weighting parameters for the weighted summation by associating the final prediction result data obtained by weighting and summing the prediction result data output by the plurality of prediction units with the second sensor data of the time series acquired in the substrate manufacturing process using weighting parameters; and the integrated prediction unit, during the prediction period, uses the weighting parameters learned by the integrated learning unit to weight and sum the prediction result data output by the plurality of prediction units, and outputs the final output prediction result data; during the learning period and the prediction period, the prediction result data output by the plurality of prediction units is the prediction result data obtained by calculating the memory feature data output by the plurality of memory feature data generation units based on the corresponding weighting parameters and correcting it according to the final prediction result data.

5. The substrate processing apparatus of claim 1, comprising a memory feature generation unit that receives first sensor data of a plurality of time series acquired in the substrate manufacturing process and outputs memory feature values, wherein the learning unit learns weight parameters such that prediction result data obtained by calculating the memory feature values ​​output by the memory feature generation unit based on weight parameters is correlated with second sensor data of time series acquired in the substrate manufacturing process, and the prediction unit calculates the memory feature values ​​output by the memory feature generation unit based on the weight parameters learned by the learning unit using the first sensor data of the plurality of time series input, and outputs prediction result data.

6. The substrate processing apparatus of any one of claims 1 to 5, wherein the first sensor data of the time series acquired in the substrate manufacturing process and input to the memory feature quantity generation unit is sensor data of the time series measured by the first sensor, and the second sensor data of the time series acquired in the substrate manufacturing process and used for learning by the learning unit is sensor data of the time series measured by the second sensor, which is different from the first sensor.

7. The substrate processing apparatus of claim 6, wherein the first sensor data of a plurality of time series obtained in the substrate manufacturing process is the sensor data of n time series measured by n sensors belonging to the first sensor (n is an integer of 2 or more), and the second sensor data of the time series obtained in the substrate manufacturing process and used for learning by the learning unit is the sensor data of m time series measured by m sensors belonging to the second sensor different from the first sensor (m is an integer of 1 or more).

8. The substrate processing apparatus of any one of claims 1 to 5, wherein the first sensor data of the time series obtained in the substrate manufacturing process is sensor data of the time series measured by the first sensor, and the second sensor data of the time series obtained in the substrate manufacturing process and used for learning by the learning unit is sensor data of the time series measured by the first sensor.

9. The substrate processing apparatus of claim 6 further comprises: a differential calculation unit that calculates the difference between the second sensor data of the time series measured by the second sensor and the prediction result data output by the prediction unit; and a determination unit that determines the state of the substrate manufacturing process based on the differential data calculated by the differential calculation unit.

10. The substrate processing apparatus of claim 8 further comprises: a differential calculation unit that calculates the difference between the first sensor data of the time series measured by the first sensor and the prediction result data output by the prediction unit; and a determination unit that determines the state of the substrate manufacturing process based on the differential data calculated by the differential calculation unit.

11. The substrate processing apparatus of claim 9, wherein the determination unit determines the status of the substrate manufacturing process by determining whether the differential data exceeds a specific threshold, or by determining whether the differential data within a specific time range is similar to a specific pattern.

12. The substrate processing apparatus of claim 11 controls the actuator based on the determination result of the state of the substrate manufacturing process.

13. A data processing method, which is a data processing method in a substrate processing apparatus, and comprising: a memory feature generation step, which inputs first sensor data of a time series acquired in a substrate manufacturing process and outputs memory feature data; a learning step, which, during a learning period, learns the weight parameters by relating prediction result data obtained by calculating the memory feature data output in the memory feature generation step based on weight parameters with second sensor data of a time series acquired in the substrate manufacturing process; a prediction step, which, during a prediction period, calculates the memory feature data to be output in the memory feature generation step based on the weight parameters learned in the learning step using the first sensor data of the time series acquired in the substrate manufacturing process, and outputs prediction result data; and a determination step, which, during the prediction period, determines the state of the substrate manufacturing process by comparing the prediction result data with the second sensor data of the time series acquired in the substrate manufacturing process.

14. A data processing program for causing a computer of a substrate processing apparatus to execute: a memory feature generation process, which inputs first sensor data of a time series acquired in a substrate manufacturing process and outputs memory features; a learning process, which, during a learning period, learns the weighting parameters by relating prediction result data obtained by calculating the memory features output in the memory feature generation process based on weighting parameters with second sensor data of a time series acquired in the substrate manufacturing process; a prediction process, which, during a prediction period, calculates the memory features to be output in the memory feature generation process based on the weighting parameters learned in the learning process using the first sensor data of the time series acquired in the substrate manufacturing process, and outputs prediction result data; and a determination process, which, during the prediction period, determines the state of the substrate manufacturing process by comparing the prediction result data with the second sensor data of the time series acquired in the substrate manufacturing process.

Images