阅读说明：本技术 状态判断装置、状态判断方法和计算机可读取的存储介质 (State determination device, state determination method, and computer-readable storage medium ) 是由 牧准之辅 丰永真臣 太田敦 冈田基 筒井拓郎 佐野圭 矢野光辉 于 2020-03-18 设计创作，主要内容包括：本发明提供状态判断装置、状态判断方法和计算机可读取的存储介质,判断构成为能够在基片处理装置中保持基片并且使之动作的驱动机构的状态。该状态判断装置包括：构成为能够获取驱动机构的动作数据的获取部；模型生成部,其构成为能够基于正常动作数据,执行使用自动编码器的机器学习,来生成驱动机构的监视模型,其中,正常动作数据来自于在驱动机构的正常动作时由获取部获取到的动作数据；和第一判断部,其构成为能够基于将评价数据输入监视模型而得到的第一输出数据,判断驱动机构的状态,其中,评价数据来自于在驱动机构的评价时由获取部获取到的动作数据。(The invention provides a state determination device, a state determination method and a computer readable storage medium, which determine the state of a driving mechanism configured to hold and operate a substrate in a substrate processing device. The state judgment device includes: an acquisition unit configured to acquire operation data of the drive mechanism; a model generation unit configured to generate a monitoring model of the drive mechanism by performing machine learning using the automatic encoder based on normal operation data from the operation data acquired by the acquisition unit during normal operation of the drive mechanism; and a first determination unit configured to determine a state of the drive mechanism based on first output data obtained by inputting evaluation data to the monitoring model, the evaluation data being derived from the operation data acquired by the acquisition unit at the time of evaluation of the drive mechanism.)

1. A state determination device characterized by:

the state determination device determines a state of a drive mechanism configured to hold and operate a substrate in a substrate processing apparatus, and includes:

an acquisition unit configured to acquire operation data of the drive mechanism;

a model generation unit configured to generate a monitoring model of the drive mechanism by performing machine learning using an automatic encoder based on normal operation data from the operation data acquired by the acquisition unit during normal operation of the drive mechanism; and

a first determination unit configured to determine a state of the drive mechanism based on first output data obtained by inputting evaluation data to the monitoring model, the evaluation data being derived from the operation data acquired by the acquisition unit at the time of evaluation of the drive mechanism.

2. The apparatus of claim 1, wherein:

the first determination unit executes the following processing:

processing for acquiring an allowable error based on a first error between the normal operation data and second output data obtained by inputting the normal operation data into the monitoring model;

a process of acquiring a first deviation ratio from the allowable error by comparing a second error between the evaluation data and the first output data with the allowable error; and

and a process of determining a state of the drive mechanism based on the first deviation ratio.

3. The apparatus of claim 2, wherein:

in the parameter mu₁、σ₁Respectively setting as follows:

μ₁: average value of the first error

σ₁: standard deviation of the first error

When the tolerance is mu₁±3σ₁The range of (1).

4. The apparatus of claim 2 or 3, wherein:

the first deviation ratio is a value obtained by calculating a Root Mean Square Error (RMSE) based on the second error and the allowable error.

5. The apparatus of any one of claims 2 to 4, wherein:

the process of determining the state of the drive mechanism based on the first deviation ratio includes a process of determining based on whether or not the first deviation ratio exceeds a predetermined threshold.

6. The apparatus of claim 5, wherein:

in the parameter sigma₂Is arranged as

σ₂: a standard deviation from a second rate of deviation of the tolerance error based on a comparison of the first error to the tolerance error

When the threshold is according to 3 sigma₂And the value obtained.

7. The apparatus of claim 6, wherein:

the second deviation ratio is a value obtained by calculating a Root Mean Square Error (RMSE) based on the first error and the allowable error.

8. The apparatus of any one of claims 5 to 7, further comprising:

a storage unit configured to store a data set in which a result of determination of the state of the drive mechanism based on the first deviation ratio is stored for a predetermined period; and

and a second determination unit configured to determine a degree to which the drive mechanism is approaching an abnormal state based on a proportion of data in the data group in which the first deviation ratio exceeds the threshold value.

9. The apparatus of any one of claims 1 to 8, wherein:

further comprising an adjusting unit configured to adjust the amount of the operation data acquired by the acquiring unit to a predetermined amount,

the normal operation data is data obtained by adjusting the amount of the operation data acquired by the acquisition unit to a predetermined amount by the adjustment unit during normal operation of the drive mechanism,

the evaluation data is data obtained by adjusting the amount of the operation data acquired by the acquisition unit at the time of evaluation of the drive mechanism by the adjustment unit to a fixed amount.

10. The apparatus of any one of claims 1 to 9, wherein:

the drive mechanism includes:

a support member supporting the substrate; and

a motor for operating the support member,

the acquisition unit is configured to be able to acquire a torque signal of the motor as the operation data.

11. A state determination method, comprising:

generating a monitoring model of the driving mechanism by performing machine learning using an automatic encoder based on normal operation data from operation data during normal operation of the driving mechanism configured to hold and operate a substrate; and

and determining a state of the drive mechanism based on first output data obtained by inputting evaluation data to the monitoring model, the evaluation data being derived from operation data at the time of evaluation of the drive mechanism.

12. A computer-readable storage medium, characterized in that:

a program for causing a drive mechanism of a substrate processing apparatus to execute the state determination method according to claim 11 is stored.

Technical Field

The invention relates to a state determination device, a state determination method, and a computer-readable storage medium.

Background

Patent document 1 discloses a substrate transfer mechanism provided in a substrate processing apparatus for processing a substrate such as a semiconductor wafer. The substrate transport mechanism is configured to be movable between different modules in the substrate processing apparatus. The transport mechanism moves, for example, between the carrier and the process module to take out one substrate from the carrier that receives a plurality of substrates and transport the substrate to the process module.

Documents of the prior art

Patent document

Patent document 1: japanese patent laid-open publication No. 2013-133192

Disclosure of Invention

Technical problem to be solved by the invention

The invention provides a state determination device, a state determination method and a computer-readable storage medium capable of determining the state of a drive mechanism of a substrate simply and accurately.

Technical solution for solving technical problem

The state determination device according to one aspect of the present invention determines the state of a drive mechanism configured to be capable of holding and operating a substrate in a substrate processing apparatus. The state judgment device includes: an acquisition unit configured to acquire operation data of the drive mechanism; a model generation unit configured to generate a monitoring model of the drive mechanism by performing machine learning using the automatic encoder based on normal operation data from the operation data acquired by the acquisition unit during normal operation of the drive mechanism; and a first determination unit configured to determine a state of the drive mechanism based on first output data obtained by inputting evaluation data to the monitoring model, the evaluation data being derived from the operation data acquired by the acquisition unit at the time of evaluation of the drive mechanism.

