JP3355849B2 - Manufacturing process quality abnormality treatment system and quality control value updating method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a system for treating abnormalities in the quality of a manufacturing process, such as semiconductor manufacturing, in which quality control is performed by measuring process capability, and to a method for updating the quality control value.

[0002]

2. Description of the Related Art Conventionally, in many manufacturing plants, including semiconductor manufacturing plants, in order to improve product quality and operation efficiency,
Aiming for automation. At that time, management is naturally required for all the manufacturing processes in order to control the quality of the product. Especially in semiconductor manufacturing plants, the number of processes is large, and it takes time to foresee the quality control of individual processes, and it is necessary to consider the relevance of other processes, etc. There is a limit.

A manufacturing factory is constructed so that the quality can be automatically controlled. For example, Japanese Patent Application Laid-Open No. 4-316156 discloses a process management system in a semiconductor manufacturing factory aiming at automation. Discloses technology related to performance data maintenance that monitors the progress of lots in real time, avoids temporary concentration of loads, and protects performance data to facilitate troubleshooting. 1 shows a configuration of a system.

[0004]

However, the disclosed technology merely collects process data, movement information, and the like, and only urges a person to deal with a trouble based on the collected data. In the case of a semiconductor manufacturing plant, especially when manufacturing in-vehicle semiconductors with strict conditions, quality assurance must be strictly performed, so that the data to be collected and the process data to be assured are enormous. Therefore, since oversight can occur sufficiently, that is, since the occurrence of an abnormality is recognized only after it has occurred, the progress of the abnormality at that time is analyzed, and countermeasures corresponding to it are implemented. Will be solved. Since this must be taken into account at every checkpoint in every step, manual control is limited.

[0005] On the other hand, in lot progress management,
The mechanism to automatically stop abnormally the lot where the abnormality occurred has been used conventionally, but in the conventional technology,
It is only necessary to provide information on the trouble countermeasure processing after the occurrence of the abnormality or to check whether the information is within a predetermined management value.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a manufacturing process quality abnormality treatment system capable of taking immediate action when an abnormality occurs in process data and taking measures to prevent occurrence of a trouble in advance, and a quality control value updating method thereof. That is.

[0007]

In order to solve the above-mentioned problems, the present invention comprises a process control device, a unit control device, an alarm device, and a lot automatic stop device under a factory control system in a manufacturing plant. , And a collected data check device, a management value for managing process data, a management upper limit, a management width determined by a management lower limit, the management value is updated based on past results, and the management width is reduced. A control value automatic updating device for automatically updating, and a means for generating an alarm for the flow of the lot when a certain lot deviates from the control range and has a predetermined tendency with respect to the control value, or a flow of the lot. And means for stopping the operation.

According to the present invention, as a method of updating a quality control value, a group of continuous data collected by a process data collection device is initially set by a collection data check device under a factory management system of a manufacturing plant. The management upper limit, the management width determined by the management lower limit, or determine whether or not a specific condition is satisfied with respect to the initially set management value, and when the condition is satisfied, the management value automatic updating device Only the side that tightens the control automatically updates, and when the data of the corresponding lot deviates from the management width, an alarm is generated for the flow of the corresponding lot, and the flow of the lot is stopped by the automatic lot stop device, It is characterized by having a characteristic configuration of:

[0009] In the related invention, in addition to the above construction, the control value automatic updating device calculates an average value and a variation from a continuous group of data with respect to certain measurement data of a certain lot, and obtains the average value. Is replaced with a new management value when it is within the management range, and furthermore, a predetermined multiple of the variation is compared with the management upper limit value and the management lower limit value,
If the difference is within the management range, the predetermined multiple of the variation is replaced with a new management upper limit value and a new management lower limit value.

[0010]

In every step of the manufacturing process, required data is collected by a process data collection device, and each collected process data is managed by a collected data check device based on product specification (spec) values. , And the management upper limit value and the management lower limit value, and it is determined whether or not it is within the management range. At the same time, the average value and the variation are calculated by sampling, the trend of the collected data is checked, and the management width is automatically reset to a stricter side in accordance with the obtained value. Under the management range in which the collected data is set and updated more strictly, if the management range is exceeded or deviates from predetermined conditions, an alarm is issued or the lot is automatically stopped.

