JP7324853B2 - SUBSTRATE PROCESSING APPARATUS, SEMICONDUCTOR DEVICE MANUFACTURING METHOD, AND SIGNAL DETECTION PROGRAM - Google Patents

Description

The present disclosure relates to a substrate processing apparatus, a semiconductor device manufacturing method, and a sign detection program.

Generally, a substrate processing apparatus that forms a thin film on a substrate such as a wafer to manufacture a semiconductor device includes a vacuum pump that evacuates a processing chamber, a mass flow controller that controls the flow rate of a reactive gas, an on-off valve, a pressure It is composed of various members such as a meter, a heater for heating the processing chamber, and a transport mechanism for transporting the substrate.

Each of these various members gradually deteriorates and breaks down as they are used, and thus need to be replaced with new members. As for the method of replacement, it may be operated by either using a member until it fails, or by determining a regular replacement cycle for each member and replacing it before it fails. . Here, if the member is used until it fails, all the substrates processed by the substrate processing apparatus at the time of the failure may become defective, resulting in a loss of production time for the substrate and the time of failure. In addition, in the case of periodic replacement before a failure occurs, it is necessary to replace the member in a period before failure occurs, that is, every short period of time.

In addition, as in Patent Document 1 and Patent Document 2, various techniques related to maintenance of these members have been proposed, but it is still sometimes impossible to detect abnormalities in the members in advance.

International Publication No. 2016-157402 International Publication No. 2017-158682

An object of the present disclosure is to provide a configuration capable of detecting a sign of abnormality of a member.

According to one aspect of the present disclosure, a configuration acquires sensor data related to a member to be detected as an abnormality sign detection target, creates a normal model, and monitors the state of an apparatus based on the normal model, wherein the abnormality sign detection target after the member is replaced or maintained, the sensor data is acquired, the normal model is recreated from the sensor data, the state of the device is monitored based on the normal model, and the device is stopped abnormally. A configuration for detecting a sign of abnormality is provided.

According to the present disclosure, a technology is provided that can detect signs of abnormality in members.

1 is a perspective view showing a schematic configuration of a substrate processing apparatus according to one embodiment; FIG. 1 is an elevation cross-sectional view showing a schematic configuration of a processing furnace of a substrate processing apparatus according to one embodiment; FIG. It is a block diagram which shows schematic structure of the main-control part of the substrate processing apparatus which concerns on one Embodiment. FIG. 3 is a flowchart showing a substrate processing process when the substrate processing apparatus according to one embodiment is used as a semiconductor manufacturing apparatus; 1 is a block diagram showing a control system of a substrate processing apparatus according to one embodiment; FIG. FIG. 4 is an explanatory diagram of singular spectrum conversion in the control system of the substrate processing apparatus according to one embodiment; FIG. 11 is a flow diagram showing a part of steps of a sign detection process according to a specific example of the third embodiment; FIG. 16 is a flow diagram showing a part of the process of sign detection processing according to the specific example of the fourth embodiment;

A method for manufacturing a semiconductor device, a sign detection program, and a substrate processing apparatus according to an embodiment of the present disclosure will be described below. In FIG. 1, arrow F indicates the front direction of the substrate processing apparatus, arrow B indicates the rear direction, arrow R indicates the right direction, arrow L indicates the left direction, arrow U indicates the upward direction, and arrow D indicates the downward direction.

<Overall Configuration of Processing Device>
A configuration of the substrate processing apparatus 10 will be described with reference to FIGS. 1 and 2. FIG. As shown in FIG. 1, the substrate processing apparatus 10 includes a housing 12 made of a pressure-resistant container. The front wall of the housing 12 is provided with an opening for maintenance, and the opening is provided with a pair of front maintenance doors 14 as an access mechanism for opening and closing the opening. In the substrate processing apparatus 10, a pod (substrate container) 18 containing a substrate (wafer) 16 (see FIG. 2) made of silicon or the like, which will be described later, is used as a carrier for transporting the substrate 16 into and out of the housing 12. be.

A pod loading/unloading port is opened in the front wall of the housing 12 so as to communicate the inside and outside of the housing 12 . A load port 20 is installed at the pod loading/unloading port. A pod 18 is placed on the load port 20 and is configured to be aligned.

A rotatable pod shelf 22 is installed at the upper portion of the substantially central portion within the housing 12 . A plurality of pods 18 are configured to be stored on the rotatable pod shelf 22 . The rotatable pod shelf 22 includes a support that is vertically erected and rotated in a horizontal plane, and a plurality of shelf plates that are radially supported by the support at upper, middle, and lower positions.

A pod transfer device 24 is installed between the load port 20 and the rotary pod shelf 22 within the housing 12 . The pod transport device 24 has a pod elevator 24A capable of moving up and down while holding the pod 18 and a pod transport mechanism 24B. By continuous operation of the pod elevator 24A and the pod transport mechanism 24B, the pods 18 are mutually transported among the load port 20, the rotary pod shelf 22, and the pod opener 26, which will be described later.

A sub-housing 28 is provided in the lower portion of the housing 12 from the substantially central portion to the rear end of the housing 12 . A pair of pod openers 26 for transporting the substrate 16 into and out of the sub-housing 28 is installed on the front wall of the sub-housing 28 .

Each pod opener 26 has a mounting table on which the pod 18 is mounted, and a cap attaching/detaching mechanism 30 for attaching/detaching the cap of the pod 18 . The pod opener 26 is configured to open and close the substrate loading/unloading port of the pod 18 by attaching/detaching the lid of the pod 18 placed on the table by means of the cap attachment/detachment mechanism 30 .

A transfer chamber 32 that is fluidly isolated from a space in which the pod transfer device 24, the rotary pod shelf 22, and the like are installed is configured in the sub-casing 28 . A substrate transfer mechanism 34 is installed in the front area of the transfer chamber 32 . The substrate transfer mechanism 34 is composed of a substrate transfer device 34A capable of horizontally rotating or linearly moving the substrate 16, and a substrate transfer device elevator 34B for raising and lowering the substrate transfer device 34A.

The substrate transfer device elevator 34B is installed between the right end of the front region of the transfer chamber 32 of the sub-housing 28 and the right end of the housing 12 . Further, the substrate transfer device 34A is provided with tweezers (not shown) as a holder for the substrate 16. As shown in FIG. By continuous operation of the substrate transfer device elevator 34B and the substrate transfer device 34A, the substrate 16 can be loaded (charged) and unloaded (discharged) from the boat 36 as a substrate holder. ing.

As shown in FIG. 2, a boat elevator 38 for raising and lowering a boat 36 is installed in the sub-casing 28 (transfer chamber 32). An arm 40 is connected to the platform of the boat elevator 38 , and a cover 42 is horizontally attached to the arm 40 . The lid 42 vertically supports the boat 36 and is configured to close the lower end of the processing furnace 44, which will be described later.

A transport mechanism that transports the substrate 16 mainly by the rotary pod shelf 22, the pod transport device 24, the substrate transfer mechanism 34, the boat 36, the boat elevator 38 shown in FIG. is configured.

As shown in FIG. 1, a processing furnace 44 is provided above a waiting section 50 that accommodates and makes the boat 36 stand by. A clean unit 52 is installed at the left end of the transfer chamber 32 opposite to the substrate transfer apparatus elevator 34B side. The clean unit 52 is configured to supply clean air 52A, which is a purified atmosphere or inert gas.

A plurality of apparatus covers (not shown) are attached to the outer circumferences of the housing 12 and the sub-housing 28 as a mechanism for entering the substrate processing apparatus 10 . A door switch 54 (only the door switch 54 of the housing 12 is shown) is provided as an entry sensor at the ends of the housing 12 and the sub-housing 28 facing these device covers.

A substrate detection sensor 56 for detecting the placement of the pod 18 is provided on the load port 20 . The switches and sensors such as the door switch 54 and the substrate detection sensor 56 are electrically connected to a substrate processing apparatus controller 58 (see FIGS. 2 and 3) as a main control section to be described later.

