Substrate processing apparatus, estimation method, method for generating learning data, learning method, learning apparatus, method for generating learned model, and learned model

Abstract

A substrate processing apparatus (100) is provided with a processing chamber (11), a processing first sensor (50), and a first control unit (102). The processing first sensor (50) detects at least one of a wind speed and a wind direction in a processing chamber (11) or at a first position at the time of processing around the processing chamber (11). The first control unit (102) inputs a detection result of the first sensor (50) for processing into the learned model (M), and acquires at least one of an estimated wind speed and an estimated wind direction at the second position during processing. The learned model M is generated by machine learning based on a first detection result of a learning first sensor (50L) at a first position during learning and a second detection result of a learning second sensor (60L) at a second position during learning.

Description

Technical Field

[0001] The present invention relates to a substrate processing device, an estimation method, a method for generating learning data, a learning method, a learning device, a method for generating a learned model, and a learned model. Background Art

[0002] Conventionally, substrate processing apparatuses for processing substrates are known. These apparatuses are suitable for use in the manufacture of semiconductor substrates. These apparatuses process substrates within a chamber using a processing liquid, such as a chemical solution. In such apparatuses, for example, changes in wind speed or direction around the substrates can affect substrate processing. Therefore, one approach is to deploy sensors to monitor wind speed or direction around the substrates. However, if chemical solution discharged from the substrates adheres to the sensors, the sensors degrade, making it impossible to deploy sensors around the substrates. Consequently, it is impossible to detect wind speed or direction around the substrates.

[0003] Here, for example, Patent Document 1 describes a method of estimating the wind speed distribution of the entire room by simulating airflow based on wind speed and temperature information measured by sensors placed indoors.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 2012-63055 Summary of the Invention

[0007] Problems to be solved by the invention

[0008] However, the method described in Patent Document 1 has a problem in that estimation takes time because the entire room is simulated.

[0009] The present invention is completed in view of the above-mentioned problems, and its purpose is to provide a substrate processing device, an estimation method, a method for generating learning data, a learning method, a learning device, a method for generating a learned model, and a learned model that can shorten the time for estimating the wind speed or wind direction at a predetermined position in a chamber.

[0010] Means for solving problems

[0011] According to one embodiment of the present invention, a substrate processing apparatus includes a substrate holding portion, a processing chamber, a first processing sensor, and a control portion. The substrate holding portion holds and rotates a substrate to be processed. The processing chamber accommodates the substrate holding portion. The first processing sensor is disposed within or around the processing chamber. The first processing sensor detects at least one of a wind speed and a wind direction at a first position within or around the processing chamber during processing. The control portion inputs the detection result of the first processing sensor into a learned model and obtains at least one of an estimated wind speed and an estimated wind direction at a second position within the processing chamber during processing. The learned model is generated by machine learning based on a first detection result of the first learning sensor and a second detection result of a second learning sensor, wherein the first learning sensor detects at least one of a wind speed and a wind direction at a first position within or around the learning chamber during training, and the second learning sensor detects at least one of a wind speed and a wind direction at a second position within the learning chamber during training.

[0012] In one embodiment, the substrate processing apparatus includes a cleaning liquid supply unit for supplying cleaning liquid to the processing target substrate. The control unit inputs the detection result of the first processing sensor during cleaning into the learned model to obtain at least one of the estimated wind speed and the estimated wind direction during cleaning.

[0013] In one embodiment, the substrate processing apparatus includes a cup disposed around the substrate holding portion and receiving the processing liquid discharged from the substrate to be processed. The second position during processing is a position on the cup.

[0014] In one embodiment, the learned model is generated by machine learning based on the first detection result and the second detection result expressed in a three-dimensional orthogonal coordinate system. The control unit inputs the detection result of the first processing sensor expressed in the three-dimensional orthogonal coordinate system into the learned model, and obtains at least one of the estimated wind speed and the estimated wind direction expressed in the three-dimensional orthogonal coordinate system.

[0015] In one embodiment, the learned model includes a regression model for each of three coordinates. The control unit inputs the detection results of the first processing sensor represented by a three-dimensional orthogonal coordinate system into each regression model to obtain the estimated wind speed for each of the three coordinates.

[0016] In one embodiment, the substrate processing apparatus includes a display unit that displays at least one of the estimated wind speed and the estimated wind direction.

[0017] In one embodiment, the control unit displays at least one of the acquired estimated wind speed and the estimated wind direction on the display unit using a coordinate system based on each of the second learning sensors.

[0018] In one embodiment, the substrate processing apparatus includes an input unit into which processing conditions for processing the processing target substrate are input, and the processing conditions that can change at least one of the estimated wind speed and the estimated wind direction are input into the input unit.

[0019] In one embodiment, the substrate processing device includes a processing detection unit. The processing detection unit detects at least one of the air pressure in or around the processing chamber, the rotation speed of the processing target substrate, the flow rate or supply pressure of the gas supplied to the processing chamber, and the flow rate or exhaust pressure of the gas discharged from the processing chamber. The learned model is generated by machine learning based on the first detection result, the second detection result, and the detection result of the learning detection unit. The learning detection unit detects at least one of the air pressure in or around the learning chamber, the rotation speed of the learning target substrate, the flow rate or supply pressure of the gas supplied to the learning chamber, and the flow rate or exhaust pressure of the gas discharged from the learning chamber. The control unit inputs the detection result of the first processing sensor and the detection result of the processing detection unit into the learned model to obtain at least one of the estimated wind speed and the estimated wind direction at the second position during the processing.

[0020] According to another aspect of the present invention, an estimation method includes: a step of detecting at least one of wind speed and wind direction at a first position during processing within a processing chamber or around a substrate holding portion that receives, holds, and rotates a substrate to be processed; and a step of inputting the detection result of at least one of the wind speed and wind direction into a learned model to obtain an estimated wind speed and wind direction at a second position during processing within the processing chamber. The learned model is generated through machine learning based on a first detection result of a first learning sensor that detects at least one of the wind speed and wind direction at the first position during training within the learning chamber or around the learning chamber, and a second detection result of a second learning sensor that detects at least one of the wind speed and wind direction at the second position during training within the learning chamber.

[0021] According to another embodiment of the present invention, a method for generating learning data includes: a process of obtaining a first detection result, wherein the first detection result is obtained by detecting at least one of the wind speed and wind direction at a first position during learning in a learning chamber for a substrate holding portion or around the learning chamber; a process of obtaining a second detection result, wherein the second detection result is obtained by detecting at least one of the wind speed and wind direction at a second position during learning in the learning chamber; and a process of storing the first detection result and the second detection result in a storage unit as learning data in association with each other.

[0022] According to another aspect of the present invention, a learning method includes: a step of acquiring learning data generated according to the method for generating learning data; and a step of performing machine learning by inputting the learning data into a learning program.

[0023] According to another aspect of the present invention, a learning device includes: a storage unit that stores learning data generated according to the learning data generation method; and a learning unit that performs machine learning by inputting the learning data into a learning program.

[0024] According to another aspect of the present invention, a method for generating a learned model includes: a step of acquiring learning data generated according to the method for generating learning data; and a step of generating a learned model through machine learning based on the learning data.

[0025] According to another aspect of the present invention, the learned model is generated by machine learning based on the learning data generated according to the above-described method for generating learning data.

[0026] Effects of the Invention

[0027] According to the present invention, a substrate processing device, an estimation method, a method for generating learning data, a learning method, a learning device, a method for generating a learned model, and a learned model can be provided, which can shorten the time for estimating the wind speed or wind direction at a predetermined position in a chamber. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a diagram showing the overall configuration of the substrate processing learning system according to the first embodiment.

[0029] Figure 2 It is a schematic diagram of a substrate processing apparatus according to a first embodiment.

[0030] Figure 3 It is a schematic diagram of the processing unit of the first embodiment.

[0031] Figure 4It is a perspective view schematically showing the processing unit 10 of the substrate processing apparatus according to the first embodiment.

[0032] Figure 5 It is a perspective view schematically showing a processing unit 10L of the substrate processing apparatus according to the first embodiment.

[0033] Figure 6 This is a block diagram showing a portion of the substrate processing apparatus 100 according to the first embodiment.

[0034] Figure 7 This is a flowchart showing a substrate processing method of the substrate processing apparatus according to the first embodiment.

[0035] Figure 8 This is a flowchart illustrating a method for estimating wind speed and wind direction at a second position during processing in a substrate processing apparatus.

[0036] Figure 9 This is a block diagram showing a portion of the substrate processing apparatus 100L according to the first embodiment.

[0037] Figure 10 This is a block diagram showing a learning data generating device.

[0038] Figure 11 This is a diagram showing an example of learning data generated by the learning data generating device.

[0039] Figure 12 This is a flowchart showing the method for generating learning data according to the first embodiment.

[0040] Figure 13 It is a block diagram showing a learning device.

[0041] Figure 14 This is a flowchart showing the learning method and the method for generating a learned model according to the first embodiment.

[0042] Figure 15 3 is a diagram showing a learned model of the second modification.

[0043] Figure 16 FIG. 1 is a diagram showing an example of a coordinate system based on the processing chamber of the third modification.

[0044] Figure 17 This is a diagram showing an example of a coordinate system based on each of the second learning sensors of the third modification.

[0045] Figure 18 This is a diagram showing an example of an image expressing the estimated wind speed and estimated wind direction using a coordinate system based on each of the second learning sensors according to the third modification.

[0046] Figure 19 It is a schematic diagram of a processing unit according to the second embodiment.

[0047] Figure 20 It is a schematic diagram of a processing unit of a substrate processing apparatus according to a second embodiment. DETAILED DESCRIPTION

[0048] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, the present invention is not limited to the following embodiments and can be implemented in various ways without departing from the scope of the present invention. In addition, for parts where the description is repeated, the description may be omitted as appropriate. In addition, in the drawings, the same reference numerals are used for the same or corresponding parts, and the description is not repeated.

[0049] (First embodiment)

[0050] Reference Figures 1 to 14 , a substrate processing learning system 1000 according to a first embodiment of the present invention is described. Figure 1 1 is a diagram showing the overall configuration of a substrate processing training system 1000 according to the first embodiment.

[0051] like Figure 1 As shown, the substrate processing training system 1000 includes a substrate processing apparatus 100 , a substrate processing apparatus 100L, a training data generating apparatus 400 , and a training apparatus 500 .

[0052] The substrate processing apparatus 100 processes a target substrate using a processing liquid. The substrate processing apparatus 100 is a single-wafer type apparatus that processes target substrates one by one. Typically, the target substrate is substantially disk-shaped.

[0053] The substrate processing apparatus 100L processes the learning object substrate using a processing liquid. In addition, the structure of the learning object substrate is, for example, the same as the structure of the processing object substrate. Typically, the learning object substrate is roughly disk-shaped. The substrate processing apparatus 100L is a single-chip device that processes the learning object substrate piece by piece. The structure of the substrate processing apparatus 100L is roughly the same as that of the substrate processing apparatus 100. The substrate processing apparatus 100L may also be the same as the substrate processing apparatus 100. For example, the same substrate processing apparatus may have previously processed the learning object substrate and then processed the processing object substrate. Alternatively, the substrate processing apparatus 100L may also be another product having a structure roughly the same as that of the substrate processing apparatus 100.

[0054] In the following description of this specification, a learning target substrate may be referred to as a "learning target substrate WL," and a processing target substrate may be referred to as a "processing target substrate Wp." Furthermore, when there is no need to distinguish between the learning target substrate WL and the processing target substrate Wp, both the learning target substrate WL and the processing target substrate Wp may be referred to as a "substrate W."

[0055] The substrate W is, for example, a semiconductor wafer, a liquid crystal display substrate, a plasma display substrate, a field emission display (FED) substrate, an optical disk substrate, a magnetic disk substrate, a magneto-optical disk substrate, a photomask substrate, a ceramic substrate, or a solar cell substrate.

[0056] The substrate processing device 100L obtains the first detection result and the second detection result described later by processing the learning object substrate WL. The first detection result and the second detection result are input to the learning data generating device 400. For example, the first detection result and the second detection result can also be input to the learning data generating device 400 through communication. Alternatively, the first detection result and the second detection result can also be input to the learning data generating device 400 via a storage medium. The storage medium can be, for example, an optical disc such as a CD (optical disc) or a storage device such as a USB (universal serial bus) memory. Alternatively, the first detection result and the second detection result can also be input by the user operating an input device such as a keyboard, a mouse or a touch sensor.

[0057] Learning data generating device 400 associates the first detection result (the detection result of first learning sensor 50L, described later) with the second detection result (the detection result of second learning sensor 60L, described later) to generate learning data 434. Specifically, learning data generating device 400 associates information indicating the first detection result detected by first learning sensor 50L and information indicating the second detection result detected by second learning sensor 60L, both detected at substantially the same time.

[0058] The learning device 500 generates the learned model M through machine learning based on the learning data 434. Specifically, the learning device 500 generates the learned model M by inputting the learning data 434 into a learning program.

[0059] The substrate processing apparatus 100 inputs the detection result of the first processing sensor 50 into the learned model M. As a result, the learned model M outputs at least one of an estimated wind speed and an estimated wind direction at a predetermined position (a second processing position P1 described later).

[0060] Below, refer to Figures 2 to 14 , the substrate processing learning system 1000 is further described. First, referring to Figure 2 and Figure 3 The substrate processing apparatus 100 will be further described. Figure 2 Schematic diagram of the substrate processing apparatus 100 according to the first embodiment. Figure 2 1 is a schematic top view of the substrate processing apparatus 100. In the first embodiment, the substrate W is a semiconductor wafer and has a substantially disk shape.

[0061] like Figure 2 As shown, the substrate processing apparatus 100 includes a plurality of processing units 10 , a fluid cabinet 110 , a plurality of fluid tanks 120 , a plurality of load ports LP, an indexer robot IR, a central robot CR, and a control device 101 .

[0062] The loading port LP accommodates a plurality of substrates W in a stacked manner. The indexer robot IR transports the substrates W between the loading port LP and the central robot CR. The central robot CR transports the substrates W between the indexer robot IR and the processing units 10. Each processing unit 10 supplies a processing liquid to the substrate W to process the substrate W. The fluid box 110 accommodates the processing liquid. The processing liquid includes, for example, a chemical solution, a rinse solution, a removal liquid, and / or a waterproofing agent. The fluid box 110 accommodates the processing liquid. In addition, the fluid box 110 may also accommodate gas.

[0063] The plurality of processing units 10 form a plurality of towers TW ( Figure 2 Each tower TW includes a plurality of processing units 10 stacked up and down ( Figure 2 The fluid tanks 120 correspond to the multiple towers TW. The treatment liquid in the fluid cabinet 110 is supplied to all the treatment units 10 included in the tower TW corresponding to the fluid tank 120 through any fluid tank 120.

[0064] The control device 101 controls the operation of each unit of the substrate processing apparatus 100. For example, the control device 101 controls the load port LP, the indexer robot IR, and the central robot CR.

[0065] Next, refer to Figure 3 The processing unit 10 according to the first embodiment will be described. Figure 3 Schematic diagram of the processing unit 10 of the first embodiment. Figure 3 is a schematic cross-sectional view of the processing unit 10 .

[0066] like Figure 3As shown, in the first embodiment, the processing liquid includes, for example, an etching liquid, and the processing unit 10 performs an etching process. The substrate W is treated (etched) by the etching liquid. The etching liquid is a chemical liquid. Examples of the etching liquid include nitric acid (a mixture of hydrofluoric acid (HF) and nitric acid (HNO3)), hydrofluoric acid, buffered hydrofluoric acid (BHF), ammonium fluoride, HPEG (a mixture of hydrofluoric acid and ethylene glycol), phosphoric acid (H3PO4), SC1 (a mixture of ammonia and hydrogen peroxide), SC2 (a mixture of hydrochloric acid and hydrogen peroxide), SPM (a mixture of sulfuric acid and hydrogen peroxide), or ammonia.

[0067] like Figure 3 As shown, the processing unit 10 includes a processing chamber 11, a blower unit 12, a substrate holder 20, a chemical supply unit 30, and a rinse liquid supply unit 60. The rinse liquid supply unit 60 is an example of a "cleaning liquid supply unit" in the present invention.

[0068] The processing chamber 11 is generally box-shaped and has an interior space. The processing chamber 11 accommodates substrates W. Here, the substrate processing apparatus 100 is a single-wafer type that processes substrates W one by one, and each substrate W is accommodated in the processing chamber 11. The substrates W are accommodated and processed within the processing chamber 11. The processing chamber 11 accommodates at least a portion of each of the substrate holder 20, the chemical liquid supply unit 30, and the rinse liquid supply unit 60.

[0069] The air supply unit 12 is disposed on or above the processing chamber 11. For example, the air supply unit 12 is disposed on the top surface of the processing chamber 11. The air supply unit 12 delivers air into the processing chamber 11. The air supply unit 12 includes, for example, a fan filter unit (FFU). The air supply unit 12 and the exhaust device 80 described below form a downward flow in the processing chamber 11.

[0070] The substrate holder 20 holds the substrate W. The substrate holder 20 holds the substrate W horizontally, with its upper surface (front surface) Wa facing upward and its lower surface (back surface) Wb facing vertically downward. Furthermore, the substrate holder 20 rotates the substrate W while holding it. For example, a laminated structure with grooves formed on the upper surface Wa of the substrate W can be provided. The substrate holder 20 rotates the substrate W while holding it.

[0071] For example, the substrate holding portion 20 may be of a clamping type that clamps the end portion of the substrate W. Alternatively, the substrate holding portion 20 may include any mechanism that holds the substrate W from the lower surface Wb. For example, the substrate holding portion 20 may be of a vacuum type. In this case, the substrate holding portion 20 holds the substrate W horizontally by adsorbing the central portion of the lower surface Wb of the substrate W, which is the non-device forming surface, onto the upper surface. Alternatively, the substrate holding portion 20 may be a combination of a clamping type in which a plurality of chuck pins are in contact with the peripheral end surface of the substrate W, and a vacuum type.

[0072] For example, the substrate holding unit 20 includes a spin base 21, a chuck member 22, a shaft 23, a rotation motor 24, and a housing 25. The chuck member 22 is provided on the spin base 21. The chuck member 22 holds the substrate W. Typically, a plurality of chuck members 22 are provided on the spin base 21.

[0073] The shaft 23 extends in the vertical direction along the rotation axis AX. The spin base 21 is coupled to the upper end of the shaft 23. The substrate W is placed above the spin base 21.

[0074] The spin base 21 is disk-shaped. The chuck assembly 22 horizontally supports the substrate W. A shaft 23 extends downward from the center of the spin base 21. A rotary motor 24 applies a rotational force to the shaft 23. The rotary motor 24 rotates the shaft 23 in the rotational direction, thereby rotating the substrate W and the spin base 21 about the rotation axis AX. A housing 25 houses the shaft 23 and the rotary motor 24.

[0075] The chemical supply unit 30 supplies chemical liquid (here, etching liquid) to the substrate W. Typically, the chemical supply unit 30 supplies chemical liquid to the upper surface Wa of the substrate W. At least a portion of the chemical supply unit 30 is housed in the processing chamber 11 .

[0076] The chemical liquid supply unit 30 includes a pipe 32 and a nozzle 36. The nozzle 36 discharges the chemical liquid toward the upper surface Wa of the substrate W. The nozzle 36 is connected to the pipe 32. The chemical liquid is supplied to the pipe 32 from a supply source.

[0077] The nozzle 36 can also be configured to be movable relative to the substrate W. The chemical liquid supply unit 30 can further include a nozzle moving unit 38. The nozzle moving unit 38 can move the nozzle 36 up and down, or can rotate the nozzle 36 horizontally about the rotation axis. The nozzle moving unit 38 moves the nozzle 36 up and down. For example, the nozzle moving unit 38 includes a ball screw mechanism and an electric motor that provides driving force to the ball screw mechanism. In addition, the nozzle moving unit 38 rotates the nozzle 36 horizontally. For example, the nozzle moving unit 38 includes an electric motor.

[0078] The chemical liquid supply unit 30 may also supply a chemical liquid other than the etching liquid to the substrate W. Furthermore, the substrate processing apparatus 100 may include a processing liquid supply unit that supplies a processing liquid other than the chemical liquid instead of the chemical liquid supply unit 30 .

[0079] The rinse liquid supply unit 60 supplies the rinse liquid to the substrate W. Typically, the rinse liquid supply unit 60 supplies the rinse liquid to the upper surface Wa of the substrate W. At least a portion of the rinse liquid supply unit 60 is accommodated in the processing chamber 11 .

[0080] For example, the rinse liquid includes deionized water (DIW), carbonated water, electrolytic ionized water, ozone water, diluted hydrochloric acid water (eg, about 10 ppm to 100 ppm), or reduced water (hydrogen water). In the first embodiment, the rinse liquid is deionized water (DIW).

[0081] The rinse liquid supply unit 60 includes a pipe 62 and a nozzle 66. The nozzle 66 sprays the rinse liquid toward the upper surface Wa of the substrate W. The nozzle 66 is connected to the pipe 62. The rinse liquid is supplied to the pipe 62 from a supply source.

[0082] The nozzle 66 may be configured to be movable relative to the substrate W. The rinse liquid supply unit 60 may further include a nozzle moving unit 68. The nozzle moving unit 68 may move the nozzle 66 upward and downward, or may rotate the nozzle 66 horizontally about the rotation axis. The nozzle moving unit 68 may be configured similarly to the nozzle moving unit 38, for example.

[0083] The processing unit 10 further includes an inert gas supply unit 70 . The type of inert gas is not particularly limited, and nitrogen gas is an example. The inert gas supply unit 70 includes a pipe 72 , a valve 74 , and a nozzle 76 .

[0084] The nozzle 76 ejects an inert gas toward the lower surface Wb of the substrate W. For example, the nozzle 76 ejects the inert gas toward the center of the substrate W. The nozzle 76 has an ejection port, from which the inert gas is ejected. For example, the nozzle 76 ejects the inert gas toward the top or horizontally. The nozzle 76 is connected to the pipe 72. The inert gas is supplied from a supply source to the pipe 72. The valve 74 can adjust the amount of inert gas supplied by adjusting the opening of the flow path of the pipe 72.

[0085] The processing unit 10 also includes an exhaust device 80. The exhaust device 80 includes an exhaust pipe 81 and an exhaust fan 82. The exhaust pipe 81 is configured to penetrate the side wall of the processing chamber 11. The exhaust pipe 81 connects the inside of the processing chamber 11 with the outside. The exhaust fan 82 discharges the air in the processing chamber 11 to the outside of the processing chamber 11 via the exhaust pipe 81. The exhaust fan 82 can change the exhaust volume by changing the rotation speed. The exhaust fan 82 is, for example, configured at the inlet of the exhaust pipe 81, the inside of the exhaust pipe 81, or the outlet of the exhaust pipe 81. In the first embodiment, the exhaust fan 82 is configured at the inside of the exhaust pipe 81. In addition, in the first embodiment, the rotation speed refers to the number of rotations per minute (rpm).

[0086] The processing unit 10 further includes a cup 90. The cup 90 recovers the processing liquid scattered from the substrate W. The cup 90 rises and falls. For example, the cup 90 rises vertically upward to the side of the substrate W while the chemical liquid supply unit 30 and / or the rinse liquid supply unit 60 supplies chemical liquid and / or rinse liquid to the substrate W. In this case, the cup 90 recovers the chemical liquid and / or rinse liquid scattered from the substrate W due to the rotation of the substrate W. In addition, when the period in which the chemical liquid supply unit 30 and / or the rinse liquid supply unit 60 supplies chemical liquid and / or rinse liquid to the substrate W ends, the cup 90 descends vertically downward from the side of the substrate W. In addition, in the first embodiment, the cup 90 has a cup 91 arranged at the outermost periphery and a cup 92 arranged on the inner side of the cup 91.

[0087] The substrate processing apparatus 100 of the first embodiment is suitable for use in the manufacture of semiconductor devices incorporating semiconductors. Typically, a semiconductor device comprises a conductive layer and an insulating layer stacked on a substrate. During semiconductor device manufacturing, the substrate processing apparatus 100 is suitable for cleaning and / or processing (e.g., etching, property modification, etc.) the conductive and / or insulating layers.

[0088] Next, refer to Figure 4 The processing unit 10 of the substrate processing apparatus 100 according to the first embodiment will be further described. Figure 4 It is a perspective view schematically showing the processing unit 10 of the substrate processing apparatus 100 according to the first embodiment.

[0089] like Figure 4 As shown, the processing unit 10 of the substrate processing apparatus 100 includes a first processing sensor 50. The first processing sensor 50 is disposed in or around the processing chamber 11. The first processing sensor 50 detects at least one of wind speed and wind direction at a first position during processing in or around the processing chamber 11. In the first embodiment, the first processing sensor 50 is disposed in the processing chamber 11. In addition, in the first embodiment, the first processing sensor 50 detects both wind speed and wind direction at a first position during processing in the processing chamber 11. The first processing sensor 50 outputs the detection result to the first control unit 102.

[0090] The detection interval of the first processing sensor 50 is not particularly limited, and is, for example, 0.1 sec to 10 sec. In the first embodiment, the first processing sensor 50 performs detection every approximately 1 sec.

[0091] The first processing sensor 50 is not particularly limited and can, for example, detect at least one of three-dimensional wind speed and wind direction. In the first embodiment, the first processing sensor 50 can detect both three-dimensional wind speed and wind direction. The first processing sensor 50 can, for example, include a semiconductor chip for detecting wind speed and wind direction mounted on a polyimide substrate.

[0092] The number of first processing sensors 50 is not particularly limited, but preferably multiple. In the first embodiment, for example, ten first processing sensors 50 are provided. That is, the processing unit 10 includes first processing sensors 50a to 50j. The first processing sensors 50a to 50j are arranged at predetermined positions within the processing chamber 11. The positions where the first processing sensors 50a to 50j are arranged are examples of the "first processing positions" of the present invention.

[0093] Specifically, the processing unit 10 includes a partition 85. The partition 85 divides the space around the cup 90 in the processing chamber 11 into upper and lower parts. The first processing sensors 50a to 50f are, for example, arranged below the air supply unit 12. The first processing sensors 50a to 50f are arranged horizontally apart from each other. The first processing sensors 50 are, for example, arranged in a staggered pattern. In addition, the first processing sensors 50a to 50f are arranged at the top of the processing chamber 11, but may also be arranged between the top of the processing chamber 11 and the bottom surface of the air supply unit 12.

[0094] The first processing sensors 50g to 50j are disposed, for example, on the partition plate 85. The first processing sensors 50g to 50j are disposed outside the opening edge 86 of the partition plate 85. The opening edge 86 is formed into a circular shape surrounding the cup 90. The first processing sensors 50g to 50j are disposed at approximately equal angular intervals in the circumferential direction centered on the cup 90.

[0095] The processing unit 10 has a second processing position P1. The number of second processing positions P1 is not particularly limited, but preferably multiple positions are provided. In the first embodiment, four second processing positions P1 are provided. That is, the processing unit 10 has second processing positions P1a to P1d.

[0096] The second positions P1a to P1d during processing are, for example, positions on the periphery of the substrate W or the substrate holding portion 20. The second positions P1a to P1d during processing are, for example, positions that the processing liquid discharged from the substrate W can reach. In the first embodiment, the second positions P1a to P1d during processing are positions on the cup 90. The second positions P1a to P1d during processing are, for example, positions on the outermost cup 91 among the cups 90. In addition, the second positions P1a to P1d during processing are positions outside the opening edge 91a of the outermost cup 91. The opening edge 91a is formed in a circular shape surrounding the substrate W. The second positions P1a to P1d during processing are positions on the inclined surface 91b that is inclined downward from the highest position in the outermost cup 91 toward the radial outside.

[0097] The second processing positions P1a to P1d are arranged at approximately equal angular intervals in the circumferential direction centered on the substrate W. In the first embodiment, the first processing sensors 50g to 50j and the second processing positions P1a to P1d are alternately arranged in the circumferential direction centered on the substrate W. Furthermore, in the first embodiment, the first processing sensors 50g to 50j and the second processing positions P1a to P1d are arranged at approximately 45° intervals in the circumferential direction centered on the substrate W.

[0098] Next, refer to Figure 5 , the substrate processing apparatus 100L according to the first embodiment will be further described. Figure 5 1 is a perspective view schematically showing a processing unit 10L of a substrate processing apparatus 100L according to the first embodiment. Figures 2 to 4 The substrate processing apparatus 100 described above has substantially the same structure, so the different structures will be mainly described. The substrate processing apparatus 100L includes a processing unit 10L.

[0099] like Figure 5 As shown, the processing unit 10L of the substrate processing apparatus 100L includes a training chamber 11L. The training chamber 11L has the same structure as the processing chamber 11 of the processing unit 10.

[0100] The processing unit 10L includes a first learning sensor 50L. The first learning sensor 50L has the same structure as the first processing sensor 50 of the processing unit 10. The first learning sensor 50L is arranged in the learning chamber 11L or around the learning chamber 11L. The first learning sensor 50L detects at least one of the wind speed and wind direction at the first learning position in the learning chamber 11L or around the learning chamber 11L. In the first embodiment, the first learning sensor 50L is arranged in the learning chamber 11L. In addition, in the first embodiment, the first learning sensor 50L detects both the wind speed and wind direction at the first learning position in the learning chamber 11L. The first learning sensor 50L outputs the detection result to the second control unit 102L described later. The first learning position in the learning chamber 11L is substantially the same position as the first processing position in the processing chamber 11.

[0101] The number of first learning sensors 50L is the same as the number of first processing sensors 50. In the first embodiment, for example, ten first learning sensors 50L are provided. That is, the processing unit 10L includes first learning sensors 50La to 50Lj. The first learning sensors 50La to 50Lj are located at positions corresponding to the first processing sensors 50a to 50j. The positions where the first learning sensors 50La to 50Lj are located are examples of the "first learning positions" of the present invention.

[0102] Here, in the first embodiment, the processing unit 10L includes a second learning sensor 60L, unlike the processing unit 10. The second learning sensor 60L has the same structure as the first processing sensor 50 and the first learning sensor 50L, for example.

[0103] The second learning sensor 60L is disposed in the learning chamber 11L. The second learning sensor 60L detects at least one of the wind speed and the wind direction at the second learning position in the learning chamber 11L. In the first embodiment, the second learning sensor 60L detects both the wind speed and the wind direction at the second learning position in the learning chamber 11L. The second learning sensor 60L outputs the detection result to the second control unit 102L. The second learning position in the learning chamber 11L is the same as the second processing position P1 in the processing chamber 11L (refer to Figure 4 ) are approximately in the same location.

[0104] Specifically, the number of second learning sensors 60L is the same as the number of second processing positions P1. In the first embodiment, four second learning sensors 60L are provided. That is, the processing unit 10L includes second learning sensors 60La to 60Ld. The second learning sensors 60La to 60Ld are arranged at the second learning positions.

[0105] Next, refer to Figure 6 The substrate processing apparatus 100 will be further described. Figure 6 1 is a block diagram showing a portion of the substrate processing apparatus 100 according to the first embodiment. Figure 6 As shown, the control device 101 includes a first control unit 102 and a storage unit 103. The substrate processing apparatus 100 further includes a display unit 104, an input unit 105, and a communication unit 106. The first control unit 102 is an example of the "control unit" of the present invention.

[0106] The first control unit 102 includes a processor. For example, the first control unit 102 includes a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). Alternatively, the first control unit 102 may include a general-purpose arithmetic unit or a dedicated arithmetic unit. The first control unit 102 may also include an NPU (Neural Network Processing Unit).

[0107] The storage unit 103 stores data and computer programs. The storage unit 103 includes a primary storage device. The primary storage device is, for example, a semiconductor memory. The storage unit 103 may also include an auxiliary storage device. The auxiliary storage device is, for example, a semiconductor memory and / or a hard disk drive. The storage unit 103 may also include removable media. The first control unit 102 controls the operation of each unit of the substrate processing apparatus 100 based on the data and computer programs stored in the storage unit 103.

[0108] Specifically, the storage unit 103 stores a recipe 131 and a control program 132. The recipe 131 specifies the processing details and processing sequence for the substrates W. Furthermore, the recipe 131 indicates processing conditions and various set values. In the first embodiment, the processing conditions include, for example, the rotational speed of the blower fan built into the blower unit 12, the rotational speed of the exhaust fan 82, the rotational speed of the substrates W, and the opening degree of the valve 74 provided in the pipe 72 that supplies nitrogen gas to the nozzle 76.

[0109] The first control unit 102 controls the operation of each unit of the substrate processing apparatus 100 based on the recipe 131 and the control program 132. In the first embodiment, the first control unit 102 causes the display unit 104 to display an image based on the estimated wind speed and estimated wind direction described later.

[0110] The storage unit 103 stores the learned model M. The learned model M is stored in the learning device 500 (see Figure 2 ) is generated, it is sent to the substrate processing apparatus 100 and stored.

[0111] The learned model M is a model for estimating at least one of the wind speed and wind direction at the second position P1 during processing based on at least one of the wind speed and wind direction at the first position of the processing chamber 11 during processing. At the second position P1 during processing, the processing liquid discharged from the substrate W scatters, causing the sensor to degrade. Therefore, it is impossible to configure a sensor for detecting wind speed, wind direction, etc. at the second position P1 during processing. Therefore, in the first embodiment, a processing unit 10L having the same structure as the processing unit 10 is used to obtain information indicating the wind speed and wind direction at the first position during learning and the wind speed and wind direction at the second position during learning. Then, using the obtained information, a model (learned model M) is generated to estimate at least one of the wind speed and wind direction at the second position P1 during processing.

[0112] Specifically, the learned model M outputs at least one of the estimated wind speed and estimated wind direction at the second position P1 during processing based on the detection results of the first processing sensor 50 as input data. In the first embodiment, the learned model M outputs both the estimated wind speed and estimated wind direction at the second position P1 during processing based on the detection results of the first processing sensor 50 as input data.

[0113] The storage unit 103 stores estimated wind speed information 133 a and estimated wind direction information 133 b indicating the estimated wind speed and estimated wind direction at the second position P1 during processing.

[0114] In addition, the method of generating the learned model M will be described later.

[0115] The display unit 104 displays various information. In the first embodiment, the display unit 104 displays, for example, various setting screens (input screens). Furthermore, the display unit 104 displays, for example, images based on the estimated wind speed and estimated wind direction described later. The display unit 104 includes, for example, a liquid crystal display or an organic EL (electroluminescence) display.

[0116] The input unit 105 receives input from the user. For example, the input unit 105 receives commands to execute or stop processing on a substrate W, change processing conditions and various set values, and the like. The input unit 105 includes, for example, a touch panel and a pointing device. The touch panel is, for example, located on the display surface of the display unit 104. The input unit 105 and the display unit 104 constitute, for example, a graphical user interface.

[0117] The communication unit 106 is connected to a network and communicates with external devices. Examples of networks include the Internet, a LAN (Local Area Network), a public telephone network, and a short-range wireless network. The communication unit 106 is a communication device, such as a network interface controller. In the first embodiment, the communication unit 106 communicates with the learning device 500. The communication unit 106 receives the learned model M from the learning device 500. The first control unit 102 causes the storage unit 103 to store the learned model M received by the communication unit 106.

[0118] Above, refer to Figures 1 to 6 The substrate processing apparatus 100 is described. In the first embodiment, as described above, the substrate processing apparatus 100 includes: a first processing sensor 50 that detects at least one of the wind speed and wind direction at the first position during processing; and a first control unit 102 that inputs the detection result of the first processing sensor 50 to the learned model M to obtain at least one of the estimated wind speed and estimated wind direction at the second position P1 during processing of the processing chamber 11. Therefore, based on at least one of the wind speed and wind direction at the first position during processing during substrate processing, at least one of the wind speed and wind direction at the second position P1 during processing can be estimated. That is, based on at least one of the wind speed and wind direction at another position (the first position during processing), at least one of the wind speed and wind direction at a position (the second position P1 during processing) where a sensor cannot be configured can be estimated.

[0119] Furthermore, in substrate processing apparatus 100, at least one of the wind speed and wind direction is calculated only for a specific position (second processing position P1) within processing chamber 11. This reduces estimation time compared to, for example, calculating at least one of the wind speed and wind direction for the entire space within the chamber. Consequently, for example, substrate processing can be performed while monitoring at least one of the wind speed and wind direction at second processing position P1.

[0120] Furthermore, as described above, the first control unit 102 inputs the detection results of the first processing sensor 50 during cleaning into the learned model M, thereby obtaining at least one of the estimated wind speed and estimated wind direction during cleaning. During cleaning, for example, the wind speed or wind direction around the substrate holder 20 may affect the cleanliness and / or dryness. In other words, the cleanliness and / or dryness may be reduced. Therefore, as described above, by obtaining at least one of the estimated wind speed and estimated wind direction at the second processing position P1 during cleaning, for example, at least one of the wind speed and wind direction around the substrate W can be estimated, thereby suppressing any adverse effects on the cleanliness and / or dryness.

[0121] As described above, the second processing position P1 is a position on the cup 90. Therefore, by estimating at least one of the wind speed and wind direction at the second processing position P1, at least one of the wind speed and wind direction around the substrate W can be easily estimated.

[0122] As described above, the display unit 104 displays at least one of the estimated wind speed and the estimated wind direction. Therefore, the user can confirm at least one of the estimated wind speed and the estimated wind direction at the second position P1 during processing simply by looking at the display unit 104.

[0123] As described above, the user can input into input unit 105 a process condition that changes at least one of the estimated wind speed and estimated wind direction. Thus, for example, a user viewing display unit 104 can change the process condition so that the estimated wind speed and estimated wind direction at second position P1 during processing fall within a predetermined range. Furthermore, the user can also stop processing.

[0124] Next, refer to Figure 7 , a substrate processing method of the substrate processing apparatus 100 is described. Figure 7 This is a flowchart illustrating a substrate processing method in the substrate processing apparatus 100 according to the first embodiment. The substrate processing method in the substrate processing apparatus 100 includes at least step S104 and step S105. In the first embodiment, the substrate processing method includes steps S101 to S107. Steps S101 to S107 are executed by the first control unit 102.

[0125] like Figure 7 As shown, in step S101, a target substrate Wp is loaded into the processing chamber 11. Under the control of the first control unit 102, the central robot CR loads the target substrate Wp into the processing chamber 11, and the substrate holder 20 holds the loaded target substrate Wp.

[0126] Next, in step S102 , the rotation of the target substrate Wp is started. Under the control of the first control unit 102 , the substrate holding unit 20 starts rotating the target substrate Wp while holding the target substrate Wp.

[0127] Next, in step S103, a chemical liquid is supplied to the target substrate Wp for processing. Under the control of the first control unit 102, the chemical liquid supply unit 30 supplies the chemical liquid from the nozzle 36 to the target substrate Wp being rotated by the substrate holder 20. The first control unit 102 then stops supplying the chemical liquid after a predetermined time has elapsed since the start of the chemical liquid supply.

[0128] Next, in step S104, a rinse liquid is supplied to the target substrate Wp for rinsing. Under the control of the first control unit 102, the rinse liquid supply unit 60 supplies the rinse liquid from the nozzle 66 to the target substrate Wp. For example, the rinse liquid supply unit 60 supplies DIW as the rinse liquid. Then, when a predetermined time has passed since the start of the rinse liquid supply, the first control unit 102 stops supplying the rinse liquid.

[0129] Next, in step S105 , the target substrate Wp is dried. Under the control of the first control unit 102 , the substrate holding unit 20 increases the rotation speed of the target substrate Wp, and blows off the rinse liquid on the target substrate Wp by centrifugal force.

[0130] Next, in step S106 , the rotation of the processing target substrate Wp is stopped. Under the control of the first control unit 102 , the substrate holding unit 20 stops the rotation of the processing target substrate Wp.

[0131] Next, in step S107, the target substrate Wp is removed from the processing unit 10. Under the control of the first control unit 102, the substrate holder 20 releases the substrate Wp from its position, and the central robot CR removes the target substrate Wp from the processing chamber 11. The target substrate Wp is then transported to the outside of the substrate processing apparatus 100 via the indexer robot IR.

[0132] As described above, the processing of the processing target substrate Wp is completed.

[0133] Next, refer to Figure 8 , a method for estimating the wind speed and wind direction at the second position P1 during processing in the substrate processing apparatus 100 will be described. Figure 8 This flowchart illustrates a method for estimating wind speed and direction at the second position P1 during processing in the substrate processing apparatus 100. The method for estimating wind speed and direction at the second position P1 during processing in the substrate processing apparatus 100 includes at least steps S201 and S202. In the first embodiment, the method for estimating wind speed and direction in the substrate processing apparatus 100 includes steps S201 through S205. Steps S201 through S205 are executed by the first control unit 102.

[0134] In addition, the method for estimating the wind speed and wind direction at the second position P1 during processing in the substrate processing apparatus 100 is similar to the method for estimating the wind speed and wind direction at the second position P1 during processing in the substrate processing apparatus 100. Figure 7 At least the above-mentioned step S104 and step S105 are executed in parallel. In the first embodiment, the estimation method of the substrate processing apparatus 100 is executed in parallel with the above-mentioned steps S102 to S106.

[0135] like Figure 8As shown, in step S201, at least one of the wind speed and wind direction at the first location during processing is detected. In the first embodiment, both the wind speed and wind direction at the first location during processing are detected. Specifically, the first processing sensor 50 detects the wind speed and wind direction and transmits a detection signal indicating the detection result to the first control unit 102. The first control unit 102 receives the detection signal from the first processing sensor 50. In addition, in the first embodiment, as described above, the first processing sensor 50 performs detection every approximately 1 second, and the first control unit 102 receives the detection signal every approximately 1 second.

[0136] Next, in step S202, at least one of the estimated wind speed and estimated wind direction at the second position P1 during processing is obtained. In the first embodiment, both the estimated wind speed and estimated wind direction at the second position P1 during processing are obtained. Specifically, the first control unit 102 inputs the detection results of the first processing sensor 50 into the learned model M to obtain the estimated wind speed and estimated wind direction at the second position P1 during processing. Furthermore, in the first embodiment, the estimated wind speed and estimated wind direction are output from the time the first control unit 102 inputs the detection results of the first processing sensor 50 into the learned model M until a time of 0.1 seconds to 0.5 seconds has passed. Therefore, the display unit 104 can display an image based on the estimated wind speed and estimated wind direction from the time the wind speed and wind direction at the second position P1 during processing change until a time of 0.1 seconds to 0.5 seconds has passed. In other words, the wind speed and wind direction at the second position P1 during processing can be monitored.

[0137] Next, in step S203, at least one of the estimated wind speed and estimated wind direction is displayed. In the first embodiment, both the estimated wind speed and estimated wind direction are displayed. Specifically, the first control unit 102 causes the display unit 104 to display an image based on the acquired estimated wind speed and estimated wind direction. Thus, the display unit 104 displays an image based on the estimated wind speed and estimated wind direction at the second position P1 during processing.

[0138] Next, in step S204 , the first control unit 102 determines whether or not there is a change in the processing conditions. Specifically, the user checks the estimated wind speed and estimated wind direction displayed on the display unit 104 .

[0139] If the estimated wind speed and estimated wind direction displayed on the display unit 104 deviate from the estimated wind speed and estimated wind direction normally displayed during cleaning, or deviate from the predetermined wind speed range and / or predetermined wind direction range, the user operates the input unit 105 to change the processing conditions. In this case, the user can change the processing condition of at least one of the estimated wind speed and estimated wind direction. For example, the user changes at least one of the rotational speed of the air supply fan built into the air supply unit 12, the rotational speed of the exhaust fan 82, and the rotational speed of the substrate Wp to be processed. In this case, the first control unit 102 determines that a change has occurred in the processing conditions. The process then proceeds to step S205.

[0140] On the other hand, if the estimated wind speed and estimated wind direction displayed on the display unit 104 do not deviate from the estimated wind speed and estimated wind direction normally displayed during cleaning, or are within the predetermined wind speed range and predetermined wind direction range, the user does not operate the input unit 4. In other words, the user does not change the processing conditions. In this case, the first control unit 102 determines that there is no change in the processing conditions. Then, the process returns to step S201.

[0141] Next, in step S205, various set values ​​are changed. Specifically, the first control unit 102 changes the set values ​​for the speed of the air supply fan, the speed of the exhaust fan 82, or the speed of the substrate Wp being processed. This changes the estimated wind speed and direction at the second position P1 during processing. During processing, the estimated wind speed and direction at the second position P1, for example, become equal to or approach the estimated wind speed and direction typically displayed during cleaning. The process then returns to step S201.

[0142] The estimation method of the substrate processing apparatus 100 repeats steps S201 to S205. Figure 7 When the substrate processing method shown enters step S107, the processing ends.

[0143] Next, refer to Figure 9 The substrate processing apparatus 100L will be further described. Figure 9 1 is a block diagram showing a portion of the substrate processing apparatus 100L according to the first embodiment. Figure 9 As shown, the substrate processing apparatus 100L is different from the substrate processing apparatus 100 and includes a control device 101L. The control device 101L includes a second control unit 102L and a storage unit 103L. The substrate processing apparatus 100L further includes a display unit 104L, an input unit 105L, and a communication unit 106L.

[0144] The second control unit 102L includes a processor. For example, the second control unit 102L includes a CPU or an MPU. Alternatively, the second control unit 102L may include a general-purpose arithmetic unit or a dedicated arithmetic unit. The second control unit 102L may also include an NPU.

[0145] The storage unit 103L stores data and computer programs. The storage unit 103L includes a primary storage device. The primary storage device is, for example, a semiconductor memory. The storage unit 103L may also include an auxiliary storage device. The auxiliary storage device is, for example, a semiconductor memory and / or a hard disk drive. The storage unit 103L may also include removable media. The second control unit 102L controls the operation of various components of the substrate processing apparatus 100 based on the data and computer programs stored in the storage unit 103L.

[0146] Specifically, the storage unit 103L stores a recipe 131L and a control program 132L. The recipe 131L specifies the processing details and processing sequence for the target substrate WL. Furthermore, the recipe 131L indicates processing conditions and various set values. In the first embodiment, the processing conditions include, for example, the rotation speed of the blower fan built into the blower unit 12, the rotation speed of the exhaust fan 82, the rotation speed of the target substrate WL, and the opening degree of the valve 74 provided in the piping 72 that supplies nitrogen gas to the nozzle 76. The recipe 131L may be the same as the recipe 131. Furthermore, the control program 132L may be the same as the control program 132.

[0147] The second control unit 102L controls the operation of each unit of the substrate processing apparatus 100L based on the recipe 131L and the control program 132L.

[0148] Second control unit 102L transmits a first detection result obtained by detecting at least one of the wind speed and wind direction at the first position during learning to learning data generation device 400. Specifically, second control unit 102L transmits first detection information 133La indicating the first detection result detected by first learning sensor 50L to learning data generation device 400.

[0149] Furthermore, the second control unit 102L transmits a second detection result obtained by detecting at least one of the wind speed and wind direction at the second position during learning to the learning data generation device 400. Specifically, the second control unit 102L transmits second detection information 133Lb indicating the second detection result detected by the second learning sensor 60L to the learning data generation device 400.

[0150] Furthermore, in the first embodiment, the second control unit 102L also transmits the processing conditions during learning to the learning data generating device 400 .

[0151] The storage unit 103L stores first detection information 133La indicating a first detection result, second detection information 133Lb indicating a second detection result, and learning processing condition information 133Lc indicating processing conditions during learning (hereinafter sometimes referred to as learning processing conditions).

[0152] The display unit 104L displays various information. In the first embodiment, the display unit 104L displays various setting screens (input screens), for example. Furthermore, the display unit 104L may also display images based on the first detection result and the second detection result, for example. Furthermore, the display unit 104L includes, for example, a liquid crystal display or an organic EL display.

[0153] The input unit 105L receives input from the user. For example, the input unit 105L receives commands to execute or stop processing on the target substrate WL, change processing conditions and various set values, and the like. Furthermore, the input unit 105L includes, for example, a touch panel and a pointing device. The touch panel is, for example, located on the display surface of the display unit 104L. The input unit 105L and the display unit 104L constitute, for example, a graphical user interface.

[0154] Communication unit 106L is connected to a network and communicates with external devices. Examples of networks include the Internet, a LAN (Local Area Network), a public telephone network, and a short-range wireless network. Communication unit 106L is a communication device, such as a network interface controller. In the first embodiment, communication unit 106L communicates with learning data generation device 400. Communication unit 106L transmits the first detection result, the second detection result, and the like to learning data generation device 400.

[0155] Next, refer to Figure 10 as well as Figure 11 The learning data generating device 400 will be further described. Figure 10 2 is a block diagram showing the learning data generating device 400 . Figure 11 4 is a diagram showing an example of the learning data 434 generated by the learning data generating device 400 .

[0156] like Figure 10 As shown, the learning data generating device 400 includes a control device 401. The control device 401 includes a control unit 402 and a storage unit 403.

[0157] The control unit 402 includes a processor. For example, the control unit 402 includes a CPU or an MPU. Alternatively, the control unit 402 may include a general-purpose arithmetic unit. The control unit 402 may also include an NPU.

[0158] The storage unit 403 stores a control program 431 and data. The control program 431 is a computer program. The storage unit 403 includes a primary storage device. The storage unit 403 may also include an auxiliary storage device. The storage unit 403 may also include removable media. The control unit 402 executes various processes based on the control program 431 and data stored in the storage unit 403.

[0159] The storage unit 403 further stores a data generation program 432. The control unit 402 stores the first detection result, the second detection result, and the learning processing conditions (hereinafter sometimes referred to as acquisition information 433) acquired from the substrate processing apparatus 100L in the storage unit 403.

[0160] The data generation program 432 is a program for generating learning data by associating the first detection result with the second detection result. In the first embodiment, the data generation program 432 is a program for generating learning data 434 by associating the first detection result, the second detection result, and the learning processing conditions.

[0161] The control unit 402 associates the first detection result with the second detection result to generate learning data 434, and stores it in the storage unit 403. In the first embodiment, the control unit 402 associates the first detection result, the second detection result, and the learning processing condition to generate learning data 434, and stores it in the storage unit 403. In addition, in the first embodiment, Figure 11 As shown, the learning data 434 generates the first detection results (DA1, DA2··), the second detection results (DB1, DB2··), and the learning processing conditions (#A1, #A2··) as time series data.

[0162] like Figure 10 As shown, the learning data generating device 400 further includes a display unit 404 , an input unit 405 , and a communication unit 406 .

[0163] The display unit 404 displays various screens and information. The display unit 404 is, for example, a liquid crystal display or an organic EL display. The display unit 404 may also display an input screen for inputting various data to the control unit 402.

[0164] The input unit 405 receives input from the user and outputs various information to the control unit 402. The input unit 405 includes, for example, a touch panel and a pointing device. The input unit 405 can also receive input of an instruction to generate the learning data 434.

[0165] The communication unit 406 is connected to a network and communicates with external devices. Examples of networks include the Internet, a local area network (LAN), a public telephone network, and a short-range wireless network. The communication unit 406 is a communication device, such as a network interface controller. In the first embodiment, the communication unit 406 communicates with the substrate processing apparatus 100L and the learning device 500. The communication unit 406 receives the first detection result and the second detection result, etc. from the substrate processing apparatus 100L. In addition, the communication unit 406 transmits the learning data 434 to the learning device 500.

[0166] Next, refer to Figures 10 to 12, describing the method for generating the learning data 434 according to the first embodiment. Figure 12 This is a flowchart illustrating a method for generating learning data 434 according to the first embodiment. The method for generating learning data 434 includes steps S301 and S302. Steps S301 and S302 are executed by control unit 402. Step S301 is an example of the "step of obtaining the first detection result" and the "step of obtaining the second detection result" of the present invention. Step S302 is an example of the "step of storing the first detection result and the second detection result as learning data in association with each other" of the present invention.

[0167] like Figure 12 As shown, in step S301, the learning data generating device 400 (control unit 402) obtains the first detection result and the second detection result. Typically, the control unit 402 obtains the first and second detection results from the substrate processing apparatus 100L. In the first embodiment, the control unit 402 obtains the first and second detection results, as well as the learning processing conditions. The storage unit 403 stores the obtained information 433.

[0168] In addition, in the first embodiment, the control unit 402 obtains the first detection result, the second detection result, and the processing conditions for learning in one process (step S301), but may obtain the first detection result, the second detection result, and the processing conditions for learning in different processes.

[0169] In step S302, the control unit 402 associates the first detection result with the second detection result and stores it as learning data in the storage unit 403. In the first embodiment, the control unit 402 associates the first detection result, the second detection result, and the learning processing conditions and stores them as learning data in the storage unit 403. Specifically, the control unit 402 associates the first detection result, the second detection result, and the learning processing conditions to generate learning data 434 and stores it in the storage unit 403.

[0170] As described above, the generation of the learning data 434 is completed.

[0171] Next, refer to Figure 13 The learning device 500 will be further described. Figure 13 is a block diagram showing the learning device 500. Figure 13 As shown, the learning device 500 includes a control device 501. The control device 501 includes a control unit 502 and a storage unit 503. The control unit 502 is an example of a "learning unit" in the present invention.

[0172] The control unit 502 includes a processor. For example, the control unit 502 includes a CPU or an MPU. Alternatively, the control unit 502 may include a general-purpose arithmetic unit. The control unit 502 may also include an NPU.

[0173] The storage unit 503 stores a control program 531 and data. The control program 531 is a computer program. The storage unit 503 includes a primary storage device. The storage unit 503 may also include an auxiliary storage device. The storage unit 503 may also include removable media. The control unit 502 executes various processes based on the control program 531 and data stored in the storage unit 503.

[0174] The storage unit 503 further stores a learning program 532. The control unit 502 causes the storage unit 503 to store the learning data 434 acquired from the learning data generating device 400.

[0175] The learning program 532 is a program for executing a machine learning algorithm that finds certain rules from the learning data 434 and generates a learned model M that represents the found rules. The control unit 502 executes the learning program 532, thereby adjusting the parameters of the inference program through machine learning based on the learning data 434. As a result, the learned model M is generated. The control unit 502 causes the storage unit 503 to store the generated learned model M.

[0176] The machine learning algorithm is not particularly limited as long as it is supervised learning, and can include, for example, a decision tree, a nearest neighbor algorithm, a naive Bayesian classifier, a support vector machine, or a neural network. Therefore, the learned model M includes a decision tree, a nearest neighbor algorithm, a naive Bayesian classifier, a support vector machine, or a neural network.

[0177] For example, a neural network includes an input layer, a single or multiple intermediate layers, and an output layer. Specifically, the neural network is a deep neural network (DNN), a recurrent neural network (RNN), or a convolutional neural network (CNN), and performs deep learning. For example, a deep neural network includes an input layer, multiple intermediate layers, and an output layer.

[0178] Thus, the machine learning algorithm is not particularly limited, but in the first embodiment, it is DiPLS (Dynamic-inner Partial Least Squares) regression. DiPLS is an algorithm that is improved based on PLS (Partial Least Squares) that is not easily affected by multicollinearity. In detail, DiPLS is an algorithm that improves PLS in a manner that allows variables such as rotational speed to be added as explanatory variables in consideration of time delays. By using DiPLS or PLS as a machine learning algorithm, even when there are similar tendencies (high correlation coefficients) between the explanatory variables (the detection results of the first processing sensor 50) as in the first embodiment, it is not easily affected by multicollinearity, so the regression equation can be prevented from becoming unstable. In addition, the machine learning algorithm may not be DiPLS, but PLS, or it may be LSTM (Long Short Term Memory), or it may be other algorithms.

[0179] The learning device 500 further includes a display unit 504 , an input unit 505 , and a communication unit 506 .

[0180] The display unit 504 displays various screens and information. The display unit 504 is, for example, a liquid crystal display or an organic EL display. The display unit 504 may also display an input screen for inputting various data to the control unit 502.

[0181] The input unit 505 receives input from the user and outputs various information to the control unit 502. The input unit 505 includes, for example, a touch panel and a pointing device. The input unit 505 may also receive input of an instruction to generate a learned model M.

[0182] The communication unit 506 is connected to a network and communicates with external devices. Examples of networks include the Internet, a local area network (LAN), a public telephone network, and a short-range wireless network. The communication unit 506 is a communication device, such as a network interface controller. In the first embodiment, the communication unit 506 communicates with the learning device 500 and the substrate processing apparatus 100. The communication unit 506 receives the learning data 434 from the learning data generating apparatus 400. Furthermore, the communication unit 406 transmits the learned model M to the substrate processing apparatus 100.

[0183] Next, refer to Figure 13 as well as Figure 14 , the learning method and the method for generating the learned model M according to the first embodiment are explained. Figure 144 is a flowchart showing the learning method and the method for generating the learned model M according to the first embodiment. The learning method of the learning data 434 includes steps S401 and S402. In addition, the method for generating the learned model M includes steps S401 to S403. Steps S401 to S403 are executed by the control unit 502. In addition, step S401 is an example of the "process of obtaining learning data" of the present invention. In addition, step S402 is an example of the "process of machine learning" of the present invention. In addition, steps S402 and S403 are an example of the "process of generating a learned model through machine learning" of the present invention.

[0184] like Figure 14 As shown, in step S401 , the control unit 502 obtains the learning data 434 from the storage unit 503 .

[0185] Next, in step S402 , the control unit 502 inputs the learning data 434 to the learning program 532 to perform machine learning.

[0186] In step S403 , the control unit 502 outputs a model (one or more functions) to which a plurality of learned parameters (coefficients) are applied as a learned model M. The storage unit 503 stores the learned model M. As a result, the learning method is completed, and the learned model M is generated.

[0187] (First Modification)

[0188] Next, a substrate processing learning system 1000 according to a first modification of the first embodiment will be described. In the first modification, an example of generating a learned model M by machine learning based on first and second detection results represented using a three-dimensional orthogonal coordinate system will be described.

[0189] In the first modified example, the first learning sensor 50L detects wind speed and direction using three values: wind speed, azimuth (0° to 360°), and depression (0° to 180°). Similarly to the first learning sensor 50L, the second learning sensor 60L and the first processing sensor 50 detect wind speed and direction using three values: wind speed, azimuth (0° to 360°), and depression (0° to 180°).

[0190] Here, when generating the learned model M, for example, if the azimuth angle is used as the first and second detection results, there is a possibility that the estimation accuracy of the learned model M will decrease due to discontinuities near 0° and 360°. Therefore, in the first modification, the first and second detection results are represented using a three-dimensional orthogonal coordinate system to suppress the decrease in estimation accuracy.

[0191] Specifically, a three-dimensional orthogonal coordinate system is defined as, for example, X-coordinates, Y-coordinates, and Z-coordinates. The control unit 402 of the learning data generating device 400 converts the three values ​​detected by the first learning sensor 50L into data using the X-coordinates, Y-coordinates, and Z-coordinates. Similarly, the control unit 402 converts the three values ​​detected by the second learning sensor 60L into data using the X-coordinates, Y-coordinates, and Z-coordinates. The control unit 402 then generates the learning data 434 using the first and second detection results represented by the three-dimensional orthogonal coordinate system (X-coordinates, Y-coordinates, and Z-coordinates).

[0192] The control unit 502 of the learning device 500 generates a learned model M by inputting the learning data 434 generated using the three-dimensional orthogonal coordinate system into the learning program 532. In other words, the learned model M is generated by machine learning based on the first detection result and the second detection result represented using the three-dimensional orthogonal coordinate system.

[0193] The first control unit 102 of the substrate processing apparatus 100 inputs the detection results of the first processing sensor 50 expressed in the three-dimensional orthogonal coordinate system into the learned model M. Thus, the first control unit 102 obtains the estimated wind speed and estimated wind direction expressed in the three-dimensional orthogonal coordinate system.

[0194] The other configuration and other methods (the method for generating the learning data 434 , the learning method, the method for generating the learned model M, and the estimation method) of the first modification are the same as those of the first embodiment.

[0195] In the first modification, as described above, the learned model M is generated by machine learning based on the first and second detection results represented using a three-dimensional orthogonal coordinate system.

[0196] Other effects of the first modification are the same as those of the first embodiment.

[0197] (Second Modification)

[0198] Next, refer to Figure 15 A second modification example of the substrate processing learning system 1000 will be described. In the second modification example, an example in which the learned model M has a regression model for every three coordinates will be described. Figure 15 3 is a diagram showing a learned model M according to a second modification.

[0199] like Figure 15As shown in FIG. 1 , in the second modification, the learned model M includes a regression model MX, a regression model MY, and a regression model MZ. The regression model MX is a regression model used to estimate the wind speed at the X coordinate of the second position P1 during processing (in other words, the X component of the wind speed). The regression model MY is a regression model used to estimate the wind speed at the Y coordinate of the second position P1 during processing (in other words, the Y component of the wind speed). The regression model MZ is a regression model used to estimate the wind speed at the Z coordinate of the second position P1 during processing (in other words, the Z component of the wind speed).

[0200] For example, the regression model MX is generated based on the detection results related to all coordinates (all X, Y, and Z coordinates) from the first and second detection results. The first control unit 102 of the substrate processing apparatus 100 inputs the detection results related to all coordinates (all X, Y, and Z coordinates) from the first processing sensor 50 into the regression model MX. This allows the first control unit 102 to obtain an estimated wind speed at the X coordinate of the second position P1 during processing.

[0201] The regression models MY and MZ are also similar to the regression model MX. Specifically, the regression models MY and MZ are generated based on the detection results related to all coordinates (all X, Y, and Z coordinates) from the first and second detection results, respectively. The first control unit 102 of the substrate processing apparatus 100 inputs the detection results related to all coordinates (all X, Y, and Z coordinates) from the first processing sensor 50 into the regression models MY and MZ, respectively. This allows the first control unit 102 to obtain the estimated wind speed for the Y coordinate and the estimated wind speed for the Z coordinate at the second position P1 during processing.

[0202] The other structures and other methods (the method for generating the learning data 434 , the learning method, the method for generating the learned model M, and the estimation method) of the second modification are the same as those of the first modification.

[0203] In the second modified example, as described above, the learned model M includes regression models MX, MY, and MZ. The first control unit 102 inputs the detection results of the first processing sensor 50, expressed in a three-dimensional orthogonal coordinate system, into the regression models MX, MY, and MZ, respectively, obtaining an estimated wind speed for each of the three coordinates. This improves the estimation accuracy of the learned model M.

[0204] Other effects of the second modification are the same as those of the first modification.

[0205] (Third Modification)

[0206] Next, refer to Figures 16 to 18A substrate processing learning system 1000 according to a third modification will be described. In the third modification, an example will be described in which the acquired estimated wind speed and estimated wind direction are displayed on the display unit 104 using a coordinate system based on each of the second learning sensors 60L. Figure 16 1 is a diagram showing an example of a coordinate system based on the processing chamber 11 according to the third modification. Figure 17 1 is a diagram showing an example of a coordinate system based on each of the second learning sensors 60L according to the third modification. Figure 18 This shows the use of the second learning sensor 60L (see Figure 5 ) are each used as a reference coordinate system to represent an example of an image of the estimated wind speed and estimated wind direction.

[0207] In the third modification, for example, a plurality of second learning sensors 60L (see Figure 5 ) are arranged rotationally symmetrically about the rotation axis AX. That is, the detection surfaces of the plurality of second learning sensors 60L are all arranged, for example, to face radially outward or radially inward. Therefore, for example, when an airflow flowing from the radially inward to the outward is generated at each second position during learning, the detection results of each second learning sensor 60L show the same trend.

[0208] On the other hand, when using Figure 16 When the coordinate system shown with the processing chamber 11 as the reference is used to display the estimated wind speed and estimated wind direction obtained using the learned model M on the display unit 104, the wind direction at the second position P1 during each treatment becomes substantially the same. That is, for example, the airflow does not flow from the radial inside to the outside. Therefore, when using Figure 16 When the display unit 104 displays an image showing the estimated wind speed and the estimated wind direction using the coordinate system based on the processing chamber 11 as shown, the user may feel uncomfortable.

[0209] Therefore, in the third modification, the first control unit 102 of the substrate processing apparatus 100 changes the estimated wind speed and estimated wind direction obtained from the following example: Figure 16 The coordinate system based on the processing chamber 11 is converted into Figure 17 The coordinate system shown is based on each second learning sensor 60L.

[0210] And, as Figure 18As shown, the first control unit 102 displays the converted estimated wind speed and estimated wind direction on the display unit 104. Thus, at each processing time second position P1, for example, an image showing airflow flowing from the radially inner side to the outer side is displayed. In this case, in the third modified example, the first control unit 102 synthesizes the estimated wind speed in the X coordinate, the estimated wind speed in the Y coordinate, and the estimated wind speed in the Z coordinate for each processing time second position P1 and displays the result on the display unit 104.

[0211] The other structures and other methods (the method for generating the learning data 434 , the learning method, the method for generating the learned model M, and the estimation method) of the third modification are the same as those of the second modification.

[0212] In the first modification, as described above, the first control unit 102 displays the acquired estimated wind speed and estimated wind direction on the display unit 104 using a coordinate system based on each of the second learning sensors 60L. This can prevent the user from feeling uncomfortable with the display unit 104.

[0213] Other effects of the third modification are the same as those of the second modification.

[0214] (Second embodiment)

[0215] Next, refer to Figure 19 A substrate processing training system 1000 according to a second embodiment of the present invention will be described. Figure 19 Schematic diagram of the processing unit 10 of the second embodiment. The substrate processing learning system 1000 of the second embodiment includes the substrate processing apparatus 100 , the substrate processing apparatus 100L, the learning data generating apparatus 400 , and the learning apparatus 500 , similarly to the first embodiment.

[0216] like Figure 19 As shown, in the second embodiment, the processing unit 10 of the substrate processing apparatus 100 further includes a processing detection unit 200. The processing detection unit 200 detects at least one of the gas pressure inside or around the processing chamber 11, the rotation speed of the processing target substrate Wp, the flow rate or supply pressure of the gas supplied into the processing chamber 11, and the flow rate or exhaust pressure of the gas exhausted from the processing chamber 11. In the second embodiment, the processing detection unit 200 includes an internal pressure sensor 201, an external pressure sensor 202, an exhaust pressure sensor 203, a flow meter 204, and a speed sensor 205.

[0217] The internal pressure sensor 201 and the external pressure sensor 202 detect the pressure inside or around the processing chamber 11. The exhaust pressure sensor 203 detects the flow rate or exhaust pressure of the gas exhausted from the processing chamber 11. The flow meter 204 detects the flow rate or supply pressure of the gas supplied into the processing chamber 11. The speed sensor 205 detects the rotational speed of the processing target substrate Wp. This is described in detail below.

[0218] The internal pressure sensor 201 is disposed within the processing chamber 11. The internal pressure sensor 201 detects the pressure (air pressure) of the gas within the processing chamber 11. The internal pressure sensor 201 outputs the detection result to the first control unit 102. The internal pressure sensor 201 is not particularly limited and may be, for example, disposed at a location where the wind speed of the gas is low. In the second embodiment, the internal pressure sensor 201 is disposed near the inner wall of the processing chamber 11. The internal pressure sensor 201 may include, for example, a pressure-sensitive element that detects the pressure of the gas.

[0219] The external pressure sensor 202 is arranged outside the processing chamber 11. The external pressure sensor 202 detects the pressure (air pressure) of the gas outside the processing chamber 11. The external pressure sensor 202 outputs the detection result to the first control unit 102. The external pressure sensor 202 is not particularly limited, and for example, it is arranged at a location where the wind speed of the gas is low. In the second embodiment, the external pressure sensor 202 is arranged near the outer wall of the processing chamber 11. In addition, the external pressure sensor 202 is arranged around the exhaust pipe 81, for example. In addition, the external pressure sensor 202 has, for example, a pressure-sensitive element that detects the pressure of the gas.

[0220] The exhaust pressure sensor 203 detects the pressure (air pressure) of the gas exhausted from the processing chamber 11. The exhaust pressure sensor 203 outputs the detection result to the first control unit 102. The exhaust pressure sensor 203 is not particularly limited and may be, for example, disposed within the exhaust pipe 81. Alternatively, the exhaust pressure sensor 203 may be disposed near the inlet or outlet of the exhaust pipe 81. Furthermore, the exhaust pressure sensor 203 may include, for example, a pressure-sensitive element that detects the pressure of the gas.

[0221] The flow meter 204 measures the amount of inert gas supplied from the inert gas supply unit 70 into the processing chamber 11. The flow meter 204 is disposed in the pipe 72. The flow meter 204 detects the flow rate of the inert gas passing through the pipe 72. The flow meter 204 outputs the detection result to the first control unit 102.

[0222] The speed sensor 205 detects the rotational speed of the substrate W (the rotational speed of the shaft 23). The speed sensor 205 outputs the detection result to the first control unit 102. The speed sensor 205 is not particularly limited, and for example, it detects the rotational speed of the spin base 21 or the shaft 23. Furthermore, the speed sensor 205 is not particularly limited, and for example, electromagnetic, Hall element, optical, or inductive sensors can be used.

[0223] The detection interval of the internal pressure sensor 201, the external pressure sensor 202, the exhaust pressure sensor 203, the flow meter 204, and the speed sensor 205 is not particularly limited, and is, for example, 0.1 seconds to 10 seconds. In the second embodiment, the internal pressure sensor 201, the external pressure sensor 202, the exhaust pressure sensor 203, the flow meter 204, and the speed sensor 205 perform detection every approximately 1 second.

[0224] In addition, other structures of the substrate processing apparatus 100 and the processing unit 10 of the second embodiment are the same as those of the first embodiment.

[0225] Next, refer to Figure 20 A substrate processing apparatus 100L according to a second embodiment will be described. Figure 20 Schematic diagram of the processing unit 10L of the substrate processing apparatus 100L according to the second embodiment. The substrate processing apparatus 100L further includes a learning detection unit 200L. The learning detection unit 200L is configured similarly to the processing detection unit 200. Specifically, the learning detection unit 200L detects at least one of the air pressure in or around the learning chamber 11L, the rotation speed of the learning target substrate WL, the flow rate or supply pressure of the gas supplied to the learning chamber 11L, and the flow rate or exhaust pressure of the gas exhausted from the learning chamber 11L. In the second embodiment, the learning detection unit 200L includes an internal pressure sensor 201L, an external pressure sensor 202L, an exhaust pressure sensor 203L, a flow meter 204L, and a speed sensor 205L.

[0226] The internal pressure sensor 201L is disposed in the training chamber 11L. The internal pressure sensor 201L detects the pressure of the gas in the training chamber 11L and outputs the detection result to the second control unit 102L.

[0227] The external pressure sensor 202L is disposed outside the training chamber 11L. The external pressure sensor 202L detects the pressure of the gas outside the training chamber 11L and outputs the detection result to the second control unit 102L.

[0228] The exhaust pressure sensor 203L detects the pressure of the gas exhausted from the learning chamber 11L to the outside, and outputs the detection result to the second control unit 102L.

[0229] The flow meter 204L measures the amount of the inert gas supplied from the inert gas supply unit 70 into the training chamber 11L, and outputs the detection result to the second control unit 102L.

[0230] The speed sensor 205L detects the rotation speed of the substrate W (the rotation speed of the shaft 23 ), and outputs the detection result to the second control unit 102L.

[0231] In addition, the other structures, actions and configuration positions of the internal pressure sensor 201L, external pressure sensor 202L, exhaust pressure sensor 203L, flow meter 204L and speed sensor 205L of the substrate processing apparatus 100L are the same as those of the internal pressure sensor 201, external pressure sensor 202, exhaust pressure sensor 203, flow meter 204 and speed sensor 205 of the substrate processing apparatus 100.

[0232] In the second embodiment, in addition to the first detection result, the second detection result and the processing conditions for learning, the second control unit 102L also sends the detection results of the internal pressure sensor 201L, the external pressure sensor 202L, the exhaust pressure sensor 203L, the flow meter 204L and the speed sensor 205L (hereinafter sometimes recorded as the third detection result) to the learning data generation device 400.

[0233] The other structures of the substrate processing apparatus 100L and the processing unit 10L of the second embodiment are the same as those of the first embodiment.

[0234] In the second embodiment, the learned model M is generated through machine learning based on the third detection results (the detection results of the internal pressure sensor 201L, the external pressure sensor 202L, the exhaust pressure sensor 203L, the flow meter 204L, and the speed sensor 205L) in addition to the first and second detection results. The detection targets of the internal pressure sensor 201L, the external pressure sensor 202L, the exhaust pressure sensor 203L, the flow meter 204L, and the speed sensor 205 all affect the wind speed and direction within the processing chamber 11. Therefore, by generating the learned model M through machine learning based on the third detection results (the detection results of the internal pressure sensor 201L, the external pressure sensor 202L, the exhaust pressure sensor 203L, the flow meter 204L, and the speed sensor 205L), it is possible to improve estimation accuracy.

[0235] Next, the learning data generating device 400 according to the second embodiment will be described. In the second embodiment, the control unit 402 generates the learning data 434 by associating the third detection result with the first detection result, the second detection result, and the learning processing conditions.

[0236] The rest of the configuration of the learning data generating device 400 of the second embodiment is the same as that of the first embodiment.

[0237] Next, a method for generating the learning data 434 according to the second embodiment will be described. Figure 12 In step S301 shown, the control unit 402 acquires the third detection result in addition to the first detection result, the second detection result, and the learning processing conditions.

[0238] In step S302 , the control unit 402 associates the third detection result with the first detection result, the second detection result, and the learning processing condition to generate the learning data 434 , and stores the data in the storage unit 403 .

[0239] As described above, the generation of the training data 434 is completed. The other generation methods of the training data 434 in the second embodiment are the same as those in the first embodiment.

[0240] The configuration of the learning device 500 , the learning method, and the method of generating the learned model M of the second embodiment are the same as those of the first embodiment.

[0241] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the above-described embodiments and can be implemented in various ways without departing from the scope of the present invention. In addition, the multiple components disclosed in the above-described embodiments can be appropriately modified. For example, a component of all the components shown in a certain embodiment can be added to the components of another embodiment, or some components of all the components shown in a certain embodiment can be deleted from the embodiment.

[0242] In the accompanying drawings, each component is schematically shown to facilitate understanding of the invention. The thickness, length, number, spacing, etc. of each component shown in the drawings may sometimes differ from the actual structure to facilitate the production of the drawings. In addition, the structures of the components shown in the above-mentioned embodiments are merely examples and are not particularly limiting. Various modifications can be made without departing from the scope of the effects of the present invention.

[0243] For example, in the first and second embodiments described above, the user changes processing conditions by confirming the estimated wind speed and estimated wind direction displayed on the display unit 104. However, the present invention is not limited to this. For example, if the estimated wind speed and estimated wind direction deviate from a predetermined range, the first control unit 102 may automatically change the processing conditions. In this case, the first control unit 102 may not display the estimated wind speed and estimated wind direction on the display unit 104. Alternatively, the first control unit 102 may automatically stop processing.

[0244] In addition, in the first and second embodiments described above, examples are described in which the first control unit 102 obtains both the estimated wind speed and the estimated wind direction. However, this is not limited to the present invention. The first control unit 102 may obtain only one of the estimated wind speed and the estimated wind direction. In other words, the learned model M may output only one of the estimated wind speed and the estimated wind direction.

[0245] In addition, in the first and second embodiments described above, the substrate processing apparatus 100 is described as including the rinse liquid supply unit 60 (cleaning liquid supply unit), but the present invention is not limited thereto. For example, the present invention can also be applied to a substrate processing apparatus that dries, heats, or etches a substrate.

[0246] In the first and second embodiments, the second processing position P1 is described as being on the cup 90 , but the present invention is not limited thereto.

[0247] Furthermore, for example, in the first modified example, the first learning sensor 50L, the second learning sensor 60L, and the first processing sensor 50 were described as detecting wind direction using two values: an azimuth angle (0° to 360°) and a depression angle (0° to 180°). However, the present invention is not limited to this. The first learning sensor 50L, the second learning sensor 60L, and the first processing sensor 50 may also output detection results using a three-dimensional orthogonal coordinate system.

[0248] In the second modification, for example, the learned model M has a regression model for every three coordinates, but the present invention is not limited thereto. For example, a single regression model may be used to estimate the wind speed for all coordinates (X, Y, and Z).

[0249] Industrial application possibilities

[0250] The present invention is applied to the field of processing substrates.

[0251] Description of Reference Numerals

[0252] 11: Processing chamber

[0253] 11L: Learning chamber

[0254] 20: Substrate holding part

[0255] 50, 50a~50j: First sensor for processing

[0256] 50L, 50La~50Lj: First sensor for learning

[0257] 60: Flushing liquid supply unit (cleaning liquid supply unit)

[0258] 60L, 60La~60Ld: Second sensor for learning

[0259] 90, 91, 92: Cup

[0260] 100: Substrate processing equipment

[0261] 102: First control unit (control unit)

[0262] 104: Display unit

[0263] 105: Input

[0264] 200: Processing detection unit

[0265] 200L: Learning detection unit

[0266] 201: Internal pressure sensor (processing detection unit)

[0267] 201L: Internal pressure sensor (detection unit for learning)

[0268] 202: External pressure sensor (processing detection unit)

[0269] 202L: External pressure sensor (detection unit for learning)

[0270] 203: Exhaust pressure sensor (processing detection unit)

[0271] 203L: Exhaust pressure sensor (detection unit for learning)

[0272] 204: Flow meter (processing detection unit)

[0273] 204L: Flow meter (detection unit for learning)

[0274] 205: Speed ​​sensor (processing detection unit)

[0275] 205L: Speed ​​sensor (detection unit for learning)

[0276] 434: Learning with Data

[0277] 500: Learning device

[0278] 502: Control Unit (Learning Unit)

[0279] 503: Storage

[0280] 532: Learning Program

[0281] M: learned model

[0282] MX, MY, MZ: regression models

[0283] P1, P1a~P1d: The second position during processing

[0284] Wp: substrate to be processed.

Claims

1. A substrate processing device, characterized in that: The substrate processing device comprises: a substrate holding portion that holds and rotates a substrate to be processed; a processing chamber accommodating the substrate holding portion; a first processing sensor disposed in or around the processing chamber and configured to detect at least one of wind speed and wind direction at a first position during processing in or around the processing chamber; and a control unit that inputs the detection result of the first processing sensor into the learned model and obtains at least one of an estimated wind speed and an estimated wind direction at a second position during processing in the processing chamber; The learned model is generated by machine learning based on a first detection result of a first learning sensor and a second detection result of a second learning sensor, wherein the first learning sensor detects at least one of the wind speed and wind direction at a first position during learning in or around a learning chamber, and the second learning sensor detects at least one of the wind speed and wind direction at a second position during learning in the learning chamber.

2. The substrate processing apparatus according to claim 1, wherein: The substrate processing apparatus includes: a cleaning liquid supply unit that supplies cleaning liquid to the processing target substrate; The control unit inputs a detection result of the first processing sensor during cleaning into the learned model, and acquires at least one of the estimated wind speed and the estimated wind direction during cleaning.

3. The substrate processing apparatus according to claim 1 or 2, wherein: The substrate processing apparatus includes: a cup disposed around the substrate holding portion and receiving the processing liquid discharged from the processing target substrate; The second position during the treatment is a position on the cup.

4. The substrate processing apparatus according to any one of claims 1 to 3, wherein: The learned model is generated by machine learning based on the first detection result and the second detection result represented using a three-dimensional orthogonal coordinate system, The control unit inputs a detection result of the first processing sensor expressed in a three-dimensional orthogonal coordinate system into the learned model, and acquires at least one of the estimated wind speed and the estimated wind direction expressed in a three-dimensional orthogonal coordinate system.

5. The substrate processing apparatus according to claim 4, wherein: The learned model has a regression model for every 3 coordinates, The control unit inputs the detection result of the first processing sensor expressed using a three-dimensional orthogonal coordinate system into each of the regression models, and obtains the estimated wind speed for each of the three coordinates.

6. The substrate processing apparatus according to any one of claims 1 to 5, wherein: The substrate processing apparatus includes a display unit that displays at least one of the estimated wind speed and the estimated wind direction.

7. The substrate processing apparatus according to claim 6, wherein: The control unit displays at least one of the acquired estimated wind speed and the estimated wind direction on the display unit using a coordinate system based on each of the second learning sensors.

8. The substrate processing apparatus according to claim 6 or 7, wherein: The substrate processing apparatus includes an input unit to which processing conditions for processing the processing target substrate are input. The processing condition capable of changing at least one of the estimated wind speed and the estimated wind direction is input to the input unit.

9. The substrate processing apparatus according to any one of claims 1 to 8, wherein: The substrate processing apparatus includes a processing detection unit, The processing detection unit detects at least one of the gas pressure in or around the processing chamber, the rotation speed of the processing target substrate, the flow rate or supply pressure of the gas supplied into the processing chamber, and the flow rate or exhaust pressure of the gas exhausted from the processing chamber. The learned model is generated by machine learning based on the first detection result, the second detection result, and the detection result of the learning detection unit. The learning detection unit detects at least one of the air pressure in or around the learning chamber, the rotation speed of the learning target substrate, the flow rate or supply pressure of the gas supplied to the learning chamber, and the flow rate or exhaust pressure of the gas exhausted from the learning chamber. The control unit inputs the detection result of the first processing sensor and the detection result of the processing detection unit into the learned model, and obtains at least one of the estimated wind speed and the estimated wind direction at the second position during the processing.

10. A method of estimating, characterized in that: The inference method includes the following steps: detecting at least one of wind speed and wind direction at a first position during processing in a processing chamber housing a substrate holding portion or around the processing chamber, the substrate holding portion rotating the substrate to be processed; as well as inputting a detection result of at least one of the wind speed and the wind direction into the learned model, obtaining at least one of an estimated wind speed and an estimated wind direction at a second position during processing in the processing chamber, The learned model is generated by machine learning based on a first detection result of a first learning sensor and a second detection result of a second learning sensor, wherein the first learning sensor detects at least one of the wind speed and wind direction at a first position during learning in or around a learning chamber, and the second learning sensor detects at least one of the wind speed and wind direction at a second position during learning in the learning chamber.

11. A method for generating learning data, characterized in that: The method for generating learning data includes the following steps: obtaining a first detection result of detecting at least one of a wind speed and a wind direction at a first position during training in a training chamber for accommodating a substrate holding portion or around the training chamber; obtaining a second detection result of detecting at least one of a wind speed and a wind direction at a second position in the learning chamber during learning; as well as The first detection result and the second detection result are associated with each other and stored in a storage unit as learning data.

12. A learning method, characterized in that The learning method comprises the following steps: acquiring learning data generated by the method for generating learning data according to claim 11; and Machine learning is performed by inputting the learning data into a learning program.

13. A learning device, characterized in that: The learning device comprises: a storage unit that stores the learning data generated by the method for generating learning data according to claim 11; and The learning unit performs machine learning by inputting the learning data into a learning program.

14. A method for generating a learned model, characterized in that: The method for generating the learned model includes the following steps: acquiring learning data generated by the method for generating learning data according to claim 11; and A learned model is generated by machine learning based on the learning data.

15. A learned model, characterized in that The learned model is generated by machine learning based on the learning data generated by the method for generating learning data according to claim 11.

Images