[0011]

[Effects of the Invention] Even if the number of processes in a manufacturing factory is so large that quality control is difficult for humans, the automation factory is operated with control values exceeding the product specifications, and the control values are recorded in actual data. In addition, the quality is improved over the entire process because it is stricter. In addition, since a countermeasure can be taken before an abnormality occurs by receiving an alarm in advance, it is possible to prevent an abnormality from being overlooked and prevent a useless lot from being generated.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to specific embodiments. FIG. 1 is a block diagram showing the overall configuration of a semiconductor process quality abnormality treatment system according to the present invention. An automated semiconductor manufacturing plant management system 100 is a higher system of the present invention and is incorporated in the higher system. is there. The automated semiconductor manufacturing plant management system 100 is shown in FIG.
As shown in the figure, process management 2, flow control device group 3, history management 4, alarm management 5, basic unit management (basic data management)
6, a production control 7, a scheduler 8, an equipment management 9, etc., and a basic unit management (basic data management) 6 performs registration and maintenance of basic data related to all of a wide range of automated semiconductor manufacturing factory management systems 100. And a management value registration device 10 and a process registration device 11.

FIG. 2 is a block diagram showing a characteristic portion of the present invention.
2, a collected data storage device 13, a collected data check device 14, and the like. The management value registration device 10 includes a management value automatic change device 15 and a management value registration storage device 16. The process data collection device 12 collects various data from the equipment currently flowing in each equipment. The collected data storage device 13 partially stores the collected process data for the past several tens of processes for each process in cooperation with the history management 4. The format of data to be collected is different for each process data.

The automatic management value changing device 15 of the management value registration device 10 updates the management limit value based on the check result of the collected data check device. However, it is not updated depending on the conditions. Then, the updated management value is held in the management value registration device, and is used for monitoring each lot under the basic unit management 6.

Each lot is started or transported under the control of each device of the flow control device 3 in the automation factory. In this control, when the collected data check device 14 determines that an abnormality exceeding the management value has occurred, the automatic lot stop device 19 is driven, and various alarms are generated to the system administrator, and the lot is stopped. Inform the danger of out, etc., to urge the adjustment, and if there is an emergency, the corresponding lot is automatically stopped.

FIG. 3 is a conceptual diagram showing automatic management value updating. The line graph collection process data 31 may be raw data, a lot average value (lotx), or a MaxMin value depending on the process or according to the specifications of the apparatus. Of those collected data, for example, the past 25
The average value and the variation are calculated from the points (calculation formulas will be described later). For example, under the following conditions, the control value (CL),
The control upper limit (UCL) and the control lower limit (LCL) are determined and automatically updated as a new control limit.

The initial value (CLo) of the control value is manually defined in the basic unit. If the collected process data falls within 25 consecutive points and within the control limit value (control width), the average value x and the variation (standard deviation) σ are calculated from the data of the 25 points.

[Equation 1] UCL = x + 3σ

## EQU2 ## CL = x

LCL = x−3σ

The counting of 25 points is started from the next lot after the control limit value is changed, and is continued as long as the collected data does not exceed the control limit value. If the collected data deviates from the control limit value within 25 points, the lot automatic stop device 19
To reset the count and restart counting from the next data. If either of the calculated upper control limit (UCL) or lower control limit (LCL) is out of the control limit before calculation, the control limit is not automatically updated, and the value before update is not executed. Leave as is, and notify the alarm management 5.

As long as the measured data satisfies the conditions (1) to (4), even if the control value (CL) changes, the control value does not exceed the initial control limit value. The management width narrows as much as possible around CL) (converges at a certain value). As a result, the control limit value is narrowed according to the actual results, it is possible to achieve better control values and maintain quality, and furthermore, it is likely that abnormalities will occur in advance due to the degree of change in the control values. A sign can be predicted, and occurrence of an abnormality can be prevented before it occurs. Note that this management value (CL) does not always coincide with the center value of the product specifications (specifications, specification values). However, most of the time, the product specification is used as the initial value, so the control value (CL) is always within the specification. Conversely, by comparing the target specification center value with the management value (CL) based on the collected data, it is possible to know the characteristics and characteristics of the production line. Naturally, when the value deviates from the standard value (specification value), the alarm and the lot automatic stop are taken similarly or more strictly.

