Monitoring method, computer-readable storage medium, monitoring device, and method of manufacturing a wafer

Abstract

The present application provides a kind of monitoring method, monitoring program, monitoring device, wafer manufacturing method and wafer capable of improving the quality of product.The monitoring method for monitoring the state of processing device (10) based on actual data obtained from wafer (W) processing includes the following steps: input actual data to analytical model representing the relationship between actual data and the state of processing device (10), obtain at least one statistical value representing the state of processing device (10) output from the analytical model; based on at least one statistical value, determine whether the state of processing device (10) is abnormal; in the case of determining that the state of processing device (10) is abnormal, specify the item that becomes the cause of abnormality from among the items of actual data as a cause item and output.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a monitoring method, a monitoring program, a monitoring device, a wafer manufacturing method, and a wafer. BACKGROUND

[0002] Conventionally, a method of accurately predicting quality management value data of a wafer is known (see Patent Literature 1, etc.).

[0003] PRIOR ART DOCUMENTS

[0004] PATENT LITERATURE

[0005] Patent Literature 1: Japanese Patent Application Publication No. 2007-242809 SUMMARY

[0006] PROBLEMS TO BE SOLVED BY THE INVENTION

[0007] A worker of a process cannot understand a specific response that should be performed on a processing device in order to improve the quality of a product based on only a prediction result of the quality of the product. By enabling the worker to understand a factor that affects the quality of the product and perform a specific response on the processing device, it is possible to improve the quality of the product.

[0008] Therefore, an object of the present disclosure is to provide a monitoring method, a monitoring program, a monitoring device, a wafer manufacturing method, and a wafer that can improve the quality of a product.

[0009] SOLUTION TO PROBLEM

[0010] One embodiment of the present disclosure that solves the above problem is as follows.

[0011] [1] A monitoring method of monitoring a state of a processing device that processes a block or a wafer based on actual data acquired from the processing device, the monitoring method comprising:

[0012] inputting the actual data to an analysis model that represents a relationship between the actual data and the state of the processing device, and acquiring at least one statistical value that represents the state of the processing device output from the analysis model;

[0013] determining whether the state of the processing device is abnormal based on the at least one statistical value; and

[0014] in a case where it is determined that the state of the processing device is abnormal, specifying a project that becomes a cause of the abnormality from among projects of the actual data as a cause project and outputting the cause project.

[0015] [2] The monitoring method according to the above [1], further comprising the step of determining that the state of the processing device is abnormal in a case where the at least one statistical value is a value outside a management range.

[0016] [3] The monitoring method according to the above [1] or [2], further comprising the step of:

[0017] estimating the quality of the wafer based on the at least one statistical value; and

[0018] determining that the state of the processing device is abnormal in a case where the estimation result of the quality of the wafer does not satisfy a quality criterion.

[0019] [4] The monitoring method according to any one of the above [1] to [3], further comprising the step of:

[0020] acquiring batch data in a period in which the processing device processes one batch;

[0021] dividing the batch data into a plurality of interval data in each interval after dividing the period in which the one batch is processed into a plurality of intervals; and

[0022] inputting the each interval data to the analysis model as the actual data, and acquiring at least one statistical value indicating the state of the processing device in the each interval.

[0023] [5] The monitoring method according to the above [4], further comprising the step of generating an interval model corresponding to the each interval as the analysis model using the processing data of the each interval as teaching data, and inputting the interval data to the interval model.

[0024] [6] The monitoring method according to the above [4] or [5], further comprising the step of determining whether the state of the processing device is abnormal based on at least one statistical value acquired by inputting interval data in an interval before the middle of the period in which the one batch is processed among the plurality of interval data to the analysis model.

[0025] [7] The monitoring method according to any one of the above [4] to [6], wherein,

[0026] the processing device is a wire saw device that cuts a block,

[0027] the monitoring method further comprises the step of determining whether the state of the processing device is abnormal based on at least one statistical value acquired by inputting interval data in an interval in which cutting of the block advances by a prescribed distance among the plurality of interval data to the analysis model.

[0028] [8] The monitoring method according to the above [7], wherein the prescribed distance is a length of a range in which a value of an item indicating a characteristic of a wafer obtained by cutting the block is calculated.

[0029] [9] The monitoring method according to the above [8], wherein the item indicating the characteristic of the wafer is an inclination angle of a cut surface at a start of cutting of the wafer obtained by cutting the block.

[0030]

[10] The monitoring method according to any one of the above [1] to [9], further comprising the steps of:

[0031] acquiring actual data in a period in which the processing device is idling as idling data; and

[0032] inputting the idling data to the analysis model to acquire at least one statistical value indicating a state of the processing device at the time of idling.

[0033]

[11] The monitoring method according to the above

[10] , further comprising the step of generating an idling model corresponding to the period in which the processing device is idling as the analysis model by using the idling data as teaching data.

[0034]

[12] A monitoring program, wherein the monitoring program causes a processor to execute the monitoring method according to any one of the above [1] to

[11] .

[0035]

[13] A monitoring device, wherein the monitoring device includes a control unit that executes the monitoring method according to any one of the above [1] to

[11] .

[0036]

[14] A wafer manufacturing method, comprising the step of processing a wafer by a processing device of a wafer that has been monitored by the monitoring method according to any one of the above [1] to

[11] .

[0037]

[15] A wafer, wherein the wafer is a wafer processed by a processing device of a wafer that has been monitored by the monitoring method according to any one of the above [1] to

[11] .

