JPH0716864B2 - Diagnostic equipment for manufacturing equipment - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention performs a failure diagnosis on a plurality of manufacturing equipments constituting a manufacturing line, and when a failure is diagnosed, measures against the failure are instructed. The present invention relates to a diagnostic device for manufacturing equipment.

[Conventional technology]

As a method for diagnosing manufacturing equipment, up to now, the quantity that is considered to be related to abnormality or failure of the target equipment is measured and processed to obtain the quantity that serves as an indicator of abnormality, failure, or deterioration. There is a method to judge the current situation based on the above, which is widely used as a self-diagnosis method for equipment (Measurement and Control Vol.24, No.4 (April 1985)) by Kageo Akizuki, "Equipment Diagnosis Technology and Safety "). However, according to this method, it is necessary to prepare a measuring device and a measurement amount processing device for each location of individual manufacturing equipment, and the cost required for that is not less than that of the equipment itself. Has become. In addition, it does not determine or instruct preventive measures for avoiding troubles and troubles in product processing. Therefore, in a production line that continuously manufactures using multiple equipment,
Since abnormalities and failures of individual equipment cause the entire line to be inoperable, it is important to take preventive measures to handle product abnormalities and reduce downtime.

[Problems to be solved by the invention]

In a production line composed of a plurality of production equipments, there are many cases where the equipments need to be temporarily stopped and repaired or adjusted due to a failure of the equipments or a change in operating state over time. For example, in the case of using several tens of types of equipment repeatedly to complete processing for a product, such as a semiconductor manufacturing line, a few downtime due to failure or adjustment of individual equipment. Not only does the time significantly reduce the productivity of the entire production line, but it is almost impossible to regenerate the product processed by the faulty equipment, and it is the largest factor of the frequent occurrence of defects.

Therefore, in the production line, preventive measures are taken to maintain the equipment before it breaks down.However, for example, in the case of semiconductor manufacturing equipment that requires delicate adjustments in an extremely clean atmosphere and intricate element devices. In this case, it takes several hours to dozens of hours for the maintenance itself.
The actual situation is that maintenance is performed only by checking, dismantling and re-adjusting after being temporarily stopped, only in the case of a complete equipment failure.

By the way, the following information can be cited as information necessary for appropriately performing preventive maintenance for the purpose of preventing equipment failure and abnormal processing.

(1) Information on past failures and countermeasure examples to estimate the cause and location of abnormalities (2) Detailed equipment configuration, causal relationships regarding abnormal states and their causes to plan maintenance methods and procedures, and Information on the frequency of occurrence Since an experienced engineer memorizes the above information as expert knowledge from past experience, it is based on the stored causal relationship based on information such as abnormal conditions and product processing accuracy. Although it is possible to determine the cause of the abnormality, its location, countermeasures, etc., the fact is that the knowledge is not fully utilized for preventive maintenance.

Therefore, an object of the present invention is to store the knowledge of a skilled engineer in a storage device, and use this as a clue for information that can be easily recognized by an unskilled person, such as an abnormal state of maintenance and product processing accuracy. Thus, a diagnostic device for manufacturing equipment is provided, which is capable of diagnosing equipment and instructing its countermeasures.

[Means for solving problems]

In the present invention, a plurality of manufacturing equipment for manufacturing the product,
A diagnostic device for manufacturing equipment in a manufacturing line, which is provided for each manufacturing equipment and has an inspection device that inspects the processing contents of each product processed by each manufacturing equipment in time series, A series of causal maintenance history data consisting of past failure points, failure states, causes, and countermeasures, quality control data such as processing dimensions and accuracy of products, and equipment operation such as processing condition setting values and processing quantity of manufacturing equipment. Storage means in which data is accumulated in advance, processing means for extracting relevant failure points based on inspection data of time-series transitions related to processing states of products, and quality control data, equipment operation data provided for each manufacturing equipment Based on the above, various preset parameters are selected, and a record of the state of failure, the cause of failure, and the machining dimensions and accuracy of the product is input. On the other hand, data transfer for exchanging data between the input / output means capable of displaying the failure portion extracted by the processing device as a countermeasure required portion, the processing means, the storage means, the input / output means, the manufacturing equipment and the inspection means. And means. Then, the processing means, when a desired parameter regarding the processing state of the product is selected via the input / output means, inspection data of time-series transition of each product obtained by the inspection means corresponding to each manufacturing facility. And a means for plotting the obtained value in time series, and a control limit value corresponding to the plotted time series data and the above selected parameter and predetermined by the type of process or product. Means for patterning the time-series data plotted by comparing with, and searching for the patterned time-series data from various past time-series transition patterns, and the patterning time-series data , When it substantially matches a specific time-series transition pattern in the past, the causal relationship maintenance history data based on the matched time-series inspection data is displayed. Means for retrieving from the record unit, and having a means for displaying as a large fault location in order of occurrence frequency associated locations via the input means based on maintenance history data the search.

[Work]

According to the present invention, the above-mentioned storage means, processing means, input / output means, and data transfer means are provided, and when the processing means selects a desired parameter relating to the processing state of manufacture via the input / output means, each manufacturing facility The inspection data of the time series transition of each product obtained by the corresponding inspection means is obtained by the cumulative sum method, and the obtained value is plotted in time series, and the plotted time series data Means for patterning the time-series data plotted by comparing with the control limit value that corresponds to the above selected parameters and is predetermined by the type of process or product, and the patterned time-series data in the past We searched from various time-series transition patterns, and the patterned time-series data almost matched the past specific time-series transition patterns. Then, a means for retrieving causal relationship maintenance history data from the storage means based on the matched time-series inspection data, and a related portion via the input / output means based on the retrieved maintenance history data are broken in descending order of influence. Since it has a means for displaying as a place, if the state regarding the processing dimension and accuracy of the product observed at that time is input by the input / output device, all the characteristic patterns conforming to that state are stored in the storage device. Taken out to the processor,
By searching the equipment status corresponding to that pattern,
Of all the basic events related to the failure of the element device that causes the state, the I / O device is displayed as failure points in descending order of influence. As a result, even an unskilled person who has insufficient empirical knowledge about equipment diagnosis can not only quickly and accurately estimate the equipment state, but also by referring to the past maintenance history, equipment abnormality can be detected. It is also possible to apply the maintenance contents at that time after predicting the above.

〔Example〕

The present invention will be described below with reference to FIGS. 1 to 7 by taking an example of the semiconductor manufacturing facility.

Before the detailed description of the present invention is given, the outline of the present invention will be described below.

As described above, in the present invention, the knowledge based on the experience of the skilled engineer, such as the failure related to each manufacturing facility and the case of the countermeasure, the component device configuration, the causal relationship between the failure status and its cause, and the failure occurrence frequency is structured. It is stored in a storage device as a product and is used. By the way, in the case of a semiconductor manufacturing line in which a subtle change in the device of an element causes a defect of the device, a failure state such as a so-called “bad condition” is frequent because of frequent occurrence of product defects. In the state where it is difficult to specify the failure mode, it is not easy to find the failure point by monitoring the individual equipment status. However, some skilled engineers analyze such a failure state by the following method.

(1) Divide the product processing dimensions and accuracy changes over time and trends, the degree of dimensional accuracy variation, the name of the device where such defects frequently occur, and the product processing status into several characteristic patterns. Recall.

(2) Next, the failure location is searched for based on the knowledge of the failure state and the factor (causal relationship of the failure) that can best explain all of the recalled patterns. When searching for a failure location, inspection is performed in order from the element device that is predicted to have the greatest impact on the occurrence of the failure.

In order for even non-skilled persons to be able to diagnose failures by the above method, these product states that can be observed by product inspection are divided into some characteristic patterns,
It is necessary to search for the causal relationship of the fault that best matches all of the actually measured patterns.

In the apparatus according to the present invention in this example, a set of basic events that can characteristically represent the product state is patterned using the cumulative sum method and the V-shaped mask, and the pattern most similar to the current state transition is the past state. It is supposed to be searched from.

Now, the present invention will be described in detail. FIG. 1 shows a schematic configuration of an example of a diagnostic apparatus for manufacturing equipment according to the present invention together with a manufacturing line. In the figure, A _{1 to} An are manufacturing facilities that form a manufacturing line, respectively, and process the material B to be processed into a product. In addition, C ₁ ~ Cn is an inspection device that is provided corresponding to the manufacturing equipment A ₁ ~ An and inspects whether or not the processing there is normally performed, so that the processing dimensions and accuracy can be measured. Has become. The production line is basically composed of these production facilities A _{1 to} An and inspection devices C _{1 to} Cn.

On the other hand, the device according to the present invention basically comprises a storage device D, a processing device F, input / output devices E _{1 to} En, and data transfer means. The storage device D has manufacturing facilities A _{1 to} An.
A history of past failure points, failure states, causes, countermeasures, and measured values of processing dimensions and accuracy of products processed in each failure state is accumulated. More specifically, a series of causal maintenance history data consisting of past failure points, failure states, causes, and countermeasures in each manufacturing facility A _{1 to} An is a maintenance database, quality control such as processing dimensions and accuracy. Data and data relating to equipment operations such as equipment condition setting values and the number of processed wafers are stored in the process database in association with each other. I / O device
To E _{1 to} En, a keyword regarding a failure state, a possible cause of failure, and product accuracy is input in order to extract a series of causal relationships accumulated in the storage device D.
The cause and measures related to the keyword are displayed. The maintenance history of the manufacturing facilities A _{1 to} An is also performed from the input / output devices E _{1 to} En. Further, in the processing device F, one causal relationship of the failure that best matches the keyword input from the input / output devices E _{1 to} En is selected, displayed on the input / output devices E _{1 to} En to instruct maintenance, and the instruction is given. The history of the contents of the maintenance actually performed is attached to the series of causal relationships and is stored in the storage device D.
Has been accumulated in. The processing result of the processing device F is also recorded and displayed on the printer G as needed. Furthermore, the data transfer means is the manufacturing equipment A ₁ ~
An, Interfaces H ₁₁ , H ₁₂ 〜Hn ₁ 〜Hn ₂ for collecting data on processing conditions and processing results for the workpiece B from the inspection devices C ₁ 〜Cn online, and a network interconnected with hardware For controlling communication by connecting a local area network including nodes I ₁₁ , I ₁₂ , H _{13 to} In ₁ , In ₂ , In ₃ for controlling the lowest layer of It is composed of a terminal control device J and a modem K.

Now, FIG. 2 shows a conceptual configuration of a plasma etching apparatus used for manufacturing a semiconductor integrated circuit as an example of an object to be diagnosed. The plasma etching device is for etching the wafer-shaped Si single crystal surface along the resist pattern drawn by the photolithography technique.
The etching is performed by the radicals excited by the high frequency discharge plasma chemically reacting with Si. In the figure, L is a wafer-shaped Si single crystal as an element to be processed, M is a storage container for a gas that is a source of radicals that react with Si, and N is an electrode that applies a high frequency to the gas to generate radicals. O is a high-frequency voltage generator, P is a vacuum pump for exhausting a gaseous reaction product of Si and radicals out of the etching chamber, and Q is a trap for trapping the reaction product when exhausting it out of the chamber. Are shown respectively.