If a lot deviates from the management value of the collected data, a message is sent to the lot automatic stop device 19 to automatically stop the lot and remove it from the line, thereby prohibiting subsequent flow. If this lot is to be reflowed, it will be performed under the direction of the responsible person. Note that the management limit value is only reduced in management width by automatic updating. This is because automatically loosening the control range reduces the quality. Therefore, the management width is not automatically loosened, and if it is desired to be loosened, the management value registration device 10 also forcibly defines the management width as required by the instruction of the person in charge.

For example, if the collected data drifts as indicated by a broken line 31 'in FIG. 3 and data that deviates from the control value comes in (point A), the data may exceed the narrowed control limit value. Then, the above condition is not satisfied, and the lot is automatically stopped. The abnormal data value is a value that sufficiently clears the control limit value in the initial stage of FIG. 3, but temporarily assumes the conventional control limit value as estimated from the movement of the broken line 31 ′. If the check was made at (32 'dashed line), the value would deviate even further and become abnormal (point B). If no action was taken, the specifications would be out of specification and the lot would be wasted. It will be done.

Therefore, as in the present invention, the data is checked in a strict control width sufficiently exceeding the specifications, and a case where the data is deviated from the data is checked, so that it is possible to prevent a large abnormality from occurring. This tight control range cannot be realized from the beginning. This is because the specifications themselves have to consider variations in the production of the product, and often have only a strict specification width with no margin. Therefore, setting a narrower management width than the specification in advance causes lots of lots to be stopped in an automated semiconductor manufacturing plant, and lowers the operation efficiency. However, in an actually manufactured product, it may appear in any range of the management range. Therefore, as in the present invention, by narrowing the management range based on the track record of collected data, the actual effect can be reduced. A certain strict control range can be realized, and loss can be reduced by preventing abnormalities beforehand.

As can be seen from FIG. 3, since the tendency of the collected data clearly indicates the tendency of the drift in the region C of the polygonal line 31, if the collected data is monitored at this time, Although it is possible to take measures, it is not efficient to always perform a visual inspection in the automated state. Therefore, as in the present invention, the management value is automatically updated and the point A in FIG. By allowing the collected data check device 14 that is the management device to recognize the limit point, it is possible to prevent the occurrence of an abnormality beforehand and to cope with it.

A mechanism for grasping the tendency of the collected process data in order to prevent the occurrence of an abnormality in advance will be described.
FIG. 4 is an explanatory diagram for grasping a tendency of collected process data in order to prevent occurrence of abnormality in advance. As described with reference to FIG. 3, it is checked whether data collected in a state where the control value is narrowed does not exceed the control value (condition (1)),
When any of the following conditions (2) to (6) is satisfied, the collected data check device 14 notifies the alarm management 5 to generate an alarm and urge the responsible person or the like to take an action. . The broken line indicated by the number in parentheses in FIG. 4 is an example in the case where the numbers of the following conditions apply.

(1) When collected data is out of the control limit value. (2) When the new control limit calculated from the average 25 points of the collected data deviates from the currently used control limit (UCL, LCL). (3) When the average value of the collected data appears continuously on one of the upper and lower sides of the control value (CL) for seven consecutive points. (4) The average value of the collected data is on one side of the control value (CL). (A) At least 10 points out of 11 consecutive points. (B) At least 12 points out of 14 consecutive points. (C) 17 consecutive points. At least 14 points out of 20 points (d) When at least 16 points out of 20 consecutive points appear (they do not have to appear on one side continuously). (5) When the average value of collected data shows either rise or fall for seven consecutive points. (6) When the average value of the collected data is greater than CL ± 2σ and less than or equal to CL ± 3σ, (e) at least 2 points out of 3 consecutive points, (f) at least 3 points out of 7 consecutive points ( g) When at least 4 points out of 10 consecutive points are satisfied.