Effects of the invention

According to the state determination device, the state determination method, and the computer-readable storage medium of the present invention, the state of the drive mechanism of the substrate can be determined easily and with high accuracy.

Drawings

Fig. 1 is a plan view schematically showing an example of a substrate processing system.

Fig. 2 is a side view schematically showing an example of the transport device.

Fig. 3 is a block diagram showing an example of the functional configuration of the controller.

Fig. 4 is a block diagram showing an example of the hardware configuration of the controller.

Fig. 5 is a flowchart showing an example of a state determination flow of the conveyance device.

Fig. 6 is a flowchart showing an example of a flow of generating a monitoring model.

Fig. 7 is a diagram for explaining adjustment of acquired data by the adjustment unit.

Fig. 8 is a diagram for explaining a monitoring model generated by machine learning.

Fig. 9 is a diagram for explaining a monitoring model generated by machine learning.

Fig. 10 is a graph for explaining an allowable error included in the monitoring model.

Fig. 11 (a) and 11 (b) are diagrams for explaining the deviation ratio between the allowable error and the output value.

Fig. 12 is a graph for explaining a method of setting the threshold value of the deviation ratio.

Fig. 13 is a flowchart showing an example of a monitoring flow of the transport apparatus.

Fig. 14 is a graph showing an example of the verification result of the monitoring model.

Fig. 15 is a graph showing an example of the result of verification of the monitoring model.

Detailed Description

Hereinafter, an example of the embodiment of the present invention will be described in more detail with reference to the drawings. In the following description, the same reference numerals are used for the same elements or elements having the same functions, and redundant description is omitted.

[ substrate processing apparatus ]

The substrate processing system 1 shown in fig. 1 is a system configured to be able to perform substrate processing on a wafer W. The substrate processing system 1 includes a substrate processing apparatus 2 and a controller 60. The wafer W may have a disc shape, a part of a circle may be notched, or a shape other than a circle such as a polygon. The wafer W may be, for example, a semiconductor substrate, a glass substrate, a mask substrate, an FPD (Flat Panel Display) substrate, or other various substrates. The diameter of the wafer W is, for example, about 200mm to 450 mm.

As shown in fig. 1, the substrate processing apparatus 2 includes processing units 3A, 3B and a conveying apparatus 10 (driving mechanism). The processing units 3A and 3B are units configured to be able to perform a predetermined process on the wafer W. The processing units 3A and 3B may be liquid processing units that supply a processing liquid to the surface of the wafer W. The processing units 3A and 3B may be heat treatment units that perform heat treatment (heating or cooling) on the coating film formed on the surface of the wafer W. The processing units 3A, 3B may have the same function as each other or different functions from each other. In the example shown in fig. 1, the processing units 3A and 3B are arranged in a horizontal direction along the direction of arrow D1 (the left-right direction in fig. 1).

[ details of the conveying apparatus ]

Next, the conveying device 10 will be described in further detail with reference to fig. 1 and 2. The transfer device 10 is configured to be able to transfer the wafer W. The transfer device 10 can transfer the wafer W between the process unit 3A and the process unit 3B, for example. The transfer apparatus 10 may transfer the wafer W from another unit in the substrate processing apparatus 2 to the processing units 3A and 3B, or may transfer the wafer W from the processing units 3A and 3B to another unit. The conveyance device 10 may be disposed opposite the process units 3A, 3B. The conveying device 10 includes a driving portion 20 and a holding portion 30.

The driving unit 20 is configured to be able to reciprocate the holding unit 30 in a predetermined direction. As shown in fig. 1, for example, the driving unit 20 can reciprocate (operate) the holding unit 30 in a direction (direction of arrow D1) in which the process units 3A and 3B are aligned. The drive unit 20 includes a housing 21, a linear moving body 22, a guide rail 23, pulleys 24 and 25, a belt 26, and a motor 27. The housing 21 houses various elements included in the driving unit 20. An opening 21a is provided in a wall of the casing 21 opposite to the process units 3A, 3B.

The linear moving body 22 is a member extending in the direction of the arrow D2 (the vertical direction in fig. 1). The base end of the linear moving body 22 is connected to a guide rail 23 and a belt 26 in the housing 21. The front end portion of the linearly movable body 22 protrudes out of the housing 21 through the opening 21 a. The guide rail 23 is laid in the housing 21 so as to linearly extend in the direction of arrow D1 (the width direction of the housing 21). The pulleys 24 and 25 are disposed at respective end portions of the housing 21 in the direction of the arrow D1. The pulleys 24 and 25 are rotatably provided in the housing 21 around a rotation shaft in the direction of arrow D2.

A belt 26 is mounted on the pulleys 24, 25. The drive belt 26 may be, for example, a timing belt. The motor 27 is a power source that generates torque, and is configured to be operable based on a control signal from the controller 60. The motor 27 may be a servo motor, for example. The motor 27 is connected to the pulley 25. When the torque (driving force) generated by the motor 27 is transmitted to the pulley 25, the belt 26 bridged over the pulleys 24, 25 moves in the direction of arrow D1. Thereby, the linearly movable body 22 can also reciprocate in the arrow D1 direction along the guide rail 23.

The holding unit 30 is configured to be able to hold a wafer W to be conveyed. For example, the holding portion 30 includes a base 31, a rotation shaft 32, a driving portion 33, and an arm 34 (support member) as shown in fig. 1 and 2. The base 31 is attached to the front end of the linearly movable body 22. Therefore, the holding portion 30 can also reciprocate in the arrow D1 direction in accordance with the movement of the linear moving body 22.

The rotation shaft 32 extends upward from the base 31 in the vertical direction. The rotary shaft 32 is rotationally driven by a motor (not shown) configured to be operable based on a control signal from the controller 60. A driving unit 33 is connected to an upper portion of the rotating shaft 32. Therefore, when the rotation shaft 32 rotates, the driving part 33 and the arm 34 rotate about the rotation shaft 32.

The driving unit 33 is configured to be able to reciprocate the arm 34 in a direction different from the moving direction of the holding unit 30 by the driving unit 20. The driving unit 33 can reciprocate (operate) the arm 34 in the direction of the arrow D2, for example. The drive unit 33 reciprocates the arm 34 to carry in and out the wafer W held by the arm 34 to and from the process units 3A and 3B. The driving unit 33 includes, for example, as shown in fig. 2, a housing 33a, a linearly movable body 33b, pulleys 33c and 33d, a belt 33e, and a motor 33 f. The housing 33a accommodates the elements included in the driving unit 33. An opening 33g is provided in the upper wall of the housing 33 a.