[0038] Effects of the Invention

[0039] According to the monitoring method, the monitoring program, the monitoring device, the wafer manufacturing method, and the wafer of the present disclosure, the quality of a product can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a block diagram showing a configuration example of a monitoring system of one embodiment of the present disclosure.

[0041] Figure 2 This is a diagram showing an example of the structure of a wire saw device as a processing apparatus.

[0042] Figure 3 This is a flowchart illustrating an example of the operation sequence for generating a parsing model in a monitoring method according to one embodiment of this disclosure.

[0043] Figure 4 This is a flowchart illustrating an example of a monitoring method according to one embodiment of the present disclosure, which uses an analytical model to determine the state of a processing device in a work sequence.

[0044] Figure 5 This is a diagram illustrating an example of dividing the period of cutting a block with a wire saw into 6 intervals.

[0045] Figure 6 This is a diagram illustrating an example of the surface shape of a wafer cut by a wire saw.

[0046] Figure 7 This is a graph showing an example of the Q statistics for each batch calculated based on interval data from the first interval.

[0047] Figure 8 This is a graph showing an example of the Q statistics for each batch calculated based on interval data from the second interval.

[0048] Figure 9 This is a graph showing an example of the Q statistics of the cause items extracted from batches where the Q statistics are above the threshold. Detailed Implementation

[0049] (Example of the structure of monitoring system 1)

[0050] like Figure 1 As shown, the monitoring system 1 includes a monitoring device 50 and a display device 60. The monitoring device 50 monitors the status of the processing apparatus 10 set up in the product manufacturing process. The processing apparatus 10 processes the product. The processing apparatus 10 has sensors that measure processing data representing the working status of the structural parts of the processing apparatus 10. The sensors of the processing apparatus 10 can measure the temperature or flow rate of auxiliary materials such as cooling water or slurry supplied to the processing apparatus 10, as well as environmental factors related to the process, such as the temperature or humidity of the process. The sensors output the processing data measured when the processing apparatus 10 is operating to the monitoring device 50. The monitoring device 50 obtains the processing data from the processing apparatus 10 and monitors the status of the processing apparatus 10 based on the processing data.

[0051] The state of the processing apparatus 10 includes a normal state and an abnormal state. The abnormal state corresponds to a state other than the normal state. The normal state corresponds to a state in which the post-processing quality of the wafer is not problematic and the state of the apparatus does not largely deviate. For example, in a case where one apparatus parameter is focused on, a state in which the amount of deviation from the average of the apparatus parameter during a period in which the post-processing quality of the wafer is not problematic is less than 3 times the standard deviation of the apparatus parameter during the period in which the post-processing quality of the wafer is not problematic is set as the normal state. Further, in a case where a plurality of apparatus parameters are used to generate a synthetic parameter such as a T2 statistic or a Q statistic, a state in which the amount of deviation from the average of the synthetic parameter calculated using the apparatus parameters during a period in which the post-processing quality of the wafer is not problematic is less than 3 times the standard deviation of the synthetic parameter during the period in which the post-processing quality of the wafer is not problematic is also set as the normal state.

[0052] In the present embodiment, it is assumed that the product processed by the processing apparatus 10 is a wafer. It is assumed that the monitoring target is set in a wafer manufacturing process. The processing apparatus 10 can be, for example, a wire saw apparatus, a lapping processing apparatus, an SMP (Single Side Polishing apparatus), a DSP (Double Side Polishing apparatus), a face polishing apparatus, or a heat treatment apparatus, or the like.

[0053] The monitoring apparatus 50 is provided with a control section 52. The control section 52 can include at least one processor. The processor can execute a program that realizes various functions of the control section 52. The processor can be realized as a single integrated circuit. The integrated circuit is also referred to as an IC (Integrated Circuit). The processor can be realized as a plurality of integrated circuits and discrete circuits that are communicably connected. The processor can also be realized based on other various known technologies.

[0054] The monitoring apparatus 50 is further provided with a storage section 54. The storage section 54 can include an electromagnetic storage medium such as a magnetic disk, and can include a semiconductor memory or a magnetic memory or the like. The storage section 54 can include a non-transitory computer readable medium. The storage section 54 stores various information such as processing data acquired from the processing apparatus 10, and a program or the like executed by the control section 52. The storage section 54 can function as a work memory of the control section 52. At least a part of the storage section 54 can be included in the control section 52. At least a part of the storage section 54 can also be configured as a storage apparatus separate from the monitoring apparatus 50.

[0055] The monitoring device 50 may also include a communication unit for transmitting and receiving data with the processing device 10 or the display device 60. The communication unit is communicatively connected to the processing device 10 or the display device 60. The communication unit may be communicatively connected to the processing device 10 or the display device 60 via a network. The communication unit may be wired or wirelessly connected to the processing device 10 or the display device 60. The communication unit may include a communication module connected to the network or the processing device 10 or the display device 60. The communication module may have a communication interface such as a LAN (Local Area Network). The communication module may also implement communication using various communication methods such as 4G or 5G. The communication methods performed by the communication unit are not limited to the examples described above and may include various other methods. At least a portion of the communication unit may be included in the control unit 52.

[0056] Display device 60 may include a display device that outputs visual information such as images, characters, or graphics. The display device may include, for example, an LCD (Liquid Crystal Display), an organic EL (Electro-Luminescence) display, an inorganic EL display, or a PDP (Plasma Display Panel). The display device is not limited to these displays and may also include various other types of displays. The display device may also include light-emitting devices such as LEDs (Light Emitting Diodes) or LDs (Laser Diodes). It is not limited to these and may also include various other devices.

[0057] (Structural example of the wire saw device as processing device 10)

[0058] In this embodiment, it is assumed that the processing device 10 is a wire saw device. For example... Figure 2 As shown, the processing apparatus 10, which is a wire saw, includes a wire bundle 16 that arranges wires 12 side-by-side and reciprocates between multiple rollers 14. The processing apparatus 10 includes a workpiece holding mechanism 18 that holds the workpiece W and moves it in a direction that pushes the workpiece W into the wire bundle 16. The processing apparatus 10 includes a pair of nozzles 20 that supply slurry to the area where the workpiece W is pushed into the wire bundle 16. The processing apparatus 10 cuts the workpiece W using the wire bundle 16. Assume the workpiece W is a block of silicon or the like (a single crystal ingot cut into a block shape). The processing apparatus 10 sends out the silicon or the like wafer obtained by cutting the workpiece W as the processed product. Hereinafter, it is assumed that the processed product is a wafer when the processing apparatus 10 is a wire saw.

[0059] The wire 12 is wound around a set of wire drums 38A and 38B. The wire 12 is pulled from one wire drum 38A to the other wire drum 38B via the guide roller 32 and the roller 14, and the like.

[0060] The wire drums 38A and 38B are rotated by the driving motor 36, respectively. The driving motor 36 is driven to rotate the wire drums 38A and 38B, whereby the wire 12 can be paid out from one wire drum 38A, travel to the other wire drum 38B via the guide roller 32 and the roller 14, and the like. The wire 12 travels via the tension imparting unit including the slack adjusting arm 33 and the slack adjusting roller 34, and the like. The wire 12 travels via the tension imparting unit, whereby the wire 12 is imparted with tension. The wire 12 travels via the contact roller 35. The contact roller 35 follows the position of the wire 12 that moves when being paid out from or wound up to the wire drums 38A and 38B.

[0061] The wire 12 is spirally wound multiple times across the plurality of rollers 14. The wire 12 that is spirally wound constitutes the wire group 16 that is arranged in parallel along a direction orthogonal to the roller axis direction X between the rollers 14. The roller 14 is structured such that polyurethane resin is pressed around a steel cylinder, and grooves are cut at a certain interval on the surface thereof. The wire 12 is embedded in the grooves cut on the surface of the roller 14, whereby the wire group 16 can travel stably.

[0062] The traveling direction of the wire 12 is controlled by the rotation direction of the driving motor 36. The wire 12 can be controlled to travel in one direction, or can be controlled to travel back and forth as needed. The magnitude of the tension imparted to the wire 12 can be set as appropriate. The traveling speed of the wire 12 can be set as appropriate.

[0063] The slurry supplied from the nozzle 20 to the wire group 16 is stored in the slurry tank 40, and is fed to the nozzle 20 from the slurry tank 40 via the slurry cooler 42 that adjusts the temperature of the slurry.

[0064] The sliced wafer that is fed out from the wire saw device is further processed in a polishing process or the like, and is shipped as a wafer as a final product.

[0065] (Working example of the monitoring device 50)

[0066] The monitoring device 50 statistically analyzes the processing data of the processing device 10 using multivariate analysis, and calculates a statistical value of the processing data. The monitoring device 50 monitors the state of the processing device 10 based on the calculated statistical value. Hereinafter, a working example of the monitoring device 50 will be described specifically.

[0067] <Generation of analysis model>

[0068] The control section 52 of the monitoring device 50 stores machining data during normal operation of the machining device 10 as teaching data in the storage section 54. The control section 52 generates an analysis model used for monitoring the state of the machining device 10 by statistically analyzing machining data of the machining device 10 based on the teaching data stored in the storage section 54. The analysis model accepts input of machining data of the machining device 10 as an analysis target. The machining data of the machining device 10 as an analysis target is also referred to as actual data. The analysis model is configured to output a statistical value calculated by statistically analyzing the input actual data. In other words, the analysis model represents a relationship between the actual data and the state of the machining device 10. Moreover, the analysis model can at least discriminate whether the machining device 10 is abnormal or not.

[0069] In the present embodiment, the analysis model is configured to calculate a statistical value of the input actual data based on a PLS (Partial Least Square regression) method. The control section 52 generates the analysis model based on the PLS method as described below.

[0070] The control section 52 selects an explanatory variable and an objective variable from each item of the machining data of the machining device 10 used as the teaching data. The control section 52 defines a latent variable (principal component) based on a relationship between a value of the explanatory variable and a value of the objective variable included in the teaching data. Specifically, the latent variable is defined as a new variable expressed by a relationship formula capable of calculating a value having a high correlation with the objective variable using the explanatory variable.

[0071] The control section 52 generates a linear regression formula representing a relationship between a value of the latent variable and a value of the objective variable by applying a least square method. The control section 52 inputs a value of the explanatory variable of which a relationship with the value of the objective variable is known in advance to the generated linear regression formula to obtain a calculation result of the linear regression formula. The control section 52 verifies the linear regression formula by calculating a difference between the calculation result of the linear regression formula and the value of the objective variable. The control section 52 repeatedly performs definition of the latent variable, generation of the linear regression formula, and confirmation of the difference in a manner that the difference between the calculation result of the linear regression formula and the value of the objective variable becomes smaller to verify the linear regression formula. The control section 52 can replace the latent variable by newly defining the latent variable in the verification of the linear regression formula or increase the number of the latent variables by adding the newly defined latent variable. The control section 52 determines the number of the latent variables and the linear regression formula in the verification of the linear regression formula in a manner that the difference between the calculation result of the linear regression formula and the value of the objective variable becomes smaller.

[0072] The control section 52 utilizes the decided linear regression formula as an analytical model. Specifically, the control section 52 calculates a T2 statistic value and a Q statistic value based on the decided linear regression formula and the actual data. The T2 statistic value represents the position of the actual data in the subspace defined by the latent variables as a statistic value, and is calculated as the square of the distance from the origin of the subspace to the point in the actual data corresponding to the value of the latent variables. The Q statistic value represents the position of the actual data in the orthogonal complement space of the subspace as a statistic value, and is calculated as the square of the length of the normal projection to the vector of the subspace from the point in the actual data corresponding to the value of the item that was not used in the calculation of the latent variables.

[0073] The T2 statistic value is a statistic value suitable for determining normality or abnormality. The Q statistic value is a statistic value suitable for determining variation that cannot be determined by the T2 statistic value. The control section 52, in the case of monitoring only the T2 statistic value, can determine that some kind of abnormality has occurred in the processing device 10 in the case where the T2 statistic value deviates from the management limit, but cannot analyze the content of the specific abnormality.

[0074] On the other hand, the control section 52, in the case of monitoring both the T2 statistic value and the Q statistic value, can determine that the correlation between the variables has not collapsed, and that no serious abnormality such as a device failure has occurred, even if the T2 statistic value deviates from the management limit, if the Q statistic value is within the management limit. Furthermore, as a reason why the T2 statistic value deviates from the management limit, the control section 52 can determine that the T2 statistic value has become large as a result of a change or a mistake in the processing conditions or the operation program, etc. set in the processing device 10. The control section 52, in the case where the Q statistic value deviates from the management limit, can determine that a serious abnormality such as a device failure has occurred.

[0075] With respect to the generated analytical model, in order to prevent overlearning of the analytical model and improve general performance, the control section 52 can perform cross-validation.

[0076] <Status determination based on T2 statistic value and Q statistic value>

[0077] The control unit 52 inputs actual data into the generated analytical model and calculates the T2 and Q statistics. The control unit 52 sets management ranges for both the T2 and Q statistics. The control unit 52 specifies the management range by setting an upper or lower management limit. If both the T2 and Q statistics are within the management range, the control unit 52 determines that the processing device 10 is in a normal state. If either the T2 or Q statistics are outside the management range, the control unit 52 determines that the processing device 10 is in an abnormal state or has a potential for abnormality. The control unit 52 outputs the determination result of the processing device 10's state to the display device 60. The display device 60 displays the determination result of the processing device 10's state and notifies the user. The user may include the operator or manager of the process.

[0078] When the Q-statistic value is outside the management range, the control unit 52 extracts the items from the actual data that affect the Q-statistic value (causing a significant change in the Q-statistic value). Items affecting the Q-statistic value can be considered as causes of abnormal changes in the state of the processing unit 10, and are also called cause items. The Q-statistic value calculated by inputting the actual data into the analytical model corresponds to the sum of the Q-statistic values ​​of each cause item. In other words, the Q-statistic value is divided into Q-statistic values ​​for each cause item.

[0079] It can be said that the larger the Q-statistic value of each cause item, the greater its impact on the Q-statistic value. The control unit 52 can rank the cause items based on their impact on the Q-statistic value. The control unit 52 can assign a higher ranking to cause items with larger Q-statistic values. The control unit 52 outputs the cause items to the display device 60. The display device 60 displays the cause items and notifies the user.

[0080] The user can determine the appropriate action for the processing device 10 based on the status determination result of the processing device 10 notified by the display device 60. For example, if the Q statistic is outside the management range, the user can confirm the actual status of the processing device 10 based on the notified reason. For example, if the T2 statistic is outside the management range, the user can confirm the processing conditions or operating procedures set in the processing device 10.

[0081] (Example of the operation sequence for monitoring methods)

[0082] The control unit 52 of the monitoring device 50 can perform actions including Figure 3 The illustrated flowchart uses a monitoring method to generate an analytical model of the work sequence. This monitoring method can also be implemented as a monitoring program executed by the control unit 52.

[0083] The control section 52 acquires machining data for a period during which the machining device 10 is normally operating as teaching data (step S1). The control section 52 generates an analysis model based on the teaching data (step S2).

[0084] The control section 52 determines whether the analysis model is appropriate (step S3). For example, the control section 52 can determine whether the difference between the calculation result of the linear regression expression and the value of the target variable is less than a prescribed value. The control section 52 returns to step S2 to regenerate the analysis model in a case where it is determined that the analysis model is not appropriate (step S3: No). The control section 52 decides the analysis model in a case where it is determined that the analysis model is appropriate (step S3: Yes) (step S4). The control section 52 ends the execution of the operation sequence of the flowchart of FIG. 10 after the execution of the operation sequence of step S4. Figure 3

[0085] The control section 52 can calculate the statistical value based on the actual data of the machining device 10 by executing the monitoring method including the operation sequence of the flowchart illustrated in FIG. 11, and monitor the state of the machining device 10. The monitoring method can also be implemented as a monitoring program that causes the control section 52 to execute. Figure 4

[0086] The control section 52 acquires the actual data from the machining device 10 (step S11). The control section 52 inputs the actual data to the analysis model, and acquires the statistical value calculated by the analysis model (step S12). Specifically, the control section 52 acquires the T2 statistical value and the Q statistical value as the statistical value.

[0087] The control section 52 determines whether the statistical value is within the management range (step S13). Specifically, the control section 52 determines whether both the T2 statistical value and the Q statistical value are within the management range. The control section 52 advances to the operation sequence of step S15 in a case where the statistical value is within the management range (step S13: Yes). The control section 52 analyzes the cause item in a case where the statistical value is not within the management range (step S13: No), that is, in a case where the statistical value is outside the management range (step S14).

[0088] The control section 52 outputs the determination result (step S15). The control section 52 outputs whether the statistical value is within the management range or outside the management range. The control section 52 outputs the analysis result of the cause item together in a case where the statistical value is outside the management range. The control section 52 ends the execution of the operation sequence of the flowchart of FIG. 12 after the execution of the operation sequence of step S15. Figure 4

[0089] (SUMMARY)

[0090] ​​​As described above, the monitoring device 50 of this embodiment monitors the state of the processing device 10 based on actual data obtained from the processing device 10. The monitoring device 50 inputs the actual data into an analytical model. The analytical model outputs at least one statistical value representing the state of the processing device 10. The monitoring device 50 obtains at least one statistical value from the analytical model and determines whether the state of the processing device 10 is abnormal based on the at least one statistical value. If the monitoring device 50 determines that the state of the processing device 10 is abnormal, it specifically designates the item that is the cause of the abnormality from the items in the actual data as a cause item and outputs it. The user performs actions on the processing device 10 based on the output cause item. By doing so, the user can understand the factors affecting the quality of the product and take specific measures against the processing device 10, thereby improving the state of the processing device 10 as early as possible. As a result, the quality of the products processed by the processing device 10 can be improved.

[0091] (Example of monitoring the status of a wire saw device)

[0092] When the processing device 10 is a wire saw, the control unit 52 of the monitoring device 50 monitors the status of the device as described below. The wire saw cuts a cylindrical block to produce a sliced ​​wafer. The sliced ​​wafer undergoes subsequent processes such as grinding to become a product wafer. It is assumed that the actual data of the wire saw, i.e., the processing data, includes the temperature of each part of the device, the position of the wire guides (guide roller 32 and roller 14, etc.), and the flow rates of cooling water and slurry. Furthermore, it is assumed that the processing data of the wire saw includes data on the length, resistance value, oxygen concentration, and crystal orientation of the processed block. Furthermore, it is assumed that the processing data of the wire saw includes data on the warp, waviness, and processing scratches of the cut wafer (sliced ​​wafer). The control unit 52 generates an analytical model based on the above items.

[0093] like Figure 5 As shown, the wire saw device cuts a block, which is a workpiece W, along the cutting direction. The cutting direction is along the diameter of the cylindrical shape of the workpiece W. The wire saw device requires a predetermined time to cut one workpiece W. The period during which the wire saw device cuts one workpiece W can be divided into several intervals. Figure 5 In the example, the period during which the wire saw device cuts a workpiece W is divided into a first interval from the start of the cut to a sixth interval from the end of the cut. The control unit 52 can also aggregate and obtain processing data covering the entire period of cutting a workpiece W. Processing data covering the entire period is also called batch data. The control unit 52 can also obtain processing data divided into intervals. Processing data divided into intervals is also called interval data. The control unit 52 in... Figure 5 The example can obtain the interval data for each of the 6 intervals.

[0094] The control section 52 acquires the batch data as actual data. The control section 52 divides the batch actual data into the interval data of each interval. The control section 52 inputs the interval data to the interval model as actual data to acquire the statistical values of each interval. The control section 52 can determine the state of the wire saw device based on the statistical values of each interval. The control section 52 can also determine that the state of the wire saw device is abnormal if the statistical value of at least one interval is outside the management range.

[0095] The control section 52 acquires the batch data as actual data. The control section 52 divides the batch actual data into the interval data of each interval. The control section 52 inputs the interval data to the interval model as actual data to acquire the statistical values of each interval. The control section 52 can determine the state of the wire saw device based on the statistical values of each interval. The control section 52 can also determine that the state of the wire saw device is abnormal if the statistical value of at least one interval is outside the management range.

[0096] The control section 52 inputs the data of the temperature of each section of the device, the position of the wire guide (the guide roller 32 and the roller 14, etc.), the flow rate of the cooling water and the slurry, and the length, the resistance value, the oxygen concentration, and the crystal orientation of the processed block as actual data to the analysis model. The control section 52 acquires two statistical values, the T2 statistical value and the Q statistical value, from the analysis model. If the acquired statistical values are within the management range, the control section 52 determines that the state of the wire saw device is normal. If the acquired statistical values are outside the management range, the control section 52 extracts the cause item and notifies the user through the display device 60. For example, in a case where the position of the wire guide is extracted as the cause item, the user makes preparations for adjusting the position of the wire guide after the processing of the current batch ends in advance, and adjusts the position of the wire guide before the processing of the next batch starts. The user makes preparations for adjustment in advance, whereby the utilization rate of the wire saw device can be improved.

[0097] <Selection of Interval as an Object of Analysis>

[0098] The control section 52 can determine the state of the wire saw device based on the statistical values acquired by inputting the interval data of the interval before the middle of the period for processing one batch to the analysis model as actual data from among the plurality of interval data included in the period for processing one batch. In this case, the control section 52 can determine the state of the wire saw device based on the statistical values acquired by inputting the interval data of the interval before the middle of the period for processing one batch to the analysis model as actual data from among the plurality of interval data included in the period for processing one batch. Figure 5In the example, the interval from the first interval to the third interval corresponds to the interval before the middle. If the control unit 52 can determine that the wire saw device is in an abnormal state based on the statistical values ​​of the interval data before the middle, it can provide the user with sufficient time to process the abnormality before starting the next batch. The user can ensure that preparation time for handling the wire saw device is performed before starting the next batch. By doing so, the utilization rate of the wire saw device can be improved.

[0099] The control unit 52 can use the initial second interval from the second interval to the fifth interval where the wire saw device stably cuts as the interval for analysis. The deviation of the processing data in the interval where the wire saw device stably cuts can become smaller than the deviation of the processing data in the interval where cutting begins. By using the processing data with small deviation as actual data to calculate statistical values ​​and determine the state of the wire saw device, the determination accuracy can be improved. Furthermore, by using the processing data with small deviation as teaching data to generate an analytical model, the stability of the analytical model can be improved.

[0100] When the processing device 10 is a wire saw, the control unit 52 can input the interval data of the interval that causes the block to cut forward a predetermined distance from among multiple interval data as actual data into the analytical model to obtain at least one statistical value. The control unit 52 can determine whether the state of the processing device 10 is abnormal based on the obtained at least one statistical value. That is, the control unit 52 can correspond the interval that causes the block to cut forward a predetermined distance to the interval used for analysis.

[0101] The intervals used for analysis can be divided based on the characteristics of the wafer. For example, as with Figure 5 The AA section in Figure 6 As shown, the wafer produced by cutting the block with a wire saw has a surface irregularity represented by WS along the cutting direction. Figure 6 In the diagram, the horizontal axis corresponds to the position along the wafer cutting direction. The vertical axis corresponds to the height of the wafer at each position along the wafer cutting direction.

[0102] The shape along the cutting direction of the surface is divided into three sections: the section where the cut begins (W1), the section where the cut ends (W2), and the section in between (W0). Each section approximates a plane. The inclination of the cut surface differs in each section. Specifically, the middle section (W0) is approximated by a double-dotted line (L0). The section where the cut begins (W1) is approximated by a double-dotted line (L1). The section where the cut ends (W2) is approximated by a double-dotted line (L2).

[0103] The height difference of the surface shape in the section (W1) where the cutting starts is calculated. The height difference of the surface shape in the section (W1) where the cutting starts is calculated as the sum of the maximum value of the length of the perpendicular line that drops from the point located above when viewed from the approximation line (L1) among the points of the surface shape of the section (W1) where the cutting starts to the approximation line (L1) and the maximum value of the length of the perpendicular line that drops from the point located below when viewed from the approximation line (L1) to the approximation line (L1). Specifically, in Figure 6 , the points that are farthest from the approximation line (L1) upward and downward among the surface shape of the section (W1) where the cutting starts and the line that is parallel to the approximation line (L1) are represented by a broken line. The height difference of the surface shape in the section (W1) where the cutting starts is calculated as the distance of the broken line upward and downward. The height difference of the surface shape in the section (W2) where the cutting ends is calculated as the sum of the maximum value of the length of the perpendicular line that drops from the point located above when viewed from the approximation line (L2) among the points of the surface shape of the section (W2) where the cutting ends to the approximation line (L2) and the maximum value of the length of the perpendicular line that drops from the point located below when viewed from the approximation line (L2) to the approximation line (L2). Specifically, in Figure 6 , the points that are farthest from the approximation line (L2) upward and downward among the surface shape of the section (W2) where the cutting ends and the line that is parallel to the approximation line (L2) are represented by a broken line. The height difference of the surface shape in the section (W2) where the cutting ends is calculated as the distance of the broken line upward and downward.

[0104] In a case where the control section 52 takes the section where the cutting of the block advances by a prescribed distance as the section for analysis, the prescribed distance corresponds to the length of the section where the value of the item indicating the characteristic of the wafer obtained by cutting the block is calculated. The item indicating the characteristic of the wafer corresponds to the tilt angle of the cutting surface at the time of the start of the cutting or the time of the end of the cutting of the wafer obtained by cutting the block. The prescribed distance corresponds to the length of the section indicated by W1 or W2. The control section 52 can determine the state of the wire saw device based on the statistical value calculated based on the section data of the section where the value of the item indicating the characteristic of the wafer is calculated. By doing so, it is possible to improve the determination accuracy of the state of the wire saw device. As a result, it is possible to improve the quality of the product.

[0105] <Analysis of Processing Data at the Time of Idling>

[0106] It is possible to perform a maintenance work such as a component replacement or cleaning on the processing device 10. In order to confirm the state of the processing device 10 where the maintenance work is performed, the processing device 10 is sometimes operated without the workpiece W. The operation without the workpiece W is also referred to as idling.

[0107] The control section 52 can generate the analysis model using the machining data of the machining device 10 at the time of the idling as the teaching data. The machining data of the machining device 10 at the time of the idling is also referred to as idling data. The analysis model generated using the idling data as the teaching data is also referred to as an idling model. Further, the control section 52 can input the idling data to the idling model as the actual data to obtain the statistical value. The control section 52 can determine the state of the machining device 10 based on the statistical value of the idling data. By doing so, it is easy to avoid an abnormality occurring when the machining device 10 is operated after the maintenance work with the workpiece W loaded. As a result, it is easy to avoid an abnormality in the machining of the workpiece W. By avoiding an abnormality in the machining, it is possible to improve the quality of the product.

[0108] The control section 52 can also generate the idling model using the idling data at the time when no abnormality has occurred over a prescribed period after the idling of the machining device 10 as the teaching data. In other words, the control section 52 can generate the idling model using the idling data other than the idling data in which an abnormality has occurred within the prescribed period after the idling of the machining device 10 as the teaching data. By doing so, it is possible to improve the stability of the idling model.

[0109] <Example of monitoring of state of wire saw device based on Q statistical value>

[0110] The control section 52 can calculate the statistical value for the batch data of each batch of the wire saw device and determine whether the statistical value of each batch is within the management range. For example, as illustrated as a graph in Figure 7 and Figure 8 , the control section 52 can determine whether the Q statistical value of each batch is less than the threshold value of the Q statistical value (QTH). In Figure 7 and Figure 8 , the horizontal axis represents the batch. The vertical axis represents the Q statistical value. Figure 7 The graph of Figure 5 represents the Q statistical value of each batch calculated from the interval data of the first interval in Figure 8 . The graph of Figure 5 represents the Q statistical value of each batch calculated from the interval data of the second interval in .

[0111] The control section 52 determines that both the Q statistical value of the first interval illustrated in Figure 7 and the Q statistical value of the second interval illustrated in Figure 8 are QTH or more with respect to the Q statistical value corresponding to the batch represented by X1. In this case, the control section 52 can determine that the state of the wire saw device when processing the batch represented by X1 is abnormal. Further, the control section 52 determines that the Q statistical value of the first interval illustrated in Figure 7 is less than QTH but the Q statistical value of the second interval illustrated in Figure 8The Q statistic value of the second interval shown is QTH or more. In this case, the control section 52 can also determine that the state of the sawing device at the time of processing the batch indicated by X1 is abnormal.

[0112] The control section 52 extracts the cause item for the batch for which the Q statistic value is QTH or more. For example, the control section 52 extracts the cause item for the batch indicated by X1. As shown in FIG. 6, it is assumed that four items indicated by F1 to F4 are extracted as the cause item. Figure 9 Figure 9 The horizontal axis in FIG. 6 indicates the magnitude of the Q statistic value of each cause item. The control section 52 can output the Q statistic value of each cause item to the display device 60, or can output the cause item for which the Q statistic value is a predetermined value or more to the display device 60. The display device 60 displays the cause item for which the Q statistic value is large and notifies the user. The user can confirm the cause item for which the Q statistic value is large to decide and perform the treatment for the sawing device.

[0113] QTH can be set to, for example, 3 times (+3σ) the average value + the standard deviation of the Q value in the past normal device state. QTH is not limited to this, and can be set to other various values.

[0114] (Other Embodiments)

[0115] Hereinafter, other embodiments will be described.

[0116] <Pass / Fail Determination of Product>

[0117] The control section 52 of the monitoring device 50 can generate an analysis model based on the actual data of the processing device 10 so that the quality of the final product when the product processed and delivered by the processing device 10 is completed to the final process can be estimated. The control section 52 can estimate whether the product processed and delivered by the processing device 10 satisfies the specifications when it becomes the final product based on the estimation result of the quality. That is, the control section 52 can estimate the final pass / fail of the product processed and delivered by the processing device 10 based on the actual data in the processing device 10.

[0118] The monitoring device 50 can determine that the state of the processing device 10 is abnormal when it is estimated based on the actual data that the product delivered by the processing in which the actual data is acquired does not satisfy the specifications when it becomes the final product. In the case where the final product is a wafer, the monitoring device 50 can determine that the state of the processing device 10 is abnormal when the estimation result of the quality of the wafer does not satisfy the quality criteria.

[0119] The control section 52 can estimate the quality of the final product based on, for example, Figure 5 ​The interval data of each interval from the first interval to the sixth interval exemplified above was used to estimate the quality of the final product. With respect to the estimation results of the quality of the final product at the time when the product to be processed by the processing device 10 and delivered is completed to the final process, it was confirmed whether or not the specifications were satisfied when actually completed to the final process to become the final product. The estimation results based on the interval data of each interval were compared with the actual results. As a result, the coincidence rate of the estimation results based on the interval data of each interval from the first interval to the sixth interval and the actual results was 96% for the first interval, 99% for the second interval, 97% for the third interval, 87% for the fourth interval, 97% for the fifth interval, and 81% for the sixth interval. It can be said that the control unit 52 can correctly determine with a precision of 97 to 99% in the case where the state of the processing device 10 is determined based on the estimation of the interval data of each interval from the first interval to the third interval. According to this data, it can be said that the precision is sufficiently high in the case where the state of the sawing device is determined based on the interval data of the interval before the period in which one batch is processed as actual data. Therefore, the advantage that the state of the device can be judged at the initial stage of the processing of the batch to prepare for the next batch can be emphasized.

[0120] <Method of manufacturing a wafer and wafer>

[0121] The processing device 10 monitored by the monitoring device 50 executing the monitoring method according to the present embodiment is used to process a product. The processing device 10 processes a wafer as a product, and thus, a wafer is manufactured. Therefore, a method of manufacturing a wafer including a step of processing a wafer by the processing device 10 monitored by the monitoring device 50 executing the monitoring method is realized. Further, a wafer processed by the processing device 10 monitored by the monitoring device 50 executing the monitoring method is realized.

[0122] <Example of product to be processed as a monitoring target>

[0123] In the present embodiment, it is assumed that the product is a wafer, but it is not limited thereto, and can be various industrial products such as industrial parts or materials, or various other products such as food.

[0124] <Example of device structure>

[0125] In the monitoring system 1, the monitoring device 50 can be included in a part of the processing device 10. The monitoring device 50 can also be configured separately from the processing device 10.

[0126] <Example of processing data in the case where the processing device 10 is a polishing device>

[0127] As described above, the processing apparatus 10 is not limited to the wire saw apparatus. In the case where the processing apparatus 10 is a polishing processing apparatus, an SMP (single-side polishing apparatus), a DSP (double-side polishing apparatus), or the like, the processing data as the actual data can include the temperature of a polishing liquid, the flow rate of the polishing liquid, a polishing pressure, or a moving speed in a horizontal direction, and the like.

[0128] While the embodiments of the present disclosure are described based on the drawings and the examples, it should be noted that various modifications or changes can be made based on the present disclosure if the person skilled in the art. Therefore, it should be noted that these modifications or changes are included in the scope of the present disclosure. For example, the functions and the like included in each structural part or each step and the like can be reconfigured in a logically non-contradictory manner, and a plurality of structural parts or steps and the like can be combined into one or divided. While the embodiments of the present disclosure are described centering on the apparatus, the embodiments of the present disclosure can also be realized as a method including the steps performed by each structural part of the apparatus. The embodiments of the present disclosure can also be realized as a method executed by a processor included in the apparatus, a program, or a storage medium in which the program is recorded. It should be understood that the scope of the present disclosure also includes these.

[0129] The drawings included in the present disclosure are schematic. The scale and the like are not necessarily consistent with the actual object.

[0130] Industrial Applicability

[0131] According to the embodiments of the present disclosure, it is possible to improve the quality of the product.

[0132] Explanation of Reference Numerals

[0133] 1: Monitoring system

[0134] 10: Processing apparatus (12: wire, 14: roller, 16: wire group, 18: workpiece holding mechanism, 20: nozzle, 32: guide roller, 33: tension adjusting arm, 34: tension adjusting roller, 35: contact roller, 36: drive motor, 38A, 38B: wire drum, 40: slurry tank, 42: slurry cooler, W: workpiece (block), X: roller axis direction)

[0135] 50: Monitoring apparatus (52: control part, 54: storage part)

[0136] 60: Display apparatus.

Claims

1. A monitoring method of monitoring a state of a processing apparatus that processes a block or a wafer, based on actual data acquired from the processing apparatus, wherein, The monitoring method includes the steps of: inputting the actual data to an analysis model that defines a latent variable based on a relationship of a value of an explanatory variable included in teaching data and a value of an objective variable, and acquiring a T2 statistic value that indicates a position of the actual data in a subspace defined by the latent variable as at least one statistic value that indicates a state of the processing device, which is output from the analysis model, the analysis model being an analysis model based on a partial least squares (PLS) regression method that is generated using processing data of the processing device as the teaching data in order to indicate a relationship of the actual data and the state of the processing device; determining whether the state of the processing device is abnormal based on the at least one statistic value; and in a case where it is determined that the state of the processing device is abnormal, specifying a cause item that is a cause of the abnormality from among items of the actual data and outputting the cause item.

2. The monitoring method according to claim 1, wherein, The monitoring method further includes a step of determining that the state of the processing device is abnormal in a case where the at least one statistic value is a value outside a management range.

3. The monitoring method according to claim 1, wherein, The monitoring method further includes the steps of: estimating a quality of the wafer based on the at least one statistic value; and determining that the state of the processing device is abnormal in a case where a result of the estimation of the quality of the wafer does not satisfy a quality criterion.

4. The monitoring method according to any one of claims 1 to 3, wherein, The monitoring method further includes the steps of: acquiring batch data in a period in which the processing device processes one batch; dividing the batch data into a plurality of interval data in each interval after dividing a period in which the one batch is processed into a plurality of intervals; and inputting the interval data to the analysis model as the actual data and acquiring at least one statistic value that indicates the state of the processing device in the each interval. The monitoring method further includes a step of generating an interval model corresponding to the each interval as the analysis model using processing data of the each interval as teaching data, and inputting the interval data to the interval model.

5. The monitoring method according to claim 4, wherein, The monitoring method further includes a step of determining whether the state of the processing device is abnormal based on at least one statistic value acquired by inputting interval data in an interval before a middle of the period in which the one batch is processed among the plurality of interval data to the analysis model.

6. The monitoring method according to claim 4, wherein, 7. The monitoring method according to claim 4, wherein the processing device is a wire saw device that cuts a block, the monitoring method further includes a step of determining whether the state of the processing device is abnormal based on at least one statistic value acquired by inputting interval data in an interval in which cutting of the block advances by a prescribed distance among the plurality of interval data to the analysis model. The prescribed distance is a length of a range in which a value of an item that indicates a characteristic of a wafer obtained by cutting the block is calculated.

8. The monitoring method according to claim 7, wherein, The item that indicates the characteristic of the wafer is an inclination angle of a cutting surface at a start of cutting of the wafer obtained by cutting the block.

9. The monitoring method according to claim 8, wherein, The monitoring method further includes the steps of:

10. The monitoring method according to any one of claims 1 to 3, wherein, acquiring actual data in a period in which the processing device is idling as idling data; and inputting the idling data to the analysis model and acquiring at least one statistic value that indicates a state of the processing device at the time of idling. ​ 11. The monitoring method according to claim 10, wherein, Further comprising a step of generating an idling model corresponding to a period of idling of the processing apparatus as the analysis model by using the idling data as teaching data.

12. A computer-readable storage medium storing a monitoring program, wherein, The monitoring program causes a processor to execute the monitoring method of any one of claims 1 to 11.

13. A monitoring device, wherein, The monitoring device includes a control unit that executes the monitoring method of any one of claims 1 to 11.

14. A method of manufacturing a wafer, wherein, Further comprising a step of processing a wafer by a processing apparatus that has performed monitoring by executing the monitoring method of any one of claims 1 to 11.

Images