The management value serving as a criterion is a value set by the automatic update described above. Since the management value is set on the basis of the actual results, human arbitrariness does not enter. Thus, it is possible to grasp the tendency centered on the control value, such as the average value of the flowing lots, the tendency of the drift thereof, and the degree of the control limit value. It should be noted that the conditions shown here are merely examples, so that even if the conditions for issuing an alarm are variously modified in accordance with the product specifications, only the time at which the alarm is issued is changed, and the effect is the same.

The formula for obtaining the average value X _M and the variation σ from the collected data is obtained by using the arithmetic mean formula and the standard deviation formula from the data from which the average value and the variation can be obtained depending on the type of the collected data. calculate. The obtained collected data includes, for example, (a) raw data, (b) average value of one lot (lotx), (c) maximum value / minimum value in each lot,
(d) Lot average value / lot maximum / minimum value; (e) Lot average value and variation between lots. Note that the average value X _M and the variation (standard deviation) σ are expressions for obtaining the standard deviation used in ordinary statistical processing.
It is as shown in the equation. The calculation of the variation σ is based on the average value X
_{Find M} and calculate using that value.

[Equation 4] X _ij : Data value P _j : Number of data

[Equation 5] X _M : average value

The data value of each collected data is applied to the data value X _ij to calculate an average value and variation. The process data collection device 12 naturally identifies the currently obtained data, and obtains the average value and the variation using mathematical expressions.

These data processes are performed by a computer belonging to the process data collection device 12. The type of collected data is not limited to the above, but is used for collecting necessary data, obtaining an average value or variation by using a formula suitable for the attribute, and automatically updating the management value.

By using the above-described method of updating the control values and the control limit values, it is possible to realize quality control for further improving the quality, and it is possible to use an alarm in advance before an abnormality occurs. The production rate of the manufacturing plant is improved without making lots.

[Brief description of the drawings]

FIG. 1 is an overall configuration block diagram of a semiconductor process quality abnormality treatment system.

FIG. 2 is a block diagram showing a characteristic portion of the present invention in FIG.

FIG. 3 is a conceptual diagram showing management value automatic updating.

FIG. 4 is an explanatory diagram for grasping a tendency of collected process data.

[Explanation of symbols]

 REFERENCE SIGNS LIST 100 semiconductor manufacturing plant management system 2 process management device 3 flow control device 4 history management 5 alarm management 6 basic unit management (basic data management) 10 management value registration device 11 process registration device 14 collected data check device 15 management value automatic update device 19 Automatic lot stop device 31 Collected process data 32 Control upper limit (UCL) 33 Control value (CL) 34 Control lower limit (LCL)

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G05B 19/418 B23Q 41/08 H01L 21/02

Claims (3)

(57) [Claims] In a manufacturing plant, a management value for managing process data, which includes a process management device, a basic unit management device, an alarm device, a lot automatic stop device, and a collected data check device under a factory management system; And management limit,
A control value automatic updating device that updates the control value based on the past performance and automatically updates the control value by narrowing the control value based on the past lower limit value, and a control when a certain lot deviates from the control value. A manufacturing process quality abnormality treatment system, comprising: means for generating an alarm for the flow of a corresponding lot when the value has a predetermined tendency, or means for stopping the flow of the lot. 2. Under a factory management system of a manufacturing factory, a group of continuous data collected by a process data collection device is determined by a management upper limit value and a management lower limit value initially set by a collection data check device. It is determined whether or not a specific condition is satisfied with respect to the management range or the initially set management value.If the condition is satisfied, the management value is automatically updated only by the stricter management by the management value automatic updating device, A quality control value updating method, wherein when the data of a relevant lot deviates from the control range, an alarm is generated for the flow of the relevant lot and the flow of the lot is stopped by a lot automatic stop device. 3. The control value automatic updating device calculates an average value and a variation from a continuous group of data with respect to certain measurement data of a certain lot, and when the average value is within the control range, a new control value is updated. Further, the predetermined multiple of the variation is compared with the management upper limit value and the management lower limit value, and if within the management range, the predetermined multiple value of the variation is replaced with a new management upper limit value and a new management lower limit value. 3. The quality control value updating method according to claim 2, wherein: