JP2011107947A - Abnormality detection method and abnormality detection system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method: for extracting the featured values of operation data indicating the operation of a processor or process data indicating processing situations at a user point; for comparing the features values of the operation data with a set value; for comparing the featured values of the process data which can not be directly compared with the set value with a reference value; and for determining the normality/abnormality of a processing without being affected by processing conditions. <P>SOLUTION: Data related with the operation of the device are collected as operation data, and the featured values of the operation data are calculated from the operation data, and compared with a set value, so that the presence/absence of the abnormality of the operation is determined, and an output when control conditions are matched with the featured values of the operation data is predetermined as a reference value. Meanwhile, data related with the output under the operation of the device are collected as process data, and a numerical value corresponding to the reference value is calculated as the featured values of the process data, and the featured values of the process data are compared with the reference value, so that the presence/absence of the abnormality of the output is determined. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention monitors the operation of a manufacturing processing apparatus such as a semiconductor or a liquid crystal panel, collects the operation data or manufacturing status data of the manufacturing processing apparatus, analyzes the operation or manufacturing status of the manufacturing apparatus from the collected data, And an abnormality detection system.

  As a background art, a case of a color filter used for a liquid crystal panel will be described as an example. FIG. 1 is a cross-sectional view showing an example of a color filter used in a color liquid crystal display device. The color filter 1 includes a black matrix (hereinafter referred to as BM) 3, a red R colored pixel (hereinafter referred to as R pixel) 4a, a green G colored pixel (hereinafter referred to as G pixel) 4b, and a blue B colored pixel on a glass substrate 2. (Hereinafter referred to as a B pixel) 4c, a transparent electrode 5, a photo spacer (hereinafter referred to as PS) 6, and a vertical alignment (hereinafter referred to as VA) 7 are sequentially formed.

  As a method for manufacturing a color filter having the above structure, a photolithography method, a printing method, and an ink jet method are known. FIG. 2 is a flowchart showing steps of a commonly used photolithography method. First, the color filter includes a step of forming a black matrix (BM) on the glass substrate (C1), a step of cleaning the glass substrate (C2), a step of applying and pre-drying a colored photoresist (C3), A pre-baking step (C4) for drying and curing the colored photoresist, a step (C5) for exposing, a step (C6) for developing, a step (C7) for curing the colored photoresist, and a film forming process for the transparent electrode A process (C8) and a process (C9) for forming PS and VA are performed in this order and manufactured.

  For example, when the pixels are formed in the order of R pixel, G pixel, and B pixel, red is performed between the process of cleaning the color filter glass substrate (C2) and the process of curing the colored photoresist (C7). The color resist is changed in the order of R, green G, and blue B, and the process is repeated three times to form R pixels, G pixels, and B pixels.

  Examples of defects that occur in the color filter manufacturing process include foreign matter adhesion in the cleaning process, colored resist application unevenness and pinholes in the colored resist application and preliminary drying processes, and excessive and insufficient exposure and mask alignment in the exposure process. There are defects such as over and under development and uneven development in the development process.

  In order to suppress the occurrence of the defect and to detect a defective product, the operation status and manufacturing status of the manufacturing processing apparatus are monitored, and a defective product is detected by an inspection device to guarantee the quality.

  In manufacturing processing apparatuses (hereinafter referred to as processing apparatuses) in manufacturing semiconductors and liquid crystal panels, processing conditions (hereinafter referred to as operating data) of processing apparatuses that indicate the operating status of processing apparatuses, manufacturing conditions such as flow rate, temperature, and pressure of cleaning liquid, In other words, there is an EES (Equipment Engineering System) that collects and analyzes process condition data (hereinafter referred to as process data) at a high rate of several ms. EES collects device operation data and process data from the processing equipment, checks whether the manufacturing process is performed under normal operation and normal process conditions, and improves the reliability and productivity of the processing equipment. It is a system to let you.

  As a process abnormality analysis method using EES, for example, there are the following methods.

Abnormality is detected from the difference from the process data during normal operation, the presence / absence of abnormality is determined, and the cause of abnormality is determined (Patent Document 1).

  A feature quantity (average, maximum, minimum, etc.) is extracted from process data cut out in a certain section, a process-quality model is created, the presence / absence of an abnormality is determined, and the occurrence of a failure is predicted (Patent Document 2).

  A process-quality model is created from analysis by data mining or multivariate analysis using feature quantities extracted from process data, and the presence or absence of abnormality is determined (Patent Document 3).

JP-A 62-42204 JP 2005-197323 A JP 2004-186445 A

  In the process abnormality analysis method using EES, in the method disclosed in Reference 1, it is difficult to detect an abnormality caused by a plurality of factors intertwined because abnormality determination is performed for each process data. Further, when the processing apparatus is operated, the reference value fluctuates due to the change of the process condition. Therefore, the reference value must be changed every time the process condition is changed.

  In the method shown in Reference 2, when a quality model is created from the extracted feature quantity, the setting value of the apparatus is compared with the feature quantity. However, depending on the process data, it is difficult to directly compare with the setting value. Is also present. Furthermore, it is possible to determine whether there is an abnormality, but it is impossible to predict what abnormality is occurring in which part of the apparatus.

  Further, the method described in Patent Document 3 can determine whether or not there is an abnormality, and can also predict what abnormality has occurred because inspection data is used. However, in general, process data varies even when a non-defective product is made, and thus the obtained process data is not necessarily accurate.

  Therefore, the present invention extracts feature data of operation data indicating operation of related processing devices (for example, operation of a pump) and process data indicating a processing state such as pressure at a use point, and sets and sets feature values of operation data. Compares the values, and can determine whether the process data features that cannot be directly compared with the set values are normal / abnormal, and provides a method for determining normal / abnormal process processing regardless of the process conditions The task is to do.

The invention according to claim 1 of the present invention is a method for detecting both an abnormality in the operation of the apparatus and an abnormality in the output based on the operation of the apparatus in the step of performing processing using the processing apparatus,
Control conditions suitable for this processing are set in advance as set values, while data relating to the operation of the apparatus, data relating to the operation of the apparatus are collected as operation data, and numerical values corresponding to the set values are obtained from the operation data. It is calculated as a feature amount of the motion data, and the feature amount of the motion data is compared with the set value to determine the presence or absence of motion abnormality,
Assuming that the device operates normally when the control condition matches the feature amount of the operation data, the output in this assumption is determined in advance as a reference value,
On the other hand, data relating to the output during operation of the apparatus is collected as process data, and a numerical value corresponding to the reference value is calculated from the process data as a feature amount of the process data,
In this abnormality detection method, the feature amount of the process data is compared with the reference value to determine whether there is an abnormality in output.

  The invention according to claim 2 of the present invention is characterized in that the operation data and the process data are each continuous analog data, and the feature amount calculated from each data is extracted and calculated by dividing the section of the analog data. The abnormality detection method according to claim 1.

  The invention according to claim 3 of the present invention uses the abnormality detection method according to claim 1 or 2, and further includes a cause identification data file comprising operation data and process data at the time of occurrence of abnormality and cause data of occurrence of abnormality. The abnormality detection system is characterized in that the cause of the abnormality is specified when the abnormality is detected.

  The invention according to claim 4 of the present invention is the abnormality detection system according to claim 3, wherein the cause identification data file has maintenance information for removing the cause of the occurrence of the abnormality.

  It is possible to detect the malfunction of the processing device by comparing the feature value of the operation data with the set value, and by comparing the feature value of the process data with the reference value obtained based on the operation of the processing device It is possible to determine normality / abnormality. Furthermore, since the cause of the abnormality can be predicted by accumulating the operation data, the process data, and the cause data of the abnormality occurrence, it is possible to suppress the occurrence of defective products.

  When an abnormality is found, it is possible to identify the location where the abnormality has occurred, so that it is possible to quickly repair the processing device and to deal with the cause of the abnormality such as maintenance, thereby preventing a reduction in production efficiency.

The figure which showed an example of the color filter board | substrate in the cross section. The flowchart of the process of the photolithographic method generally used. The figure which shows the outline of a structure of the abnormality detection system to which the abnormality detection method which concerns on this invention is applied. The figure which shows an example of the waveform data in the abnormality detection method which concerns on this invention. The figure which shows an example of the extraction method of the area of the waveform data in the abnormality detection method which concerns on this invention The figure for demonstrating the example of calculation of the feature-value of the analog data in the output side in the abnormality detection method which concerns on this invention. The figure which shows the flow of the determination in the abnormality detection method which concerns on this invention.