As shown in FIG. 2 , the substrate processing apparatus 10 includes a gas supply unit 60 and an exhaust unit 62 outside the housing 12 . A processing gas supply system and a purge gas supply system are housed in the gas supply unit 60 . The processing gas supply system includes a processing gas supply source and opening/closing valve (not shown), a mass flow controller (hereinafter abbreviated as MFC) 64A as a gas flow controller, and a processing gas supply pipe 66A. Also, the purge gas supply system includes a purge gas supply source and an opening/closing valve (not shown), an MFC 64B, and a purge gas supply pipe 66B.

In the exhaust unit 62, a gas exhaust mechanism including an exhaust pipe 68, a pressure sensor 70 as a pressure detection unit, and a pressure adjustment unit 72 including, for example, an APC (Auto Pressure Controller) valve is housed. . Although not shown, a vacuum pump 74 as an exhaust device is connected to the exhaust pipe 68 on the downstream side of the exhaust unit 62 . Note that the vacuum pump 74 may also be included in the gas exhaust mechanism.

As shown in FIG. 2, the substrate processing apparatus controller 58 as a main controller is connected to the transfer controller 48, the temperature controller 76, the pressure controller 78, and the gas supply controller 80, respectively. Further, as shown in FIG. 5, the substrate processing apparatus controller 58 is connected to a sign detection controller 82 as a sign detection unit, which will be described later.

<Configuration of processing furnace>
As shown in FIG. 2, the processing furnace 44 has a reaction tube (process tube) 84 . The reaction tube 84 includes an inner reaction tube (inner tube) 84A and an outer reaction tube (outer tube) 84B provided outside thereof. The internal reaction tube 84A is formed in a cylindrical shape with an upper end and a lower end opened, and a processing chamber 86 for processing the substrate 16 is formed in the cylindrical hollow portion of the internal reaction tube 84A. The processing chamber 86 is configured to accommodate the boat 36 .

A cylindrical heater 88 is provided outside the reaction tube 84 so as to surround the side wall surface of the reaction tube 84 . Heater 88 is mounted vertically by being supported by heater base 90 .

A cylindrical furnace opening (manifold) 92 is disposed below the outer reaction tube 84B so as to be concentric with the outer reaction tube 84B. The furnace throat 92 is provided to support the lower end of the inner reaction tube 84A and the lower end of the outer reaction tube 84B, and is engaged with the lower end of the inner reaction tube 84A and the lower end of the outer reaction tube 84B. are doing.

An O-ring 94 as a sealing member is provided between the furnace port 92 and the outer reaction tube 84B. By supporting the furnace port 92 on the heater base 90, the reaction tube 84 is vertically installed. The reaction tube 84 and the furnace port 92 form a reaction container.

A processing gas nozzle 96 A and a purge gas nozzle 96 B are connected to the furnace opening 92 so as to communicate with the processing chamber 86 . A processing gas supply pipe 66A is connected to the processing gas nozzle 96A. A processing gas supply source (not shown) and the like are connected to the upstream side of the processing gas supply pipe 66A via the MFC 64A. A purge gas supply pipe 66B is connected to the purge gas nozzle 96B. A purge gas supply source (not shown) and the like are connected to the upstream side of the purge gas supply pipe 66B via the MFC 64B.

An exhaust pipe 68 for exhausting the atmosphere of the processing chamber 86 is connected to the furnace port 92 . The exhaust pipe 68 is arranged at the lower end of a cylindrical space 98 formed by a gap between the inner reaction pipe 84A and the outer reaction pipe 84B and communicates with the cylindrical space 98. As shown in FIG. A pressure sensor 70, a pressure regulator 72, and a vacuum pump 74 are connected to the downstream side of the exhaust pipe 68 in this order from the upstream side.

A disk-shaped lid 42 capable of air-tightly closing the lower end opening of the furnace throat 92 is provided below the furnace throat 92 . An O-ring 100 is provided as a contact sealing member.

A rotating mechanism 46 for rotating the boat 36 is installed on the opposite side of the processing chamber 86 near the center of the lid 42 . A rotating shaft 102 of the rotating mechanism 46 penetrates the lid 42 and supports the boat 36 from below. Further, the rotation mechanism 46 has a built-in rotation motor 46A. By rotating the rotation shaft 102 of the rotation mechanism 46 with the rotation motor 46A and rotating the boat 36, the substrate 16 is rotated. ing.

The lid 42 is configured to be lifted vertically by a boat elevator 38 provided outside the reaction tube 84 . The boat 36 can be transported to the processing chamber 86 by raising and lowering the lid 42 . A transport controller 48 is electrically connected to the rotation motor 46A of the rotation mechanism 46 and the boat elevator 38 .

The boat 36 is configured to hold a plurality of substrates 16 in a horizontal posture and in a multi-tiered state in which the centers of the substrates are aligned with each other. In addition, a plurality of disc-shaped heat insulating plates 104 as heat insulating members are arranged in a horizontal posture in a multi-tiered manner at the bottom of the boat 36 . The boat 36 and the heat insulating plate 104 are made of a heat-resistant material such as quartz or silicon carbide. The heat insulating plate 104 is provided to make it difficult for the heat from the heater 88 to be transferred to the furnace mouth portion 92 side.

A temperature sensor 106 as a temperature detector is installed in the reaction tube 84 . A temperature controller 76 is electrically connected to the heater 88 and the temperature sensor 106 .

<Operation of Substrate Processing Apparatus>
Next, a method of forming a thin film on the substrate 16 will be described as one step of the semiconductor device manufacturing process with reference to FIGS. 1 and 2. FIG. The operation of each part constituting the substrate processing apparatus 10 is controlled by a substrate processing apparatus controller 58 .

As shown in FIG. 1, when the pod 18 is supplied to the load port 20 by the in-process transfer device (not shown), the substrate detection sensor 56 detects the pod 18, and the pod loading/unloading port is opened by the front shutter (not shown). is released by Then, the pod 18 on the load port 20 is loaded into the housing 12 from the pod loading/unloading port by the pod transport device 24 .

The pod 18 carried into the housing 12 is automatically transferred by the pod transfer device 24 onto the shelf plate of the rotary pod shelf 22 and temporarily stored. After that, the pod 18 is transferred from the shelf board onto the mounting table of one of the pod openers 26 . The pod 18 carried into the housing 12 may be directly transferred onto the mounting table of the pod opener 26 by the pod transfer device 24 .

The lid of the pod 18 mounted on the mounting table is removed by the cap attaching/detaching mechanism 30 to open the substrate loading/unloading port. After that, the substrate 16 (see FIG. 2) is picked up from the pod 18 through the substrate loading/unloading port by the tweezers of the substrate transfer device 34A, and after its orientation is aligned by a notch aligning device (not shown), it is placed in the rear of the transfer chamber 32. is carried into the waiting section 50 located in the carriage and loaded into the boat 36 . Then, the substrate transfer device 34A returns to the mounting table on which the pod 18 is mounted, takes out the next substrate 16 from the pod 18, and loads the boat 36 with it.

During the operation of loading the substrates 16 into the boat 36 by the substrate transfer mechanism 34 in one of the pod openers 26 (upper or lower), another pod is placed on the mounting table of the other (lower or upper) pod opener 26. 18 are transported from the rotary pod shelf 22 by the pod transporter 24 . When the other pod 18 is transferred to the mounting table, the opening operation of the pod 18 by the pod opener 26 proceeds simultaneously.

When a predetermined number of substrates 16 are loaded into the boat 36, the lower end of the processing furnace 44 is opened by a furnace port shutter (not shown). Subsequently, the boat 36 holding the group of substrates 16 is carried (loaded) into the processing furnace 44 as the lid 42 is lifted by the boat elevator 38 .

As described above, when the boat 36 holding a plurality of substrates 16 is carried (loaded) into the processing chamber 86 of the processing furnace 44, the lid body 42 is pushed into the furnace via the O-ring 100 as shown in FIG. The lower end of the mouth portion 92 is sealed.

Thereafter, the processing chamber 86 is evacuated by the vacuum pump 74 to a desired pressure (degree of vacuum). At this time, based on the pressure value measured by the pressure sensor 70, (the opening of the valve of) the pressure adjustment unit 72 is feedback-controlled. A heater 88 heats the processing chamber 86 to a desired temperature. At this time, based on the temperature value detected by the temperature sensor 106, the amount of power supplied to the heater 88 is feedback-controlled. Subsequently, the boat 36 and the substrate 16 are rotated by the rotating mechanism 46 .

Next, the processing gas supplied from the processing gas supply source and controlled to have a desired flow rate by the MFC 64A flows through the processing gas supply pipe 66A and is introduced into the processing chamber 86 from the processing gas nozzle 96A. The introduced processing gas rises in the processing chamber 86, flows out into the tubular space 98 through the upper end opening of the internal reaction tube 84A, and is exhausted through the exhaust pipe 68. As shown in FIG. As the process gas passes through the process chamber 86, it contacts the surface of the substrate 16, causing a thermal reaction to deposit a thin film on the surface of the substrate 16. FIG.

When the preset processing time has elapsed, the purge gas supplied from the purge gas supply source and controlled to have a desired flow rate by the MFC 64B is supplied to the processing chamber 86, and the processing chamber 86 is replaced with an inert gas. At the same time, the pressure in the processing chamber 86 is returned to normal pressure.

After that, the lid 42 is lowered by the boat elevator 38 to open the lower end of the furnace throat 92 , and the boat 36 holding the processed substrates 16 moves from the lower end of the furnace throat 92 to the outside of the reaction tube 84 . It is carried out (unloaded). After that, the processed substrates 16 are taken out (discharged) from the boat 36 and stored in the pod 18 .

After the discharge, the pod 18 containing the processed substrates 16 is carried out of the housing 12 by almost the reverse of the procedure described above, except for the alignment process by the notch aligning device.

<Structure of Controller for Substrate Processing Apparatus>
Next, referring to FIG. 3, the substrate processing apparatus controller 58 as a main control section will be specifically described.

The substrate processing apparatus controller 58 mainly includes an arithmetic control unit 108 such as a CPU (Central Processing Unit), a RAM 110, a ROM 112, and a storage unit 114 having an HDD (not shown), an input unit 116 such as a mouse and a keyboard, and a monitor. etc., and a display unit 118 . Each data can be set by the arithmetic control unit 108, the storage unit 114, the input unit 116, and the display unit 118. FIG.

The arithmetic control unit 108 constitutes the core of the substrate processing apparatus controller 58, executes a control program stored in the ROM 112, and according to instructions from the input unit 116, recipes are stored in the storage unit 114, which also constitutes the recipe storage unit. A recipe (eg, a process recipe as a substrate processing recipe) is executed.

The ROM 112 is a recording medium composed of a flash memory, a hard disk, or the like, and stores an operation program or the like of the arithmetic control unit 108 that controls the operation of each member (for example, the vacuum pump 74 and the like) of the substrate processing apparatus 10 . Also, the RAM 110 (memory) functions as a work area (temporary storage unit) for the arithmetic control unit 108 .

Here, the substrate processing recipe (process recipe) is a recipe in which processing conditions, processing procedures, and the like for processing the substrate 16 are defined. In the recipe file, setting values and transmission timings to be transmitted to the transport controller 48, the temperature controller 76, the pressure controller 78, the gas supply controller 80, etc. are set for each step of the substrate processing recipe.

The arithmetic control unit 108 controls the temperature and pressure in the processing furnace 44, the flow rate of the processing gas introduced into the processing furnace 44, etc. so that the substrate 16 loaded in the processing furnace 44 is subjected to a predetermined process. has a function to control

The transport controller 48 controls the transport operations of the rotary pod shelf 22, the boat elevator 38, the pod transport device 24, the substrate transfer mechanism 34, the boat 36, and the rotation mechanism 46, which constitute the transport mechanism for transporting the substrate 16. is configured as

Sensors are built into the rotary pod shelf 22, the boat elevator 38, the pod transfer device 24, the substrate transfer mechanism 34, the boat 36, and the rotation mechanism 46, respectively. When each of these sensors indicates a predetermined value or an abnormal value, the substrate processing apparatus controller 58 is notified to that effect. A system for detecting signs of abnormality in each member of the substrate processing apparatus 10 will be described in detail later.

The storage unit 114 is provided with a data storage area 120 storing various data and the like, and a program storage area 122 storing various programs including substrate processing recipes. The data storage area 120 stores various parameters related to recipe files. The program storage area 122 stores various programs necessary for controlling the apparatus, including the substrate processing recipes described above.

Further, the display unit 118 of the substrate processing apparatus controller 58 is provided with a touch panel (not shown). The touch panel is configured to display an operation screen for accepting input of operation commands to the substrate transfer system and the substrate processing system. The substrate processing apparatus controller 58 may have a configuration including at least a display unit 118 and an input unit 116, like an operation terminal (terminal device) such as a personal computer or a mobile device.

The temperature controller 76 regulates the temperature inside the processing furnace 44 by controlling the temperature of the heater 88 of the processing furnace 44 . When the temperature sensor 106 indicates a predetermined value, an abnormal value, or the like, the substrate processing apparatus controller 58 is notified to that effect.

The pressure controller 78 controls the pressure adjustment section 72 based on the pressure value detected by the pressure sensor 70 so that the pressure in the processing chamber 86 reaches a desired pressure at a desired timing. When the pressure sensor 70 indicates a predetermined value, an abnormal value, or the like, the substrate processing apparatus controller 58 is notified to that effect.

The gas supply controller 80 is configured to control the MFCs 64A and 64B so that the flow rate of the gas supplied to the processing chamber 86 reaches a desired flow rate at a desired timing. When a sensor (not shown) included in the MFCs 64A, 64B or the like indicates a predetermined value or an abnormal value, the substrate processing apparatus controller 58 is notified to that effect.

<Substrate processing step>
Next, an outline of a substrate processing process for processing a substrate using the substrate processing apparatus 10 of this embodiment as a semiconductor manufacturing apparatus will be described with reference to FIG. This substrate processing step is, for example, one step of a method of manufacturing a semiconductor device (IC, LSI, etc.). In the following description, the operation and processing of each part constituting the substrate processing apparatus 10 are controlled by the substrate processing apparatus controller 58 .

Here, an example of forming a film on the substrate 16 by alternately supplying a source gas (first processing gas) and a reaction gas (second processing gas) to the substrate 16 will be described. Further, hereinafter, a silicon nitride (SiN) film is formed as a thin film on the substrate 16 using hexachlorodisilane (Si ₂ Cl ₆ , hereinafter abbreviated as HCDS) gas as a source gas and ammonia (NH ₃ ) as a reaction gas. An example will be described. For example, a predetermined film may be formed in advance on the substrate 16, and a predetermined pattern may be formed in advance on the substrate 16 or the predetermined film.

(Substrate loading step S102)
First, in the substrate loading step S102, the boat 36 is loaded with the substrates 16 and loaded into the processing chamber 86. As shown in FIG.

(Film formation step S104)
In the film forming step S104, a thin film is formed on the surface of the substrate 16 by sequentially executing the following four steps. During steps 1 to 4, the substrate 16 is heated to a predetermined temperature by the heater 88. Next, as shown in FIG.

[Step 1]
In step 1, the opening/closing valve (not shown) provided in the processing gas supply pipe 66A and the pressure adjustment unit 72 (APC valve) provided in the exhaust pipe 68 are opened to allow the HCDS gas whose flow rate is adjusted by the MFC 64A to flow into the processing gas supply pipe 66A. pass through Then, the HCDS gas is exhausted from the exhaust pipe 68 while being supplied to the processing chamber 86 from the processing gas nozzle 96A. At this time, the pressure in the processing chamber 86 is kept at a predetermined pressure. Thereby, a silicon thin film (Si film) is formed on the surface of the substrate 16 .

[Step 2]
In step 2, the opening/closing valve of the processing gas supply pipe 66A is closed to stop the supply of the HCDS gas. The pressure adjustment portion 72 (APC valve) of the exhaust pipe 68 is kept open, and the processing chamber 86 is evacuated by the vacuum pump 74 to remove residual gas from the processing chamber 86 . In addition, an open/close valve provided in the purge gas supply pipe 66B is opened to supply an inert gas such as N ₂ to the processing chamber 86 to purge the processing chamber 86. discharge to

[Step 3]
In step 3, an opening/closing valve (not shown) provided in the purge gas supply pipe 66B and a pressure adjustment unit 72 (APC valve) provided in the exhaust pipe 68 are both opened to allow the NH ₃ gas whose flow rate is adjusted by the MFC 64B to flow into the purge gas supply pipe. Pass through 66B. Then, the NH ₃ gas is exhausted from the exhaust pipe 68 while being supplied to the processing chamber 86 from the purge gas nozzle 96B. At this time, the pressure in the processing chamber 86 is kept at a predetermined pressure. As a result, the Si film formed on the surface of the substrate 16 by the HCDS gas reacts with the NH ₃ gas to form a SiN film on the substrate 16 .

[Step 4]
In step 4, the open/close valve of the purge gas supply pipe 66B is closed to stop the supply of _NH3 gas. The pressure adjustment portion 72 (APC valve) of the exhaust pipe 68 is kept open, and the processing chamber 86 is evacuated by the vacuum pump 74 to remove residual gas from the processing chamber 86 . Also, an inert gas such as _N2 is supplied to the processing chamber 86 to purge the processing chamber 86 again.

A SiN film having a predetermined thickness is formed on the substrate 16 by repeating steps 1 to 4 above as one cycle.

(Substrate Unloading Step S106)
In the substrate unloading step S106, the boat 36 on which the substrates 16 having the SiN films formed thereon are unloaded from the processing chamber 86. FIG.

<Control system in this embodiment>
Next, a control system for detecting signs of abnormality (signs of failure) of each member of the substrate processing apparatus 10 will be described with reference to FIGS. 5 and 6. FIG. An example in which a thin film is formed on the substrate 16 by the substrate processing apparatus 10 will be described below.

As shown in FIG. 5, the control system includes a substrate processing apparatus controller 58 as a main control section, a sign detection controller 82 as a sign detection section, various sensors 124, and a data collection unit (hereinafter referred to as Data Collection Unit). DCU) 126 and an Edge Controller (hereinafter abbreviated as EC) 128, which are connected by wire or wirelessly.

The substrate processing apparatus controller 58 is connected to a host computer (not shown) including a customer host computer, and an operating unit (not shown). The operation unit is configured to exchange various data (sensor data, etc.) acquired by the substrate processing apparatus controller 58 with the host computer.

The sign detection controller 82 acquires sensor data from sensors of various members provided in the substrate processing apparatus 10 and monitors the state of the substrate processing apparatus 10 . Specifically, the sign detection controller 82 uses data from various sensors 124 to calculate a numerical index and compares it with a predetermined threshold value to detect a sign of abnormality. The sign detection controller 82 incorporates a sign detection program for detecting signs of abnormality based on movement of sensor data.

The sign detection controller 82 has two systems, one directly connected to the substrate processing apparatus controller 58 and the other connected to the substrate processing apparatus controller 58 via the DCU 126 . Therefore, when a sign of abnormality is detected by the sign detection controller 82, a signal is output directly to the substrate processing apparatus controller 58 without going through the DCU 126 to generate an alarm, and an alarm is provided to the member where the sign of abnormality is recognized. It is possible to display the information of the sensor data obtained from the sensor on the screen of the display unit 118 (see FIG. 3).

The various sensors 124 are sensors (for example, the pressure sensor 70, the temperature sensor 106, etc.) provided in various members provided in the substrate processing apparatus 10, and measure the flow rate, concentration, temperature, humidity (dew point), Detect pressure, current, voltage, voltage, torque, vibration, position, rotation speed, etc.

The DCU 126 collects and accumulates data from various sensors 124 during execution of process recipes. In addition, the EC 128 once takes in sensor data as necessary depending on the type of sensor, performs processing such as Fast Fourier Transform (hereinafter abbreviated as FFT) on the raw data, and then transmits the data to the sign detection controller 82. .

Further, the various sensors 124 are divided into a first sensor system 124A and a second sensor system 124B having different transmission paths. The first sensor system 124A is a system that captures raw data in real time in units of 0.1 seconds, and raw data is sent from the first sensor system 124A to the sign detection controller 82 in real time via the substrate processing apparatus controller 58 and DCU 126. sent. The first sensor system 124A includes, for example, sensors such as a temperature sensor, a pressure sensor, and a gas flow rate sensor.

On the other hand, the second sensor system 124B is a system that extracts only the part necessary for analysis by applying processing such as FFT in the EC 128 and transmits data in a processed file format. The processed data is transmitted to the sign detection controller 82 . The second sensor system 124B includes sensors such as vibration sensors.

When the sensor is a vibration sensor, since vibration data is accumulated in units of milliseconds, the amount of data becomes enormous, and if the data is sent to the omen detection controller 82 as it is, the storage capacity of the omen detection controller 82 is consumed in large amounts. Since the data of this vibration sensor is finally processed by FFT etc. and used for analysis, by executing the processing in advance in the EC 128, the amount of information is reduced and the data format is easy to analyze. It can be sent to controller 82 .

(First embodiment)
A first embodiment of a process of detecting a sign of abnormality of each member of the substrate processing apparatus 10 using the control system described above will be specifically described below.

[Calculation of degree of anomaly]
First, by using multiple values detected by the sensors directly installed on the member to be detected as an anomaly sign and the values detected by the sensors of other members directly or indirectly affected by the state of the member, Abnormality" is calculated. In the present embodiment, for example, when a member for which an abnormality sign is detected approaches an abnormal state, the value of the degree of non-normality generally increases. It should be noted that the degree of non-normality may be configured to have a property that the value decreases when the member subject to detection of an abnormality sign approaches an abnormal state.

[Original data that constitutes the degree of anomaly]
The sequence of substrate processing includes, for example, carrying the substrate 16 into the processing chamber 86, vacuuming the processing chamber 86, raising the temperature, purging with an inert gas, waiting for the temperature to rise, processing the substrate 16 (for example, film formation), It consists of many events each having its own purpose, such as replacing the gas in the processing chamber 86, returning to atmospheric pressure, and unloading the substrate 16 after processing. Note that the above events are an example of the substrate processing sequence, and each event may be further divided into smaller parts.

In this embodiment, instead of using all the sensor data in the sequence, the values of one or more sensors at one or more specific events in these events are used as numerical indicators in the algorithm, "abnormal It is used as the original data for calculating the degree. In addition, the anomaly degree value for each Run is monitored to detect a sign of abnormality of each member of the substrate processing apparatus 10 . In this way, by using only the data of specific events, it is possible to save the amount of accumulated data.

For example, the detection of an abnormality sign of the vacuum pump 74 is easily detected at the timing when the vacuum pump 74 is heavily loaded. The step of reducing the pressure of the processing chamber 86 from the atmospheric pressure to a predetermined pressure, that is, the time when the vacuuming is started and the pressure band close to the atmospheric pressure for several minutes after the vacuuming is started correspond to the timing when the vacuum pump 74 is heavily loaded. .

Specifically, one substrate processing apparatus 10 is in charge of a plurality of processes, and there are cases where different process recipes, such as those with different film forming conditions, are mixed and started. Since the raw material gas flows during film formation on the substrate 16, the raw material gas may react or thermally decompose to form solid matter, which may put a load on the vacuum pump 74. Therefore, the film formation event is monitored. This is also effective for anomaly sign detection.

On the other hand, the evacuation event before substrate processing is often common even if subsequent substrate processing events are different. In other words, even if a plurality of recipes with different film formation conditions are started in the same apparatus, by monitoring the state at the start of the common evacuation in each run and acquiring sensor data, the process can be performed without depending on the details of substrate processing. Therefore, it is possible to know the change over time of the same state, and to make a highly accurate prediction.

[Calculation example of the degree of anomaly]
Here, when using the sensor data of the vibration sensor, and when using the sensor data of sensors other than the vibration sensor (for example, current sensor, temperature sensor, exhaust pressure sensor, torque value data, current data, etc.) Calculation examples are shown respectively.

First, when determining whether or not there is an abnormality for each individual frequency using the sensor data (vibration data) of the vibration sensor, the procedure is as follows.

(1) Vibration data (raw data) detected by a vibration sensor is acquired from among sensor data in a specified step among steps constituting a process recipe.
(2) Convert the acquired vibration data into a vibration frequency spectrum by processing such as FFT, and extract frequencies in a predetermined range (for example, 10 Hz to 5000 Hz) of the converted vibration frequency spectrum at predetermined frequency intervals (for example, every 10 Hz). (The numerical value is the amplitude (envelope) of the vibration, which is 500-dimensional in the example).
(3) For each extracted frequency, calculate the average value μ and standard deviation σ of the amplitude of the vibration frequency spectrum using the data for a predetermined number of times of the normal process recipe (for example, 30 Runs). Assuming that it follows a normal distribution N(μ, σ), this is the normal model.
(4) Using the numerical value of (2) after creating the normal model as a non-normality vector, comparing the amplitude value of the normal model with a predetermined threshold value for each extracted frequency, If the amplitude value of the frequency of () or more) is outside the threshold, it is determined that there is a sign of abnormality (presence of abnormality). Note that the threshold value is calculated, for example, by using the average value μ and the standard deviation σ obtained in (3), and adding or subtracting the value obtained by multiplying the standard deviation σ by 3 to the average value μ (μ±3σ).

Further, when the sensor data (vibration data) of the vibration sensor is used and the sum of the amplitudes of each frequency is used for determination, the procedure is as follows.

(1) Vibration data (raw data) detected by a vibration sensor is acquired from among sensor data in a specified step among steps constituting a process recipe.
(2) Convert the acquired vibration data into a vibration frequency spectrum by processing such as FFT, and extract frequencies in a predetermined range (for example, 10 Hz to 5000 Hz) of the converted vibration frequency spectrum at predetermined frequency intervals (for example, every 10 Hz). (The numerical value is the amplitude (envelope) of the vibration, which is 500-dimensional in the example).
(3) Add all the extracted sums of amplitudes for each frequency in each normal run (since one amplitude sum is obtained for each run, 30 numbers are obtained for 30 runs).
(4) Calculate the average μ and standard deviation σ from the numerical value group obtained for each Run, assume that the sum obtained for each Run follows the normal distribution N (μ, σ), and use this as a normal model .
(5) The value of (3) after creating the normal model is used as the degree of non-normality, and the amplitude value of the normal model is compared with a predetermined threshold value. has occurred). Note that the threshold value is calculated, for example, by using the average value μ and the standard deviation σ obtained in (3), and adding or subtracting the value obtained by multiplying the standard deviation σ by 3 to the average value μ (μ±3σ).

Moreover, when making a determination for each basic statistic using sensor data other than vibration data, the following procedure is used.

(1) Select one or more data from the basic statistics of the average value, standard deviation, N-quantile, maximum value, and minimum value of the sensor data of the target event during normal operation.
(2) Determine the average μ and standard deviation σ for each statistic of the selected normal basic statistic, and assume that each basic statistic follows a normal distribution. Let this be a normal model for each basic statistic of the sensor.
(3) After the normal model is created, the value of (1) is defined as the degree of non-normality, and if the value of each basic statistic exceeds a predetermined threshold value, it is determined that there is a sign of abnormality. Note that the threshold value is calculated, for example, using the average value μ and the standard deviation σ obtained in (2), and is calculated within a range (μ±3σ) by adding or subtracting the value obtained by multiplying the standard deviation σ by 3 to or from the average value μ.

Moreover, as shown in FIG. 6, when making a determination using singular spectrum conversion using sensor data other than vibration data, the procedure is as follows. In the following procedure, two data matrices X and Z are created on the past and present sides using the partial time series of window width n around Run p. The following procedure is a general practice for singular spectral transformations.

(1) Considering each of them as an M-dimensional vertical vector, prepare an Mn-dimensional vertical vector formed by vertically connecting n vertical vectors from the top S (pn+1, 1) to the bottom S (p, M). .
Sensor data at times 1, 2, . . . , M of the target event of Run pn+1 n+1, M)}
・・・
Sensor data at times 1, 2, . 1, M)}
Sensor data at times 1, 2, .
(2) Considering each of them as an M-dimensional vertical vector, n vertically connected vertical vectors ((1 ) is shifted to a Run group one older than ).
Sensor data at times 1, 2, . n, M)}
・・・
Sensor data at times 1, 2, . 2, M)}
Sensor data at times 1, 2, . 1, M)}
(3) Similar to the above (1) and (2), K vertical vectors are prepared in order, and the vertical vectors are arranged from left to right from oldest to newest to form an Mn×K-dimensional Create a matrix X(p). Thus, a history matrix for performing singular spectrum transformation is created.
(4) Consider each of them as an M-dimensional vertical vector, and prepare an Mn-dimensional vertical vector formed by connecting n vertically from the top S (p+L, 1) to the bottom S (p+L-n+1, M). . Note that L is a positive integer.
Sensor data at time 1, 2, .
・・・
Sensor data at times 1, 2, . . . , M of the target event of Run p+L-n+2 n+2, M)}
Sensor data at times 1, 2, . . . , M of the target event of Run p+L-n+1 n+1, M)}
(5) Considering each as an M-dimensional vertical vector, an Mn-dimensional vertical vector ( (4) shifted to a Run group one older than (4)) is prepared.
Sensor data at times 1, 2, . . . , M of the target event of Run p+L-1 1, M)}
・・・
Sensor data at times 1, 2, . . . , M of the target event of Run p+L-n+1 n+1, M)}
Sensor data at times 1, 2, . . . , M of the target event of Run p+L-n n, M)}
(6) As in (4) and (5) above, prepare R vertical vectors arranged in order, and arrange the vertical vectors from the oldest to the newest from left to right. Create a matrix Z(p). The test matrix for the singular spectrum transformation is created above.
(7) Perform singular value decomposition on the above matrix X(p) and matrix Z(p) to perform singular spectrum transformation.
(8) Select r left singular vectors in X(p) and m in Z(p) from the left singular vectors obtained by the singular value decomposition, construct U(r) and Q(m) and matrices respectively, and Find the largest singular value of the product U(r) ^T Q(m). Let it be λ (0≦λ≦1), and let 1−λ be the degree of abnormality (degree of change). If the degree of non-normality exceeds a predetermined threshold value, it is determined that there is an anomaly sign.

[Determination of anomaly signs using the degree of anomaly]
Moreover, as a method of determining whether or not there is a sign of abnormality using the degree of anomaly, for example, the following method is conceivable. When it is determined that there is a sign of abnormality, the substrate processing apparatus controller 58 is notified.

(1) A method of determining that there is a sign of abnormality when the degree of abnormality of at least one sensor data exceeds a threshold.
(2) A method of determining that there is an anomaly sign when the degree of anomaly of two or more sensor data is out of the threshold.
(3) A method of determining that there is an anomaly sign when the degree of abnormality of one or more sensor data exceeds a threshold a predetermined number of times (for example, three times).
(4) A method of judging that there is a sign of abnormality when the degree of abnormality of sensor data other than vibration data continuously deviates from the threshold for a predetermined number of times (for example, three times).
(5) A method of not determining that there is an anomaly sign if the degree of abnormality of the vibration data does not deviate from the threshold, even if the degree of abnormality of the sensor data other than the vibration data does not deviate from the threshold.
(6) A method of determining that there is a sign of abnormality when both the degree of abnormality of vibration data and the degree of abnormality of sensor data other than vibration data are outside the threshold.

For example, in the above methods (2), (5), and (6), a plurality of pieces of sensor data are used to determine signs of abnormality, so erroneous sensor detection can be reduced. In addition, since the movement of the degree of abnormality is not necessarily monotonous, the above methods (3) and (4) can reduce erroneous judgments when the value of the degree of abnormality fluctuates around the threshold. The formula for calculating the degree of anomaly, the threshold value, and the program differ for each component and each device, and are incorporated in advance into the sign detection controller 82 .

[Display of analysis screen for anomaly sign detection]
The analysis screen of the abnormality sign detection can be displayed on the display unit 118 (see FIG. 3) of the substrate processing apparatus controller 58 . For this reason, it is possible to visually observe the transition of the degree of abnormality, the threshold value, the number of times the threshold value is exceeded, and the like, so that the state of the member can be confirmed by the degree of abnormality.

[Cases involving EC]
Here, the case of the second sensor system 124B shown in FIG. 5, that is, the case where the EC 128 intervenes between the sensor and the portent detection controller 82 will be described.

[Time synchronization]
Since the data of the vibration sensor is converted by the EC 128, it is sent to the sign detection controller 82 with the time of the EC 128. In order to simultaneously use the vibration sensor data and other sensor data having the time on the DCU 126 and substrate processing apparatus controller 58 side for analysis, it is necessary to synchronize the times of both for analysis. Therefore, the EC 128, the DCU 126, and the sign detection controller 82 periodically acquire the time using the time of the substrate processing apparatus controller 58 as a reference time, and synchronize the time. This synchronizes the time of all members and enables accurate analysis.

Here, the method of detecting a sign of abnormality in the members of the substrate processing apparatus 10 will be specifically described, taking the vacuum pump 74 (see FIG. 2) as an example.

In the processing chamber 86 of the substrate processing apparatus 10, reaction by-products of the processing gas accumulate inside, and when the amount and height of the reaction by-products reach a certain level, the rotation of the vacuum pump 74 suddenly stops. .

Here, at least one sensor data of the current data, temperature data, exhaust pressure data, and vibration data of the vacuum pump 74 is continuously monitored, and changes in the behavior of the sensor data are detected by the sign detection program in the sign detection controller 82. By analyzing, it is possible to detect signs of abnormality of the vacuum pump 74 . When a sign of abnormality is detected, the information is transmitted to the substrate processing apparatus controller 58, and the operator is notified to replace the vacuum pump 74 or perform maintenance.

(Second embodiment)
Next, a second embodiment of the abnormality sign detection process for each member of the substrate processing apparatus 10 using the control system described above will be specifically described. The configuration of the sign detection controller 82 and the like, and abnormality sign determination using the degree of abnormality are the same as in the first embodiment.

[Calculation of degree of anomaly]
In this embodiment, a plurality of sensor values installed on a member to be detected as an anomaly sign detection and sensor values of other members directly or indirectly affected by the state of the member are used to It learns sensor data and calculates "abnormality" using learned data and data during operation.

In the present embodiment, for example, when a member for which an abnormality sign is detected approaches an abnormal state, the value of the degree of non-normality generally increases. It should be noted that the degree of non-normality may be configured to have a property that the value decreases when the member subject to detection of an abnormality sign approaches an abnormal state.

Here, the method of detecting a sign of abnormality in the members of the substrate processing apparatus 10 will be specifically described, taking the vacuum pump 74 (see FIG. 2) as an example.

In general, when the processing chamber 86 is being evacuated by the vacuum pump 74, an inert gas or a film-forming gas flows into the vacuum pump 74 and the load is high, which makes it easy to detect signs of abnormality. . On the other hand, when the vacuum pump 74 does not evacuate the processing chamber 86, the load on the vacuum pump 74 is small, making it difficult to detect a sign of an abnormality or to cause an abnormality. Therefore, conventionally, the vacuum pump 74 is monitored while the processing chamber 86 is being evacuated.

In contrast, in the present embodiment, in the event that the processing chamber 86 is not evacuated by the vacuum pump 74 and the substrate 16 is not in the processing chamber 86 , a large amount of gas is intentionally supplied to the vacuum pump 74 . to increase the load on the vacuum pump 74. By monitoring current data, vibration data, temperature data, back pressure data, etc. of the vacuum pump 74 in this state, signs of abnormality can be easily detected.

Thus, by applying a load to the vacuum pump 74 while the processing chamber 86 is not vacuumed, even if the vacuum pump 74 stops when the load is applied, the substrate 16 can be prevented from being damaged. can be done. In addition, when the vacuum pump 74 stops when a load is applied while the processing chamber 86 is not evacuated, it is considered that the vacuum pump 74 was in a state immediately before failure. As a result, it is possible to avoid a situation in which the processing chamber 86 is evacuated, that is, a situation in which the vacuum pump 74 stops during substrate processing.

(Third embodiment)
Next, a specific description will be given of a third embodiment of a process of detecting signs of abnormality in each member of the substrate processing apparatus 10 using the control system described above. The configuration of the sign detection controller 82 and the like, and abnormality sign determination using the degree of abnormality are the same as in the first and second embodiments.

In the present embodiment, when replacement or maintenance is performed on a member to be detected as an abnormality predictor detection target, a normal model after replacement or maintenance is created, and the substrate processing apparatus 10 is monitored based on the normal model to detect an abnormality predictor. make judgments.

In the present embodiment, replacement or maintenance of a member subject to abnormality sign detection is automatically or semi-automatically detected. For example, if a component for which an abnormality sign is to be detected has cumulative operating time information, component replacement can be detected using the cumulative operating time information. Since the accumulated operating time of the member subject to abnormality sign detection is normally stored in a non-volatile storage medium, the operating time is accumulated until the member is replaced, and the operating time is reset by the replacement. Therefore, it is possible to monitor the accumulated operating time of the member for which an abnormality sign is to be detected, and to detect that the member has been replaced when the accumulated operating time has decreased.

Specifically, the substrate processing apparatus controller 58 transmits the accumulated operating time of the member for which an abnormality sign is detected to the sign detection controller 82 at predetermined time intervals, and the sign detection controller receives the newly sent accumulated operating time. , is shorter than the accumulated operating time stored before. If the determination is affirmative, it can be determined that the member has been replaced.

In addition, even if the member subject to abnormality sign detection does not have the accumulated operating time information instead of the accumulated operating time information, it is possible to detect the replacement of the member by using the work of disconnecting the signal connector at the time of member replacement. When the signal connector is removed at the time of component replacement, the signal line of the component will be open (disconnected). Alternatively, the operator is urged to input confirmation as to whether or not maintenance has been performed. For example, other work cannot be started unless a confirmation input is made on the operation screen. This makes it possible to semi-automatically determine that the member has been replaced.

When it is determined that the member targeted for abnormality sign detection has been replaced or maintained, the sign detection controller 82 acquires new sensor data for the member targeted for abnormality sign detection as part of the sign detection processing. to update the normal model. Then, the degree of non-normality is calculated based on the updated normal model. Calculation of the degree of anomaly and detection of a sign by monitoring the value of the degree of anomaly can be performed in the same manner as in the first and second embodiments.

Here, a method for detecting a sign of abnormality in the substrate processing apparatus 10 will be described in detail, taking as an example the case where the vacuum pump 74 (see FIG. 2) is replaced.

When the substrate processing apparatus controller 58 acquires the operating time as sensor data of the vacuum pump 74 until the vacuum pump 74 is replaced, the substrate processing apparatus controller 58 accumulates the acquired operating time of the vacuum pump 74. time) is configured to reset. Further, the substrate processing apparatus controller 58 is connected to a portent detection controller 82 as shown in FIG. 5, and transmits the integrated operation time to the portent detection controller 82 at predetermined time intervals. The accumulated operating time of the vacuum pump 74 (accumulated operating time) can be managed by the sign detection controller 82 by directly acquiring the operating time from the vacuum pump 74 .

As part of the sign detection process, the sign detection controller 82 acquires the accumulated operation time transmitted from the substrate processing apparatus controller 58 (S10), as shown in FIG. It is determined whether the time is shorter than the time (S12), and if the determination is affirmative, it is determined that the vacuum pump 74 has been replaced, and the sensor data necessary for creating a normal model after the replacement is acquired ( S14). For example, sensor data for a predetermined number of times of the process recipe (for example, 30 runs) are acquired. Then, a normal model is created based on the acquired sensor data (S15). For example, using sensor data for a predetermined number of times of the process recipe, obtain the average value μ and standard deviation σ, assume that each sensor data in normal times follows the normal distribution N (μ, σ), and use this as a normal model . Based on the obtained normal model, the degree of abnormality is calculated (S16), and the previously stored data of the degree of abnormality is rewritten with the calculated degree of abnormality (S17). Then, the substrate processing apparatus 10 is monitored (S18), and abnormality sign determination is performed. Calculation of the degree of anomaly and monitoring of the value of the degree of anomaly can be performed in the same manner as in the first and second embodiments.

According to the present embodiment, a new normal model is created after replacement or maintenance of a member targeted for abnormality sign detection. can be detected). In addition, since the replacement or maintenance of the member subject to abnormality sign detection is detected automatically or semi-automatically, it is possible to appropriately change the abnormal value to be monitored as required.

(Fourth embodiment)
Next, a specific description will be given of a fourth embodiment of the abnormality sign detection process for each member of the substrate processing apparatus 10 using the control system described above. The configuration of the sign detection controller 82 and the like, and abnormality sign determination using the degree of abnormality are the same as in the first to third embodiments.

In the present embodiment, when replacement or maintenance is performed on a member to be detected as an abnormality sign detection target, before creating a new normal model after replacement or maintenance, a new normal model is created or a replacement or maintenance model is created. Determine whether to continue using the previous normal model. The automatic or semi-automatic detection of the replacement or maintenance of the member for which the abnormality sign is to be detected is performed in the same manner as in the third embodiment.

Specifically, when it is determined that there has been replacement or maintenance of a member for which anomaly sign detection is to be performed, the sign detection controller 82 detects a sensor with a smaller amount of data than is required to create a normal model. Get data. Then, based on the acquired sensor data, it is determined whether or not the normal model before replacement or maintenance can continue to be used.

When it is determined that the normal model before replacement or maintenance can be used continuously, the normal model before replacement or maintenance is used without acquiring the sensor data required for creating the normal model. Therefore, it is not necessary to calculate the degree of anomaly, and the same anomaly degree value as before replacement or maintenance is monitored to detect a sign.

If it is determined that the normal model before replacement or maintenance cannot continue to be used, further acquisition of sensor data is performed to obtain the sensor data required to create a normal model and create a new normal model. Then, based on the new normal model, the degree of anomaly is calculated, and the new value of the degree of anomaly is monitored to perform sign detection.

Here, as a specific example, a method for detecting a sign of abnormality of the substrate processing apparatus 10 will be specifically described by taking as an example the case where the vacuum pump 74 (see FIG. 2) is replaced.

[Concrete example]
As shown in FIG. 8, the sign detection controller 82 acquires the accumulated operation time transmitted from the substrate processing apparatus controller 58 (S30), and determines whether it is shorter than the previously stored accumulated operation time. It is judged (S32), and if the judgment is affirmative, it is judged that the vacuum pump 74 has been replaced, and sensor data (judgment sensor data) is acquired (S33). The amount of sensor data for judgment is less than the amount of sensor data for a predetermined number of times (for example, 30 runs) of the process recipe required to create a normal model. It's becoming Then, it is statistically determined whether or not the distribution of the acquired determination sensor data is equal to the data distribution of the normal model before replacement, and it is determined whether or not the normal model before replacement can be used (S34).

Statistical judgment can be performed as follows as an example.
(1) normality is determined by the Shapiro-Wilk test for the data group before exchange and the data group after exchange,
(2) Determine whether the variances of the data group before exchange and the data group after exchange are equal by F test,
(3) Based on the results of (1) and (2) above, the difference between the mean values (representative values) is tested by Student's t-test, Wilch's t-test, or Mann-Whitney U-test.

When the distribution of the acquired sensor data is equal to the data distribution of the normal model before replacement, it is determined that the normal model before replacement can be used (Y), and the sensor data required to create the normal model. (S39) to detect a sign by monitoring the non-normality value of the normal model before replacement.

If the distribution of the acquired sensor data is not equal to the data distribution of the normal model before replacement, it is determined that the normal model before replacement cannot be used (N), sensor data is further acquired (S35), and the normal model is A new normal model is created by obtaining sensor data required to create a model (S36). Based on the obtained normal model, the degree of abnormality is calculated (S37), and the previously stored data of the degree of abnormality is rewritten with the calculated degree of abnormality (S38). Then, the substrate processing apparatus 10 is monitored (S39), and abnormality sign determination is performed. Calculation of the degree of anomaly and monitoring of the value of the degree of anomaly can be performed in the same manner as in the first and second embodiments.

According to the present embodiment, it is determined whether or not a normal model before replacement or maintenance of a member can be used by acquiring a smaller amount of sensor data than required to create a normal model for the member for which anomaly sign detection is to be performed. do. Therefore, the creation of the normal model can shorten the time during which the monitoring for detecting an anomaly sign is stopped.

(action, effect)
According to the above-described embodiment, the substrate processing apparatus 10 includes a control system for detecting a sign of abnormality of a member. Therefore, when the sign of abnormality of a member is detected by the control system, the member can be replaced or maintained. can be done. In particular, with respect to failure sign detection of the vacuum pump 74, by continuously monitoring sensor data such as current data, temperature data, exhaust pressure data, and vibration data of the vacuum pump 74, the accuracy of anomaly sign can be increased. becomes possible.

As a result, it is possible to take measures such as replacing the member before it breaks down, and to use the member until just before it breaks down, so that the frequency of replacement can be reduced. In addition, by preventing failures during substrate processing, it is possible to improve the operating rate of the apparatus, prevent a decrease in the yield of products (substrates 16), and reduce unnecessary maintenance time.

Further, according to the above embodiment, the sign detection controller 82 for detecting an abnormality sign is connected to the substrate processing apparatus controller 58 . Therefore, it is possible to obtain and analyze data limited to a specific substrate processing sequence in which signs of abnormality are easily detected.

Further, even after replacement and maintenance of the member for which anomaly sign detection is to be performed, an appropriate normal model can be used to detect an anomaly sign of the member for which an anomaly sign detection is to be performed.

(Other embodiments)
Although the embodiments of the present disclosure have been specifically described above, the present disclosure is not limited to the above-described embodiments, and can be variously modified without departing from the scope of the present disclosure.

For example, in the above-described embodiments, an example of forming a thin film on the substrate 16 has been described. However, the present disclosure is not limited to such an aspect, and for example, when a thin film or the like formed on the substrate 16 is subjected to oxidation treatment, diffusion treatment, annealing treatment, etching treatment, or the like, it is also suitable. applicable to

Further, in the present embodiment, an example of forming a thin film using the substrate processing apparatus 10 having the hot wall type processing furnace 44 has been described, but the present disclosure is not limited to this, and the cold wall type processing furnace is provided. It can also be suitably applied when forming a thin film using a substrate processing apparatus. Furthermore, in the above-described embodiment, an example of forming a thin film using the batch-type substrate processing apparatus 10 that processes a plurality of substrates 16 at once has been described, but the present disclosure is not limited to this.

In addition, the present disclosure is not limited to semiconductor manufacturing apparatuses for processing semiconductor substrates, such as the substrate processing apparatus 10 according to the above-described embodiment, but can also be applied to LCD (Liquid Crystal Display) manufacturing apparatuses for processing glass substrates. can be done.

10 Substrate processing apparatus 16 Substrate 58 Substrate processing apparatus controller (an example of a main control unit)
74 vacuum pump 82 sign detection controller (an example of sign detection unit)
86 Processing chamber μ Mean value σ Standard deviation

Claims (12)

A substrate processing apparatus that acquires sensor data relating to a member to be detected as an abnormality sign detection target, creates a normal model, and monitors the state of the apparatus based on the normal model,
Acquiring the sensor data and recreating the normal model from the sensor data after replacement or maintenance of the member subject to the abnormality sign detection,
a sign detection unit configured to monitor the state of the device based on the normal model and detect a sign of abnormality before the device stops abnormally;
The sign detection unit uses the normal model before replacement or maintenance as the normal model after replacement or maintenance based on the sensor data having a data amount smaller than the amount of data required to create the normal model. determine whether it is possible
Substrate processing equipment. The sign detection unit is
When it is determined that the normal model before replacement or maintenance can be used as the normal model after replacement or maintenance, the normal model before replacement or maintenance is used as the normal model after replacement or maintenance. ,
When it is determined that the normal model before replacement or maintenance cannot be used as the normal model after replacement or maintenance, the sensor data is acquired and the normal model after replacement or maintenance is created from the sensor data. do,
The substrate processing apparatus according to claim 1 . A substrate processing apparatus that acquires sensor data relating to a member to be detected as an abnormality sign detection target, creates a normal model, and monitors the state of the apparatus based on the normal model,
Acquiring the sensor data and recreating the normal model from the sensor data after replacement or maintenance of the member subject to the abnormality sign detection,
a sign detection unit configured to monitor the state of the device based on the normal model and detect a sign of abnormality before the device stops abnormally;
a main control unit that executes a process recipe including a plurality of steps to perform a predetermined process on the substrate;
The sign detection unit is
Collecting the sensor data while executing the process recipe;
Acquiring, from the collected sensor data, vibration data detected by a vibration sensor from among the sensor data in a specified step among steps constituting the process recipe;
converting the obtained vibration data into a vibration frequency spectrum;
The converted vibration frequency spectrum is extracted at predetermined frequency intervals, and the average value and standard deviation of the amplitude of the vibration frequency spectrum are calculated for each extracted frequency using data for a predetermined number of times of the process recipe under normal conditions. and creating a normal model using the average value and standard deviation of the amplitude of the vibration frequency spectrum obtained,
Substrate processing equipment. 4. The substrate processing apparatus according to claim 3 , wherein said specifying step is a step of reducing the pressure of said processing chamber for processing said substrate from atmospheric pressure to a predetermined pressure. The sign detection unit is
The normal model is created using the average value and standard deviation of the amplitude of the vibration frequency spectrum, and the amplitude value of the normal model is compared with a predetermined threshold for the extracted frequency, and a predetermined number or more frequencies If the amplitude value of is out of the threshold value, it is determined that there is an abnormality sign,
The substrate processing apparatus according to claim 3 . The predetermined threshold is calculated by adding or subtracting a value obtained by multiplying the standard deviation by three to the average value using the average value and the standard deviation,
The substrate processing apparatus according to claim 5 . The sign detection unit is
6. The substrate processing apparatus according to claim 5 , wherein when it is determined that there is a sign of abnormality, an alarm is generated and the sensor data of the member with a sign of abnormality is displayed on the screen. When the member to be detected for an abnormality sign is an exhaust device for exhausting the atmosphere of a processing chamber for processing substrates,
The sign detection unit is
At least one of the sensor data selected from the group consisting of vibration data detected by the vibration sensor, current data of the exhaust system, temperature data of the exhaust system, and exhaust pressure data of the exhaust system is acquired, and configured to create a normal model,
The substrate processing apparatus according to claim 3 . A method of manufacturing a semiconductor device having a substrate processing step of performing a predetermined processing on a substrate by executing a process recipe including a plurality of steps, comprising:
The substrate processing step includes a step of acquiring sensor data related to the member for abnormality sign detection after replacement or maintenance of the member for abnormality sign detection, and recreating a normal model;
a step of monitoring the state of the device based on the normal model and detecting signs of abnormality before the device stops abnormally;
has
The step of detecting a sign of anomaly includes determining the normal model before replacement or maintenance as the normal model after replacement or maintenance based on the sensor data having a data amount smaller than the amount of data required to create the normal model. A method of manufacturing a semiconductor device , the method being configured to determine availability of a model . A sign detection program executed in a substrate processing apparatus that acquires sensor data relating to a member to be subjected to abnormality sign detection, creates a normal model, and monitors the state of the apparatus based on the normal model, comprising:
a procedure of acquiring the sensor data after replacement or maintenance of the member subject to the abnormality sign detection, and recreating a normal model from the acquired sensor data;
a procedure of monitoring the state of the device based on the created normal model and detecting signs of abnormality of the device;
a procedure for determining whether the normal model before replacement or maintenance can be used as the normal model after replacement or maintenance based on the sensor data having a data amount smaller than the data amount required to create the normal model;
a sign detection program that causes the substrate processing apparatus to execute the above by a computer. A method of manufacturing a semiconductor device having a substrate processing step of performing a predetermined processing on a substrate by executing a process recipe including a plurality of steps, comprising:
The substrate processing step includes a step of acquiring sensor data related to the member for abnormality sign detection after replacement or maintenance of the member for abnormality sign detection, and recreating a normal model;
a step of monitoring the state of the device based on the normal model and detecting signs of abnormality before the device stops abnormally;
The step of detecting a sign of abnormality includes:
collecting the sensor data while executing the process recipe;
obtaining vibration data detected by a vibration sensor from among the sensor data in a designated step among the steps constituting the process recipe from the collected sensor data;
converting the obtained vibration data into a vibration frequency spectrum;
extracting the transformed vibrational frequency spectrum at predetermined frequency intervals;
calculating the average value and standard deviation of the amplitude of the vibration frequency spectrum for each extracted frequency using data for a predetermined number of times of the process recipe under normal conditions;
A step of creating a normal model using the obtained average value and standard deviation of the amplitude of the vibration frequency spectrum;
A method of manufacturing a semiconductor device having A sign detection program to be executed in a substrate processing apparatus that acquires sensor data relating to a member to be subjected to abnormality sign detection, creates a normal model, and monitors the state of the apparatus based on the normal model,
a procedure of acquiring the sensor data after replacement or maintenance of the member subject to the abnormality sign detection, and recreating a normal model from the acquired sensor data;
a procedure of monitoring the state of the device based on the created normal model and detecting signs of abnormality of the device;
has
In the procedure for detecting a sign of abnormality,
collecting the sensor data while executing a process recipe comprising multiple steps;
a procedure of obtaining, from the collected sensor data, vibration data detected by a vibration sensor among the sensor data in a specified step among the steps constituting the process recipe;
a procedure for converting the obtained vibration data into a vibration frequency spectrum;
extracting the transformed vibration frequency spectrum at predetermined frequency intervals;
a procedure of calculating the average value and standard deviation of the amplitude of the vibration frequency spectrum using data for a predetermined number of times of the process recipe in the normal state for each extracted frequency;
A procedure for creating a normal model using the obtained average value and standard deviation of the amplitude of the vibration frequency spectrum;
is executed by a computer in the substrate processing apparatus.