The linearly moving body 33b is a member extending in the vertical direction. The lower end of the linear moving body 33b is connected to a belt 33e inside the housing 33 a. The upper end of the linearly moving body 33b protrudes out of the housing 33a through the opening 33 g. The pulleys 33c and 33D are disposed at respective end portions of the housing 33a in the direction of arrow D2. The pulleys 33c and 33D are rotatably provided in the housing 33a around a rotation shaft along the arrow D1 direction.

The belt 33e is mounted on pulleys 33c and 33 d. The belt 33e may be a timing belt, for example. The motor 33f is a power source that generates torque, and is configured to be operable based on a control signal from the controller 60. The motor 33f may be a servo motor, for example. When the torque (driving force) generated by the motor 33f is transmitted to the pulley 33D, the belt 33e stretched over the pulleys 33c and 33D moves in the direction of arrow D2. Thereby, the linear moving body 33b also reciprocates in the arrow D2 direction.

The arm 34 is configured to surround the peripheral edge of the wafer W and support the back surface of the wafer W. The arm 34 is attached to the distal end portion of the linearly movable body 33 b. Accordingly, the arm 34 can also reciprocate in the arrow D2 direction in accordance with the movement of the linearly moving body 33 b. The holding portion 30 may include a plurality of arms 34 arranged to overlap in the vertical direction.

[ controller ]

Next, the controller 60 will be described in more detail with reference to fig. 3 and 4. The controller 60 controls the substrate processing apparatus 2 partially or entirely. As shown in fig. 3, the controller 60 includes a state determination unit 70 (state determination means). The state determination unit 70 determines the state of the transport apparatus 10 that holds and operates the wafer W. An example of the case where the state determination unit 70 determines the state of the driving unit 33 (for example, whether or not the tension of the belt 33e is appropriate) will be described below.

The state determination section 70 includes, as functional blocks, for example, a reading section 71, a storage section 72, an instructing section 73, an acquiring section 74, an adjusting section 75, a model generating section 76, a determining section 77 (first determining section), a determining section 78 (second determining section), and an output section 79. These functional blocks are merely a plurality of blocks into which the functions of the controller 60 are divided for convenience, and do not necessarily mean that the hardware constituting the controller 60 is divided into such blocks. The functional blocks are not limited to being implemented by executing programs, and may be implemented by dedicated circuits (e.g., logic circuits) or integrated circuits (ASIC) in which they are integrated.

The reading unit 71 has a function of reading a program from a computer-readable storage medium RM. The storage medium RM stores a program for operating each unit in the transport apparatus 10 accompanying the transport of the wafer W and a program for determining the state of the transport apparatus 10 by the state determination unit 70. The storage medium RM may be, for example, a semiconductor memory, an optical disk, a magnetic disk, or an opto-magnetic disk.

The storage unit 72 has a function of storing various data. The storage unit 72 stores, for example, a program read from the storage medium RM by the reading unit 71, various data for determining the state of the conveying device 10, a determination result regarding the state of the conveying device 10, and the like.

The instructing unit 73 has a function of transmitting a control signal based on a program for operating each unit in the conveying device 10 stored in the storage unit 72. Specifically, the instructing unit 73 drives the motor 33f of the driving unit 33 to generate a control signal for moving the arm 34 in the direction of the arrow D2. The instructing unit 73 drives the motor 27 of the driving unit 20 to generate a control signal for moving the arm 34 in the direction of the arrow D1.

The acquisition unit 74 has a function of acquiring operation data of the conveyor 10. The acquisition unit 74 may acquire, for example, a torque signal of the motor 33f as operation data. The acquisition section 74 may acquire a torque signal for each action of the arm 34. The torque signal may be time-series data obtained at a predetermined sampling period from a time change (analog signal) of the torque of the motor 33 f. One motion of the arm 34 may be, for example, a motion in which the arm 34 is moved in a single direction in the direction of the arrow D2 by driving the motor 33 f. The acquiring unit 74 may acquire, for example, discrete values of about 100 to 200 for each operation of the arm 34 from the time variation of the torque. The acquisition unit 74 outputs the acquired torque signal to the adjustment unit 75.

The adjusting unit 75 has a function of adjusting the amount of the operation data (torque signal) acquired by the acquiring unit 74 to a predetermined amount. The operation time of one operation of the arm 34 may vary slightly even in the same operation. Therefore, when the discrete value of the torque signal is obtained by the obtaining unit 74 at a constant sampling cycle, the data amount may differ for each operation of the arm 34. The adjusting unit 75 adjusts the data amount of the torque signal, which differs for each operation of the arm 34, to a fixed amount. The adjusting unit 75 may, for example, perform Discrete Fourier Transform (DFT) on the torque signal to obtain frequency data, and perform Inverse Discrete Fourier Transform (IDFT) on the frequency data so that the amount of data after the Transform becomes a predetermined number (for example, 128).

By adjusting the data amount of the torque signal, for example, a torque signal whose data amount is compressed can be generated. That is, the operation data having a data amount larger than the predetermined number may be adjusted to the operation data (compressed operation data) compressed to the predetermined number of data amounts. The adjusting unit 75 outputs the operation data with the adjusted data amount to the storage unit 72 and the determining unit 77. The adjustment unit 75 may adjust the amount of the motion data by another method. For example, when the amount of the operation data exceeds 128, the adjusting unit 75 may remove 129 th or subsequent data. Instead of compressing the data amount, the adjusting unit 75 may adjust the data amount so that the data amount is increased with respect to the data amount of the torque signal before adjustment. Hereinafter, a case of compressing the data amount will be described as an example.

The model generation unit 76 has a function of generating a monitoring model relating to the transport apparatus 10. When the driving unit 33 is the object of the state determination by the state determination unit 70, the model generation unit 76 generates the monitoring model by performing machine learning using the automatic encoder based on the normal operation data from the operation data (torque signal) acquired by the acquisition unit 74 when the driving unit 33 is operating normally. The model generation unit 76 generates a monitoring model and outputs the monitoring model to the storage unit 72. The normal operation is an operation of the driving unit 33 in a state where it is determined that deterioration, abnormality, or the like of the driving unit 33 has not occurred. The normal operation data may be operation data (compressed operation data) in which the adjustment unit 75 compresses the data amount, or may be operation data acquired by the acquisition unit 74. The details of the method of generating the monitoring model will be described later.

The determination unit 77 has a function of determining the state of the conveyance device 10. The determination unit 77 determines the state of the drive mechanism based on output data (first output data) obtained by inputting evaluation data from the operation data acquired by the acquisition unit 74 at the time of evaluation of the conveyor 10 into the monitoring model. The evaluation means a state in which the wafer W is continuously processed in the substrate processing apparatus 2 in a state in which, for example, an operator or the like cannot grasp the state of the transport apparatus 10. The evaluation data may be operation data (compressed operation data) obtained by compressing the data amount by the adjusting unit 75, or may be operation data obtained by the obtaining unit 74. The determination method of the determination unit 77 will be described later. The determination unit 77 outputs the determination result to the storage unit 72.

The determination unit 78 has a function of determining the degree to which the conveyance device 10 is approaching the abnormal state based on the determination result of the determination unit 77 stored in the storage unit 72 for a predetermined period. The determination method by the determination unit 78 will be described later. The determination unit 78 outputs the determination result to the output unit 79.

The output unit 79 has a function of outputting the determination result of the determination unit 78. The output unit 79 may output a signal indicating the determination result to other elements in the controller 60, or may output a signal indicating the determination result to the outside of the controller 60, for example. The output unit 79 may output a signal (hereinafter referred to as a "warning signal") indicating that the conveyor 10 is approaching an abnormal state as a signal indicating the determination result. When the warning signal is output, the controller 60 may temporarily stop the transport operation by the transport apparatus 10, or may temporarily stop the processing of the wafer W in the substrate processing apparatus 2. Alternatively, the substrate processing apparatus 2 further includes a notification unit (not shown) that notifies an operator or the like that the transport apparatus 10 is approaching an abnormal state when receiving the warning signal from the output unit 79.

The hardware of the controller 60 is constituted by one or more control computers, for example. The controller 60 has a hardware configuration, for example, a circuit 81 shown in fig. 4. The circuit 81 may be constituted by a circuit device (circuit). Specifically, the circuit 81 includes a processor 82, a memory 83 (storage unit), a memory 84 (storage unit), and an input/output port 85. The processor 82 executes a program in cooperation with at least one of the memory 83 and the storage 84, and executes input and output of signals via the input/output port 85, thereby configuring each of the above-described functional modules. The input/output port 85 inputs/outputs signals between the processor 82, the memory 83, and the storage 84, and various devices (the transfer device 10) of the substrate processing apparatus 2.

In the present embodiment, the substrate processing system 1 has one controller 60, but may have a controller group (control unit) including a plurality of controllers 60. In the case where the substrate processing system 1 has a controller group, the above-described functional blocks may be realized by one controller 60, or may be realized by a combination of 2 or more controllers 60. When the controller 60 is configured by a plurality of computers (circuits 81), the above-described functional blocks may be realized by one computer (circuit 81) or a combination of 2 or more computers (circuits 81). The controller 60 may have a plurality of processors 82. In this case, the above-described functional modules may be respectively implemented by one or more processors 82.

[ method for determining State ]

Next, a method of determining the state of the conveyance device 10 will be described with reference to fig. 5.

First, the controller 60 generates a monitoring model of the driving unit 33 based on the operation data when the driving unit 33 normally operates (step S10 in fig. 5). Next, the controller 60 monitors the state of the driving unit 33 based on the operation data at the time of evaluation of the driving unit 33 using the generated monitoring model (step S20 in fig. 5). The controller 60 may repeatedly execute the process of step S20. The stage in which the processing in step S10 is performed is the "learning stage", and the stage in which the processing in step S20 is continued is the "monitoring (evaluating) stage".

[ method of generating monitoring model ]

Next, the method of generating the monitoring model in step S10 will be described in more detail with reference to fig. 6 to 12. The monitor model may be generated, for example, when the substrate processing apparatus 2 does not process the wafer W. Further, when the operator determines that the state of the driving unit 33 is normal, the monitoring model may be generated.

First, the state determination unit 70 acquires operation data when the driving unit 33 normally operates (step S11 in fig. 6). In step S11, first, the instructing unit 73 controls the motor 33f to operate the arm 34 once in the direction of the arrow D2. Next, the acquisition unit 74 acquires, as operation data, a torque signal obtained by sampling the time change of the torque during the operation at a predetermined sampling period. The acquisition unit 74 may acquire a torque signal obtained from a current command value of the instruction unit 73 for the motor 33f as operation data, or may acquire a torque signal obtained from a detection result of a torque sensor provided in the motor 33f as operation data. Fig. 7 shows an example of the operation data obtained by the obtaining unit 74 as operation data T1. In this example, the amount of motion data T1 is 136. That is, the motion data T1 is represented by 136 discrete values. The acquiring unit 74 outputs the acquired operation data to the adjusting unit 75.

Next, the state determination unit 70 adjusts the operation data acquired by the acquisition unit 74 (step S12 in fig. 6). In step S12, the adjustment unit 75 adjusts the amount of the operation data to a fixed amount. The adjusting unit 75 may generate the compressed operation data by performing discrete fourier transform and inverse discrete fourier transform on the operation data, for example. In fig. 7, an example of the compression operation data generated by the adjustment unit 75 is shown as compression operation data T2. In this example, the compressed motion data T2 is generated by adjusting (compressing) 136 pieces of motion data T1 to 128 pieces of data. The horizontal axis sample count corresponds to time, and as shown in fig. 7, the compressed motion data T2 is a waveform obtained by compressing the motion data T1 on the time axis (horizontal direction in the drawing). The adjusting unit 75 outputs the generated compression operation data to the storage unit 72. The compressed operation data obtained during the normal operation is used as learning data (normal operation data) for generating the monitoring model.

Next, the state determination unit 70 determines whether or not the number of normal operation data generated by the adjustment unit 75 has reached a predetermined number (hereinafter referred to as "collection number") (step S13 in fig. 6). When determining that the number of normal operation data has not reached the collection number (no in step S13), the state determination unit 70 repeats steps S11 and S12. At this time, the state determination unit 70 repeats the same operation of the drive unit 33 to acquire a plurality of normal operation data. The state determination unit 70 repeats, for example, an operation of the drive unit 33 when the wafer W is carried into the processing unit 3A (the processing unit 3B) or an operation of the drive unit 3A when the wafer W is carried out.

Thus, a plurality of (for example, 600 to 1800) normal operation data are stored in the storage unit 72 as a learning data set. The state determination unit 70 may store a plurality of data sets for learning in the storage unit 72 for each case where the tension (vibration frequency) of the belt 33e is set to different values. For example, the data set for learning may include: 200-600 pieces of normal action data are acquired under the condition that the vibration frequency corresponding to the tension is 140 Hz; 200-600 pieces of normal action data are acquired under the condition that the vibration frequency is 130 Hz; and 200-600 pieces of normal action data acquired under the condition that the vibration frequency is 120 Hz.

When determining that the number of normal operation data included in the learning data group has reached the collection number (yes in step S13), the state determination unit 70 generates a monitoring model AE (see fig. 8) based on the stored learning data group (step S14 in fig. 6). The monitoring model AE is a model based on the characteristics of the driving unit 33, and is used to determine the state of the driving unit 33. In step S14, the model generation unit 76 performs machine learning based on the plurality of normal operation data in the learning data group to generate a monitoring model corresponding to the specific operation of the drive unit 33.

The model generation section 76 generates a monitoring model AE based on a plurality of normal torque signals included in a learning data group by machine learning using an automatic encoder (AutoEncoder) which is a kind of neural network. A model having an intermediate layer, which is a layer in which output data of the same data amount outputs the same value as input data with respect to input data of a certain data amount, is generated by machine learning using an automatic encoder. The intermediate layers of the model include a plurality of layers for sequentially reducing the dimension of the feature amount from the input data and sequentially restoring the feature amount. When the amount of data adjusted by the adjusting unit 75 is 128, for example, a monitoring model AE in which 128 pieces of data are input and 128 pieces of output data are obtained is generated. The model generation unit 76 outputs the generated monitoring model AE to the storage unit 72.

Fig. 8 shows an example of output data obtained when the normal operation data Tin1 is input to the monitoring model AE. Since the monitor model AE is a model generated based on the normal operation data, when the normal operation data Tin1 is input to the monitor model AE, the output data Tout1 having a waveform close to the normal operation data Tin1 is output from the monitor model AE. On the other hand, fig. 9 shows an example of output data obtained when the monitoring model AE is input to the compression operation data Tin2 in a case where the driving unit 33 is not in the normal state. In this case, output data Tout2 having a large waveform deviation from the compression operation data Tin2 is output from the monitoring model AE. That is, the state of the driving unit 33 can be determined by utilizing the fact that the error (deviation) of the output data output from the monitoring model AE with respect to the input data input to the monitoring model AE becomes large when the driving unit 33 approaches an abnormality.

Next, the state determination unit 70 (determination unit 78) calculates the allowable error Ea in the monitoring model AE (step S15 in fig. 6). Here, although the monitoring model AE is generated based on the normal operation data, even if the same normal operation data is input again, the output data that completely matches the input data cannot be output. That is, when the normal operation data is input data, an error (deviation) may occur between the input data and the output data due to the monitoring model AE itself. Therefore, in the present embodiment, the state determination unit 70 calculates an error caused by the monitoring model AE itself as the allowable error Ea. In step S15, first, the model generation unit 76 inputs the plurality of normal operation data included in the learning data group to the monitoring model AE, and calculates the difference between the input data and the output data (second output data) as an error Eb (first error).

Fig. 10 shows an example of the calculation result of the error Eb. In fig. 10, for example, when 10 normal operation data are input to the monitoring model AE, "number (Tick no)" shows the calculation result of the error Eb of 111 to 114 as an example. Here, the "number" corresponds to the sample count value shown in fig. 7, and for example, when the "number" is 111, it indicates 111 th data. Hereinafter, data having "numbers" of 1 to 128 will be referred to as "1 st data" to "128 th data", respectively.

The determination unit 78 may calculate the error Eb of each of the 1 st to 128 th data in all or a part of the normal operation data included in the learning data group. The model generation unit 76 may calculate the allowable error Ea due to the monitoring model AE itself based on the error Eb. The model generation unit 76 generates the parameter μ₁、σ₁Are respectively provided with

μ₁: average value of error Eb

σ₁: standard deviation of error Eb

Mu of time₁±3σ₁Is set as the allowable error Ea. The determination unit 78 stores the calculated allowable error Ea in the storage unit 72.

Next, the controller 60 calculates a threshold value Th1 of the deviation ratio da (first deviation ratio) (step S16 in fig. 6). The deviation rate da is an index indicating how close the driving unit 33 is to the abnormal state at the time of evaluation. The threshold Th1 indicates that the driving mechanism of the subject is approaching an abnormal state. Here, after the method of calculating the deviation ratio is described, a specific example of the method of calculating the threshold Th1 of the deviation ratio da will be described.

In step S16, the determination unit 77 calculates the difference between the error Eb and the allowable error Ea as the corrected error Ec for each of the 1 st data to the 128 th data with respect to the plurality of normal operation data. The judgment section 77 may calculate the corrected error Ec as 0 if the value of the error Eb is included in the range of the allowable error Ea. If the value of the error Eb is out of the range of the allowable error Ea, the judgment section 77 may calculate the difference between the upper limit value or the lower limit value of the allowable error Ea and the value of the error Eb as the correction error Ec.

Fig. 11 (a) shows an example of the calculation results of the error Eb and the allowable error Ea. In fig. 11 (a), the value of the error Eb is indicated by a black dot, and the range of the allowable error Ea is indicated by a vertical solid line. In the example shown in fig. 11 (a), the error Eb is out of the allowable error Ea for the 7 th to 9 th data, and the error Eb is within the allowable error Ea for the 10 th to 12 th data. Fig. 11 (b) shows, as an example, the calculation result of the difference (corrected error Ec) between the error Eb and the allowable error Ea shown in fig. 11 (a). Since the error Eb is out of the range of the allowable error Ea for the 7 th to 9 th data, the corrected error Ec is not 0. On the other hand, for the 10 th to 12 th data, since the error Eb is within the allowable error Ea, the corrected error Ec is 0.

The determination unit 77 may perform processing of calculating a deviation ratio dr (second deviation ratio) for learning with respect to the plurality of normal operation data based on the correction error Ec of each of the 1 st data to the 128 th data. The determination unit 77 may perform, for example, a process of calculating a Root Mean Square Error (RMSE) of the correction Error Ec (difference between the Error Eb and the allowable Error Ea) of the 1 st to 128 th data as the deviation ratio dr for learning on the plurality of normal operation data. The root mean square error based on the error Eb and the allowable error Ea can be obtained by calculating the square root of the average value obtained by averaging the square values of the correction errors Ec of the 1 st to 128 th data for each normal operation data.

Fig. 12 shows an example of the calculation result of the deviation ratio dr of the data set for learning. In fig. 12, the calculation result of the deviation ratio dr for learning is represented by a "box chart". Since the calculation result of the deviation ratio dr based on the learning data set is shown in fig. 12, a box indicating the quartile range is drawn in the vicinity of the deviation ratio dr being 0, and the box is in a state where it cannot be observed by eyes. The judgment unit 77 sets the parameter σ to the threshold Th1 of the deviation ratio da used in the evaluation₂Is arranged as

σ₂: the standard deviation from the deviation rate dr for learning the allowable error Ea, which is obtained based on the comparison between the error Eb and the allowable error Ea, is used

Th1＝3σ₂

A threshold Th1 may be calculated. The determination unit 77 outputs the threshold Th1 to the storage unit 72.

[ State monitoring method of conveying device ]

Next, a method of monitoring the state of the driving unit 33 in step S20 shown in fig. 5 will be described in more detail with reference to fig. 13. The state monitoring of the driving unit 33 can be continued, for example, when the wafer W is processed by the substrate processing apparatus 2.

First, the state determination unit 70 acquires operation data at the time of evaluation of the driving unit 33 (step S21 in fig. 13). In step S21, the instructing unit 73 controls the motor 33f in synchronization with (in accordance with) the processing of the wafer W in the substrate processing apparatus 2, thereby causing the arm 34 to perform one operation in the direction of the arrow D2. Next, the acquisition unit 74 acquires, as operation data, a torque signal (evaluation torque signal) obtained by sampling a temporal change in torque during the operation at a predetermined sampling period. Step S21 is performed in the same manner as step S11, except that it is not clear whether the state of the driver 33 is normal or not. The acquiring unit 74 outputs the acquired operation data to the adjusting unit 75.

Next, the state determination unit 70 adjusts the operation data acquired by the acquisition unit 74 (step S22 in fig. 13). In step S22, the adjustment unit 75 adjusts the amount of motion data to a fixed amount (for example, 128 pieces) to generate compressed motion data, as in step S12. The adjusting unit 75 outputs the generated compression operation data to the determining unit 77. The compression operation data obtained at the time of evaluation is used as evaluation data (evaluation data) for determining the state of the driving unit 33.

Next, the state determination unit 70 calculates the deviation ratio da based on the evaluation data generated by the adjustment unit 75 (step S23 in fig. 13). In step S23, the determination unit 77 calculates the deviation ratio da based on the monitoring model AE stored in the storage unit 72. First, the determination unit 77 may calculate an error Ed (second error) between output data obtained by inputting evaluation data to the monitoring model AE and the input evaluation data, for example. Then, the determination unit 77 compares the error Ed with the allowable error Ea of the monitoring model AE to calculate the deviation rate da. This deviation rate da can be obtained by calculating a root mean square error (obtained by comparing the error Ed with the allowable error Ea) based on the error Ed and the allowable error Ea, as in the calculation of the deviation rate dr in step 16. The root mean square error is obtained by calculating the square root of the average value obtained by averaging the square values of the differences between the error Ed and the allowable error Ea of the 1 st to 128 th data for each evaluation data based on the root mean square error of the error Ed and the allowable error Ea.

Next, the state determination unit 70 makes a primary determination of determining the state of the target drive mechanism based on the calculated deviation da (step S24 in fig. 13). In step S24, the determination unit 77 may determine the state of the target drive mechanism based on whether or not the deviation ratio da exceeds the threshold Th1 stored in the storage unit 72. Determination unit 77 may output the result of determination as to whether deviation ratio da exceeds threshold Th1 to storage unit 72.

Next, the state determination unit 70 determines whether or not a predetermined period of time has elapsed since the start of monitoring of the drive unit 33 (step S25 in fig. 13). When determining that the predetermined period has not elapsed (no in step S25), the state determination unit 70 repeats steps S21 to S25. Thus, a data set (hereinafter referred to as "determination data set") in which the determination result of the state of the drive unit 33 based on the deviation ratio da is stored for a predetermined period is stored in the storage unit 72. The storage unit 72 may store a predetermined period, and the predetermined period may be set in advance by an operator, for example. The predetermined period may be set to, for example, 1 hour, several hours, half a day, 1 day, or 1 week.

When it is determined that the predetermined period has elapsed (yes in step S25), the state determination unit 70 makes 2 determinations based on the determination data set to determine how close the drive unit 33 is to the abnormal state (step S26 in fig. 13). In step S26, for example, the determination unit 78 determines the degree to which the drive unit 33 is approaching an abnormal state based on the proportion of data (hereinafter referred to as "data proportion") in the determination data group for which the deviation ratio da exceeds the threshold Th 1. The data rate is a rate of the number of determination results determined to exceed the threshold Th1 for all the determination times determined by the determination unit 77 in a predetermined period.

When the data rate exceeds a preset threshold Th2, the determination unit 78 may determine that the target drive mechanism is approaching an abnormal state. The threshold Th2 may be set to any value by an operator or the like, and may be set in the range of 70% to 100%, 80% to 100%, or 90% to 100%. The determination unit 78 outputs the determination result to the output unit 79.

Next, the state determination unit 70 outputs the determination result (step S27 of fig. 13). In step S27, the output unit 79 may output a signal (warning signal) indicating that the target drive mechanism is approaching an abnormal state as a signal indicating the determination result, for example.

[ test results ]

Next, the verification result regarding the determination of the transport mechanism using the monitoring model will be described with reference to fig. 14 and 15. Fig. 14 shows the result of calculating the deviation rate da using the monitoring model AE based on a plurality (500) of normal operation data for verification when the belt 33e of the driving unit 33 has different tensions from each other.

In the example shown in fig. 14, the deviation ratios da in the case where the vibration frequency corresponding to the tension of the belt 33e of the conveying mechanism is changed in units of 10Hz within the range of 140Hz to 60Hz are calculated. Indicating that the lower the vibration frequency, the lower the tension. Further, it is shown that the lower the tension, the more the belt 33e of the conveying mechanism deteriorates. In fig. 14, the distribution of the calculation result of the deviation ratio da for each tension (vibration frequency) is shown as a box chart. From the calculation results shown in fig. 14, it is found that the maximum value of deviation ratio da increases as the tension decreases, and deviation ratio da included in the quartile range indicated by the box increases.

Fig. 15 shows a data rate, which is a rate at which the deviation rate da exceeds the threshold value Th1, under the same conditions as the verification result of the deviation rate da shown in fig. 14. As shown in fig. 15, the data rate is 75% or more when the vibration frequency is 90Hz or less, and 90% or more when the vibration frequency is 80Hz or less. Since the drive unit 33 approaches the abnormal state as the vibration frequency (tension) decreases, the determination unit 78 can determine that the drive unit 33 approaches the abnormal state by setting the threshold Th2 to 75%, for example. Alternatively, by setting the threshold Th2 to 90%, the determination unit 78 can determine that the drive unit 33 is approaching an abnormal state.

[ Effect ]

In the above example, the state of the conveyor 10 is determined based on the output data obtained by inputting the evaluation data from the operation data acquired by the acquisition unit when evaluating the conveyor 10 into the monitoring model. In this case, when normal operation data is input to the monitoring model generated by machine learning based on normal operation data using the automatic encoder, and when operation data at the time of abnormal operation of the conveyor 10 is input, a value having a large difference can be output. Therefore, the state of the conveyor 10 can be determined easily and accurately based on the first output data from the monitoring model.

In accordance with the above example, the determination unit 77 executes processing for acquiring the allowable error Ea based on the error Eb between the normal operation data and the output data obtained by inputting the normal operation data to the monitoring model (step S15). Further, the determination section 77 performs: a process of obtaining a deviation rate da from the allowable error Ea by comparing an error Ed between the evaluation data and the output data with the allowable error Ea; and a process of determining the state of the conveying device 10 based on the deviation da (step S24).

In this case, the monitoring model can be generated by machine learning using the automatic encoder so that an error between the normal operation data and the output data output from the monitoring model to which the normal operation data is input becomes very small. In other words, when the operation data at the time of the abnormal operation of the conveyor 10 is input to the monitoring model, an error between the operation data at the time of the abnormal operation and the output data from the monitoring model becomes large. Therefore, the state of the conveyor 10 can be determined easily and accurately.

According to the above example, the allowable error Ea is μ₁±3σ₁The range of (1). In this case, the range in which the abnormal value that may be included in the normal operation data is removed becomes the allowable error. By comparing the error Ed with the allowable error, a value having a large error can be distinguished from the values included in the error Ed with high accuracy. Therefore, the abnormal operation of the conveyor 10 can be determined more accurately.

In the above example, the deviation ratio da at the time of evaluation is a value obtained by calculating a root mean square error based on the error Ed at the time of evaluation and the allowable error Ea. In this case, the deviation rate da indicates how much the evaluation data deviates from the allowable error as a whole. By determining the state of the conveyor 10 based on the deviation da, the accuracy of the abnormality determination can be further improved.

In the above example, the process of determining the state of the conveyor 10 based on the deviation da at the time of evaluation includes a process of determining whether or not the deviation da exceeds a predetermined threshold Th 1. In this case, the state of the conveyance device 10 can be determined by a very simple method of comparing the deviation ratio da with the threshold Th 1.

According to the above example, the threshold Th1 is based on 3 σ₂The value is obtained. By comparing the deviation ratio da at the time of evaluation with the threshold Th1, it is possible to accurately distinguish a portion exceeding the deviation ratio that can be inherent in the normal operation data among the obtained deviation ratios. Therefore, the abnormal operation of the conveyor 10 can be determined more accurately.

In the above example, the deviation ratio dr during learning may be a value obtained by calculating a root mean square error based on the error Eb during normal operation and the allowable error Ea. In this case, the deviation ratio dr in the normal operation indicates how much the error Eb in the normal operation deviates from the allowable error Ea as a whole. By determining the state of the conveyor 10 using the threshold Th1 obtained based on the deviation ratio dr in the normal operation, the accuracy of the abnormality determination can be further improved.

According to the above example, further comprising: a storage unit 72 that stores a data set obtained by accumulating the determination result of the state of the conveyor 10 based on the deviation ratio da at the time of evaluation for a predetermined period; and a determination unit 78 that determines the degree to which the conveyance device 10 is approaching an abnormal state based on the proportion of data in the data group for which the deviation ratio da exceeds a predetermined threshold Th 1. In this case, the maintenance timing of the conveyor 10 can be grasped based on the determination result of the determination unit 78.

According to the above example, the adjustment unit 75 is provided to adjust the data amount of the operation data acquired by the acquisition unit 74 to a fixed amount. In this case, the subsequent data processing can be easily performed.

According to the above example, the transport apparatus 10 includes the arm 34 for supporting the wafer W and the motor 33f for operating the arm 34, and the acquisition unit 74 acquires a torque signal of the motor as operation data. In this case, the torque signal that can be easily acquired can be used as the operation data of the conveyor 10 to determine the abnormal operation of the conveyor 10.

[ modified examples ]

All matters hithertofore set forth herein are to be interpreted in an illustrative and non-limiting sense. Various omissions, substitutions, changes, and the like may be made to the above examples without departing from the scope of the claims and their spirit.

(1) The object of the state determination by the state determination unit 70 may be the holding unit 30 which conveys the wafer W in the direction of the arrow D1 of the conveying device 10. Alternatively, the object of the state determination may be a driving mechanism that drives the rotation shaft 32, or may be a mechanism that moves the arm 34 in the vertical direction.

(2) The state determination unit 70 may not include the determination unit 78. In this case, the state determination unit 70 may perform the determination only once based on the deviation rate da at the time of evaluation in one operation of the conveyance mechanism. The state determination unit 70 may store only the primary determination result in the storage unit 72, and may output the primary determination result.

(3) The allowable error Ea is not limited to the above example. The allowable error Ea may be, for example, μ₁±2σ₁May be in the range of μ₁±σ₁May be in the range of μ₁±n×σ₁(n is an arbitrary number).

(4) The threshold Th1 is not limited to the value obtained by the above example. The threshold Th1 may be n1 × σ₂(n1 is an arbitrary positive number).

(5) The state determination unit 70 (state determination means) may be housed in a case different from the controller 60 and configured as a computer (circuit) different from the controller 60. The state determination unit 70 may be constituted by a computer or a servo device that can be connected to the substrate processing apparatus 2 from the outside. In this manner, the state determination unit 70 does not need to be integrally configured with the substrate processing apparatus 2 or the controller 60, and may be implemented as an external apparatus capable of performing communication connection by wire or wireless as necessary.

(6) The model generation unit 76 may be realized by a controller different from the controller 60. For example, a servo device or the like separate from the substrate processing apparatus 2 may have the other controller. In this case, the controller 60 may obtain the monitoring model generated by the model generation unit 76 of the other controller by communicating with the other controller via a predetermined communication method such as a network.

[ other examples ]

Example 1. a state determination device (70) according to an example of the present invention determines the state of a drive mechanism (10) configured to be capable of holding and operating a substrate (W) in a substrate processing apparatus (2). The state determination device (70) includes: an acquisition unit (74) configured to acquire operation data of the drive mechanism (10); a model generation unit (76) configured to be able to execute machine learning using an automatic encoder based on normal operation data from the operation data acquired by the acquisition unit (74) during normal operation of the drive mechanism (10) and generate a monitoring model of the drive mechanism (10); and a first determination unit (77) configured to determine the state of the drive mechanism (10) based on first output data obtained by inputting evaluation data from the operation data acquired by the acquisition unit (74) at the time of evaluation of the drive mechanism (10) into the monitoring model. In this case, when normal operation data is input to the monitoring model generated by machine learning based on normal operation data using the automatic encoder, and when operation data at the time of abnormal operation of the drive mechanism is input, a value having a large difference can be output. Therefore, the state of the drive mechanism can be determined easily and accurately based on the first output data from the monitoring model.

Example 2 in the apparatus of example 1, the first determination unit (77) may execute: a process of obtaining an allowable error (Ea) based on a first error (Eb) between normal operation data and second output data obtained by inputting the normal operation data to the monitoring model; a process of obtaining a first deviation ratio (da) from the allowable error (Ea) by comparing a second error (Ed) between the evaluation data and the first output data with the allowable error (Ea); and a process of determining the state of the drive mechanism (10) based on the first deviation ratio (da). In this case, the monitoring model is generated by machine learning using the automatic encoder so that an error between the normal operation data and output data output from the monitoring model to which the normal operation data is input becomes very small. In other words, when the operation data at the time of the abnormal operation of the driving mechanism is input to the monitoring model, an error between the operation data at the time of the abnormal operation and the output data output from the monitoring model becomes large. Therefore, the state of the drive mechanism can be determined easily and accurately.

Example 3 in the apparatus of example 2, the parameter μmay be set₁、σ₁Respectively setting as follows:

μ₁: average value of the first error (Eb)

σ₁: standard deviation of first error (Eb)

When the tolerance (Ea) is mu₁±3σ₁The range of (1). In this case, the range in which the abnormal value that may be included in the normal operation data is removed becomes the allowable error. By comparing the second error with the allowable error, a value having a large error can be distinguished with high accuracy from values included in the second error. Therefore, the abnormal operation of the drive mechanism can be determined more accurately.

Example 4 in the apparatus of example 2 or 3, the first deviation ratio may be a value obtained by calculating a Root Mean Square Error (RMSE) based on the second error and the allowable error. In this case, the first deviation ratio indicates how much the evaluation data as a whole deviates from the allowable error. By determining the state of the drive mechanism based on such a first deviation ratio, the accuracy of the abnormality determination can be further improved.

Example 5 in any one of the apparatuses of examples 2 to 4, the process of determining the state of the drive mechanism (10) based on the first deviation ratio (da) may include a process of determining whether or not the first deviation ratio (da) exceeds a predetermined threshold value (Th 1). In this case, the state of the drive mechanism can be determined by a very simple method of comparing the first deviation ratio with a threshold value.

Example 6 in the apparatus of example 5, the parameter σ may be set₂Is arranged as

σ₂: a standard deviation from a second deviation rate (dr) of the allowable error (Ea) obtained based on a comparison of the first error (Eb) with the allowable error (Ea)

The threshold value (Th1) is based on 3 sigma₂To find the value. By comparing the first deviation ratio with the threshold value, it is possible to accurately distinguish a portion exceeding the deviation ratio that can be specific to the normal operation data among the obtained first deviation ratios. Therefore, the abnormal operation of the drive mechanism can be determined more accurately.

Example 7 in the apparatus of example 6, the second deviation ratio (dr) may be a value obtained by calculating a Root Mean Square Error (RMSE) based on the first error (Eb) and the allowable error (Ea). In this case, the second deviation ratio indicates how much the first error deviates from the allowable error as a whole. By determining the state of the drive mechanism based on the threshold value obtained using such a second deviation ratio, the accuracy of abnormality determination can be further improved.

Example 8 in any one of the apparatuses of examples 5 to 7, the apparatus may further include: a storage unit (72) configured to store a data set in which the determination result of the state of the drive mechanism (10) based on the first deviation ratio (da) is stored for a predetermined period; and a second determination unit (78) configured to determine the degree to which the drive mechanism (10) is approaching the abnormal state, based on the proportion of data in the data group for which the first deviation ratio (da) exceeds a predetermined threshold value (Th 1). In this case, the maintenance timing of the drive mechanism can be grasped based on the determination result of the second determination unit.

Example 9. in any one of the apparatuses of examples 1 to 8, the apparatus may further include an adjusting unit (75) configured to be capable of adjusting the data amount of the operation data acquired by the acquiring unit (74) to a fixed amount, the normal operation data may be data obtained by the adjusting unit (75) adjusting the data amount of the operation data acquired by the acquiring unit (74) to a fixed amount during the normal operation of the drive mechanism (10), and the evaluation data may be data obtained by the adjusting unit (75) adjusting the data amount of the operation data acquired by the acquiring unit (74) to a fixed amount during the evaluation of the drive mechanism (10). In this case, the subsequent data processing can be easily executed.

Example 10 in any one of the apparatuses of examples 1 to 9, the drive mechanism (10) may include a support member (21) that supports the substrate (W) and a motor (33f) that operates the support member, and the acquisition unit (74) may be configured to acquire a torque signal of the motor (33f) as the operation data. In this case, the abnormal operation of the drive mechanism can be determined using the torque signal that can be easily acquired as the operation data of the drive mechanism.

Example 11. a state determination method according to another example of the present invention includes: a step of generating a monitoring model of the driving mechanism (10) by executing machine learning using an automatic encoder based on normal operation data from operation data when the driving mechanism (10) configured to hold and operate the substrate (W) is operating normally; and a step of determining the state of the drive mechanism based on first output data obtained by inputting evaluation data from the operation data of the drive mechanism (10) at the time of evaluation. In this case, the same operational effects as in example 1 can be obtained.

Example 12 a computer-readable storage medium according to another example of the present invention stores a program for causing a state determination device (70) to execute the method of example 11. In this case, the same operational effects as those of the method of example 11 can be obtained. In this specification, a computer-readable storage medium includes a non-transitory tangible medium (non-transitory computer-readable storage medium) (e.g., various main storage devices or auxiliary storage devices), a propagated signal (transitory computer storage medium) (e.g., a data signal that can be provided via a network).

Description of the reference numerals

2 … … substrate processing apparatus, 10 … … conveying apparatus, 33 … … driving section, 34 … … arm, 60 … … controller, 70 … … state judging section, 74 … … acquiring section, 75 … … adjusting section, 76 … … model generating section, 77, 78 … … judging section.