  Hereinafter, an embodiment of an abnormality detection method and an abnormality detection system according to the present invention will be described with reference to the drawings, taking as an example the case of manufacturing a glass substrate for a color filter.

  FIG. 3 shows an outline of the configuration of an abnormality detection system to which the abnormality detection method according to the present invention is applied.

The abnormality detection system includes a personal computer (hereinafter referred to as PC) 14 that performs data collection and data processing, ID information 15 acquired from an LCS (Line Control System) 13, operation data and process data 16 collected from the processing device 11 by the ESS 12. And the waveform data 17 created in units of processes from the operation data and the process data 16, the waveform data 19 of the operation data for calculating the feature amount of the operation data from the waveform data 17, and the feature amount of the process data. Waveform data 18 of process data. Further, feature data information 21 of motion data calculated from the waveform data 19 of the apparatus for calculating the feature data of the generated motion data, and waveform data 18 of process data for calculating the feature data of the process data. The feature amount information 20 of the process data for which the feature amount has been calculated from the above is included, and the cause specifying data file for storing the cause specifying data is also included (not shown). The PC 14 has a function of displaying the collected data as waveform data on the terminal of the PC 14 in real time.

  The operation data and process data 16 collected by the EES 12 are analog data for pump operation, pump pressure, stage vibration, cleaning liquid flow rate, etc., and the operation signal of the apparatus is digital data. It is displayed on the data collection PC terminal 14 in real time as waveform data, and it can be used as means for determining whether or not the apparatus is operating normally by determining whether or not the collected waveform is normal.

  FIG. 4 is a diagram showing an example of waveform data, showing an example in which a colored resist is applied to a glass substrate. Data on the operation time of the resist supply pump and the discharge amount of the colored resist is collected, and the collected data Waveform data indicating the elapsed time indicated by ms (milliseconds) on the horizontal axis and the discharge amount indicated by ml (milliliter) on the vertical axis.

  The setting value, feature amount, and reference value will be described.

  The operation data collected by the EES 12 and the analog data or digital data of the process data 16 include the substrate ID that has undergone the process processing while the operation data and the process data are being collected, and the processing apparatus that performs the process processing. Control conditions (hereinafter referred to as set values) are collected in association with each other. For example, in the case of the above-described resist supply pump, the set value here refers to the average value, maximum value, minimum value of the waveform data, the slope at the rise of the waveform data, the time until the waveform data reaches a steady state, and the like. It is a value and is set for each processing device.

  The board ID and the set value associated with the operation data and the process data 16 can be acquired from the LCS 13 which is a host system called a line management system (the line management system refers to the production status of the equipment in the factory, production results, etc. In the system that collects and manages this data in real time, it is possible to acquire the substrate ID and the set value during the process processing by including the substrate ID and the set value of the apparatus in the collected data. ).

  One method of determining whether the waveform data is normal is a method of comparing analog data acquired with a predetermined set value of control conditions suitable for this processing. Analog data displays continuous data during process processing, whereas setting values indicate local or average values, so the analog data feature values (average value, maximum value, minimum value, rise of waveform data) Extract and compare the slope of time).

  The feature value is extracted from the value of the analog data being processed, and the extracted value is compared with the set value to determine whether the extracted value is within the normal range. Can be determined.

  As a result of the normal / abnormal determination, if it is determined as abnormal, the analog data (waveform data) during the process processing in which the abnormality was found on the data collection PC terminal, or the extracted section or calculation in the waveform data described later The feature value is displayed and an alarm is generated.

When comparing the setting value of the processing device with the feature value of analog data, comparison is possible for analog data that is directly controlled, but setting value for analog data that is not directly controlled Comparison with is difficult. That is, for example, when the operation of the pump is controlled, the feature value of the analog data indicating the pump operation can be compared with the set value, but the use point (for example, the discharge pressure at the discharge nozzle of the colored resist) Since it does not control the operation itself, the setting value cannot be set, so it cannot be compared with the setting value.

  Furthermore, when an abnormality occurs in the analog data of the use point, it becomes difficult to isolate the location of the abnormality occurrence (identification of the location causing the abnormality) using only the analog data of the use point (the use in the above example) When an abnormality is detected in the analog data of the point, it is impossible to determine whether an abnormality has occurred in the pump itself or until the use point (for example, in the colored resist piping).

  Further, the reference value of the process data corresponding to the set value of the operation data with respect to the analog data of the use point (for example, the process data corresponding to the pressure), that is, the time until the waveform data of the pump becomes a steady state, for example, By setting the time until the steady state of the pressure corresponding to the set value as a reference value, it becomes possible to compare with the feature amount. In that case, change of the process condition (for example, by changing the color resist) Since the output value with respect to the set value may change due to a change in process conditions), a change with time, etc., it is necessary to set a determination condition together with the change of the setting condition. Therefore, instead of creating a reference value based on the analog data of the process data and setting the determination condition every time the setting condition is changed. Since a reference value based on analog data is used, it is possible to change process conditions and set determination conditions following changes with time.

  A feature amount calculation method will be described.

  First, the feature amount needs to be calculated. In the EES 12, in the case of a colored resist coating apparatus, analog data is divided and collected for each process operation such as before coating, during coating, and after coating. However, the change in analog data for each process operation is large, making it difficult to calculate the average value and extract the peak of a specific part. Waveform data needs to be extracted. That is, a numerical value corresponding to the set value is calculated as a feature amount of the motion data by dividing the section of the waveform data indicating the motion data.

  FIG. 5 is a diagram illustrating an example of a section extraction method. FIG. 5A shows waveform data showing the relationship between the time lapse of the pump for supplying the colored resist and the internal pressure. The OUT valve is opened at the time indicated by the arrow 31, and the arrow 32 is 4 seconds after the OUT valve is opened. Indicates the time point. The feature amount is calculated by dividing a section finely in the process operation when such an event occurs. FIG. 5B is an enlarged view of the waveform data 33 showing the timing from when the OUT valve opens at the time indicated by the arrow 31 in FIG. 5A to the time 4 seconds after the OUT valve indicated by the arrow 32 opens. It is. In this case, the feature amount to be extracted is calculated from the time at the rising timing 34, the timing 35 at which the peak appears, and the time 36 at the steady state timing 36, the value of the internal pressure, and the average value. As described above, the timing at which the analog data is divided and collected is the ON / OFF timing of the digital data (in the case of FIG. 5, the digital data indicates the timing at which the OUT valve is opened) and the analog data has an arbitrary threshold value. The timing at the moment of exceeding, the timing when a certain time has elapsed from these timings, and the like are used. By extracting the section and calculating the feature amount for each processing device, it is possible to determine whether the operation for each processing device is normal or abnormal.

The feature values calculated after section extraction include the average, maximum, and minimum values of waveform data corresponding to the set values that have already been set, the slope of waveform data, the rise timing of waveform data, and the waveform data in a steady state. It is possible to consider the time until the peak and the appearance timing of the peak of the waveform data.

  When obtaining feature values for both operation data and process data and determining whether the operation data is normal / abnormal, compare the device settings obtained from the host system with the feature values of the analog data in the extracted section. Perform normal / abnormal judgment. When determining whether the process data is normal or abnormal, the determination is performed by comparing the feature value of the analog data of the process data in the extracted section with the reference value of the process data.

  An example of calculating the feature amount of the analog data of the process data will be described with reference to FIG. Of the two waveform data shown in FIG. 6, reference numeral 41 is waveform data showing the relationship between the passage of time of the pump and the internal pressure (left scale), and reference numeral 42 is the discharge pressure (right scale) at the discharge nozzle of the colored resist in the same time course. Is waveform data. 35 is the timing when the OUT valve of the pump opens (same as 35 in FIG. 5), 36 is the timing when the internal pressure of the pump peaks (same as 36 in FIG. 5), 37 is the timing when the discharge pressure at the discharge nozzle rises , 38 indicate the timing when the discharge pressure at the discharge nozzle peaks.

  As with the calculation of the feature value of the operation data, the process data feature value is calculated as the process data feature value by dividing the section finely into the waveform data in the event of an event. To do. From the discharge pressure waveform data 42 at the colored resist discharge nozzle shown in FIG. 6 to the timing 37 at which the discharge pressure at the discharge nozzle rises from the timing 35 at which the OUT valve of the pump opens to the timing 37 at which the discharge pressure rises at the discharge nozzle, and the timing at which the discharge pressure rises. A time lapse 44 from 37 to a timing 38 at which the discharge pressure reaches a peak, a discharge pressure 45 at a timing at which the internal pressure of the pump reaches a peak, a timing 38 at which the discharge pressure reaches a peak, or a point at which the discharge pressure reaches a peak. The discharge pressure value 46 is calculated and used as the feature amount of the process data.

  A method for setting a reference value of process data will be described.

  Assuming that the processing device operates normally when the feature value of the operation data matches the set value, the output in this assumption is determined in advance as a reference value of the process data. That is, when the analog data indicating the operation of the processing device is within the set value, the feature amount is obtained from the output analog data based on the operation of the processing device, and the reference value is set for each feature amount. The reference value is preferably set at the beginning of changing process conditions such as changing the color resist. As the reference value, for example, for the process data corresponding to the discharge pressure, a reference value is set for each feature amount as shown below. The reference value setting method is that the time lapse 43 from the timing 35 when the OUT valve of the input side pump opens to the timing 37 when the discharge pressure at the discharge nozzle rises is within 150% of the rise time of the pump or the timing 37 when the discharge pressure rises. The time lapse 44 from when the discharge pressure reaches a peak to timing 38 is within 110% of the time lapse from the timing when the pump starts up to the timing when the internal pressure of the pump reaches a peak, or at the timing when the internal pressure of the pump reaches a peak. By setting the discharge pressure 45 to 70% or more of the peak pressure of the pump internal pressure, the reference value of the process data can be set from the operation data of the processing apparatus. Thus, by setting the reference value of the process data based on the waveform data indicating the operation of the processing device, the process data feature amount can be obtained from the pump waveform data even when the process conditions accompanying the change of the colored resist are changed. It is possible to set a reference value to be compared with the process data, and even if the value of the operation data changes due to the aging of the processing apparatus, the process data can be determined based on the same reference. This reference value can be set without manual intervention every time the process condition is changed.

  A normal / abnormal determination method will be described.

  FIG. 7 shows a determination flow. It is assumed that the set value to be compared with the feature value of the operation data can be acquired from the LCS, and the reference value to be compared with the feature value of the process data is already set by the above method. After the start (S1), first, a section of waveform data from which a feature amount is extracted is extracted (S2). After calculating the feature amount from the motion data (S3), if the feature amount of all motion data is calculated (YES in S4), the feature amount of the motion data is compared with the set value (S5). ) If within the normal range (YES in S6), then the feature amount of the process data is calculated (S7). If all the feature amounts have been calculated (YES in S8), the calculated process data The feature quantity is compared with the reference value, and if it is within a normal range with respect to the reference value (YES in S9), the process is immediately terminated.

  In step (S4), when the feature values of all motion data have not been calculated (NO in S4), the process returns to step (S3) to calculate the feature values.

  In step (S6), the feature value of motion data is compared with the set value, and if it is not within the normal range (NO in S6), an alarm is generated (S10), and the waveform data, section, and feature value of occurrence of abnormality are displayed. (S11), and further, it is determined whether or not there is specific information of the abnormal part in the section where the abnormality has occurred and the characteristic amount (S12). If there is specific information (YES in S12), the abnormal part The maintenance method is displayed (S13), and the process is immediately terminated (S14).

  In step (S9), if the process data feature value is compared with the reference value and it is not within the normal range (NO in S9), an alarm is generated (S10) as in the case of the operation data, and the waveform data indicating the occurrence of abnormality, The section and the feature amount are displayed (S11), and further, it is determined whether or not there is specific information on the abnormal portion in the section and the feature amount (S12). (YES), an abnormal part and a maintenance method are displayed (S13), and it complete | finishes immediately (S14).

  If there is no specific information in step (S12) (NO in S12), the process is immediately terminated (S14). In this case, the abnormality occurrence part is manually searched to correct the abnormality, and further maintenance information is stored in the system to obtain new abnormality part specifying information. The determination flow is performed for each waveform data.

  Describes how to deal with errors. When an abnormality occurs, an alarm is generated, the input side or output side abnormality is displayed on the PC terminal, and a related waveform is further displayed. Also, the cause of the abnormality is registered for each extracted section and feature quantity.

  Although the embodiment for carrying out the present invention has been shown as an example of manufacturing a glass substrate of a color filter, the present invention is not limited to this, and when detecting abnormalities in processing devices and manufacturing conditions in a wide range of manufacturing processes, Applicable.

  As described above, according to the abnormality detection method and the abnormality detection system according to the present invention, it is possible to determine the operation abnormality of the processing apparatus and the abnormality of the manufacturing situation in the manufacturing process, and the defect is caused by accumulating the cause data of the occurrence of the defect. Therefore, it is possible to identify the cause of the occurrence of defective products, and thus it is possible to suppress the occurrence of defective products.

  When an abnormality is found, it is possible to identify the location where the abnormality has occurred, and it is possible to promptly respond to the cause of the abnormality such as repair and restoration of the processing apparatus or maintenance, thereby preventing a reduction in production efficiency.

DESCRIPTION OF SYMBOLS 1 ... Color filter 2 ... Glass substrate 3 ... Black matrix 4a ... Red R colored pixel 4b ... Green G colored pixel 4c ... Blue B colored pixel 5 ... Transparent Electrode 6 ... Photo spacer 7 ... Vertical alignment 11 ... Processing device 12 ... ESS (Equipment Engineering System)
13 ... LCS (Line Control System)
14 ... Personal computer (PC)
15... ID information 16... Operation data and process data 17... Waveform data 18... Process data waveform data 19... Operation data waveform data 20. ... Process data feature information 31 ... Arrow 32 indicating when the OUT valve is opened ... Arrow 33 indicating the time 4 seconds after the OUT valve is opened ... Expanding waveform data 34 Timing of rising 35: Timing of appearance of peak 36 Timing of reaching steady state 37 Timing of rising discharge pressure at discharge nozzle 38 Timing of discharge pressure at discharge nozzle 41 ..Waveform data 42 showing the relationship between the passage of time and the internal pressure of the pump 42... Showing the discharge pressure at the discharge nozzle of the colored resist The discharge pressure at the discharge nozzle from the timing OUT valve shape data 43 ... pump is opened
Elapsed time 44 until the rising timing From the timing when the discharge pressure rises to the timing when the discharge pressure peaks
Time lapse 45 ... discharge pressure 46 at the timing when the internal pressure of the pump reaches a peak ... discharge pressure value at the time when the discharge pressure reaches a peak

Claims (4)

In the process of performing processing using a processing device, a method for detecting both abnormal operation of the device and abnormal output based on the operation of the device,
Control conditions suitable for this processing are set in advance as set values, while data relating to the operation of the apparatus, data relating to the operation of the apparatus are collected as operation data, and numerical values corresponding to the set values are obtained from the operation data. It is calculated as a feature amount of the motion data, and the feature amount of the motion data is compared with the set value to determine the presence or absence of motion abnormality,
Assuming that the device operates normally when the control condition matches the feature amount of the operation data, the output in this assumption is determined in advance as a reference value,
On the other hand, data relating to the output during operation of the apparatus is collected as process data, and a numerical value corresponding to the reference value is calculated from the process data as a feature amount of the process data,
An abnormality detection method characterized in that the feature quantity of the process data is compared with the reference value to determine the presence or absence of an output abnormality.   2. The abnormality detection method according to claim 1, wherein the operation data and the process data are continuous analog data, and the feature amount calculated from each data is extracted and calculated by dividing a section of the analog data. .   The abnormality detection method according to claim 1 or 2, further comprising a cause identification data file including operation data and process data at the time of occurrence of an abnormality and cause data of the occurrence of the abnormality, and at the time of detecting an abnormality, the cause of the occurrence of the abnormality is specified. An abnormality detection system characterized by that.   4. The abnormality detection system according to claim 3, wherein the cause identification data file has maintenance information for removing the cause of occurrence of the abnormality.

Images