By the way, since the amount of etching is proportional to the reaction time of Si and radicals, it is usually necessary to react for a fixed time in order to obtain the required etching dimensional accuracy. However, the reaction for etching a certain amount due to changes over time in the equipment state, such as reactants adhering to the inner wall of the chamber, the degree of vacuum being lowered, and the density of radicals being lowered due to electrode deterioration. Time may fluctuate, and in the worst case, etching may not proceed at all, resulting in a defective product. In such a case, since the progress of etching and the degree of contamination of the inner wall of the chamber cannot be directly measured, the clues to discover the cause of failure must depend on the processing accuracy and electrical characteristics of the product. Changes in processing accuracy and electrical characteristics can be associated in advance with changes over time in equipment conditions, and the maintenance and adjustment of the apparatus have been performed by such methods up to now.

FIG. 3 shows how the time required to perform a certain amount of etching changes depending on the adjustment of the vacuum degree and the cleaning of the inner wall of the chamber by contamination. It can be seen that it is closely related to the particular change in state of the product. In the figure, oil exchange is to replace the lubricating oil of the vacuum pump, vacuum degree adjustment is to readjust the vacuum degree by adjusting the pressure gauge, and
Complete cleaning means cleaning the inner wall of the chamber. In the past, when changes over time in equipment conditions occurred in combination, it was clarified what changes in product status should correspond to what changes in equipment status over time. I didn't get it.

In the present invention, when the processing device F obtains the time-series transition of the actual process or product by the inspection devices C _{1 to} Cn, the processing device F
Is to solve the above problems by searching for a pattern similar to the inspection data from various time-series patterns in the past, and referring to the state of the equipment in the particular time-series pattern found. .

The present invention will be described below with reference to the example shown in FIG.

That is, first, as a method of grasping the state of the process and the product, for example, in the etching step, the deviation between the time required to reach the complete dimension and the preset specified time is obtained for each product, The cumulative sum of deviations is plotted for each lot. In this example, V with a certain angle and length is used to detect a significant change in the process or product state.
A V-shaped mask (hereinafter referred to as a V mask) is used. This V-mask is superposed on the latest plotted point with a leading distance d as shown in FIG. 4, but its shape is determined by the process, the type of product, the number of samples of measured values, and the required prediction accuracy. θ
The standard of ₁ , θ ₂ and the leading distance d is set as follows.

θ ₂ = tan ⁻¹ (δ) (2) d = −2δ ⁻² loge α ₀ (3) where δ is the standard deviation of the sample mean and α ₀ is the risk rate of the test.

The angle θ _{1 on} the inner side of the V mask represents a range in which the change of the average level is less than or equal to the standard deviation σ when looking back at each point in the past from the present time, and the angle θ ₂ of the outer side changes the average level by 2σ or less. Represents the range of. That is, the inner angle θ ₁ represents the area where the average level is considered to be constant, and the outer angle θ ₂ represents the area where the cumulative sum is entered when the average level fluctuates by the deviation. .

Now, as a method of patterning the time-series transition, attention is paid to five regions divided by the V mask, and each region is represented by which region it passes through. Furthermore,
By associating characters such as *, +, O,-, and # with each area, the time-series transition pattern is stored as a character string. As a result, the function of searching the past state for a specific time-series pattern that is similar to the time-series transition pattern up to the present time stored as a character string is a character string that represents the time-series transition pattern up to the present time. It is realized by comparing and extracting a matching result. In this way, if a pattern similar to the one at the present time can be searched, the current equipment state can be estimated from the maintenance history at that time. This makes it possible to predict equipment abnormalities and apply the maintenance content at that time.

As a specific example, a case where the present method is applied by actually observing an etching time which is one of the process parameters of the plasma etching apparatus in the etching step as shown in FIG. 4 will be described. As mentioned above, the etching time is set to Si if silicon is etched.
And is a reaction time between F ^* radicals, so that the reaction time to achieve a predetermined etching amount varies due to changes in equipment status. Here, the cumulative deviation Sr of etching time is defined as follows.

However, ti is an etching time and t ₀ is a standard value.

When a V mask having a standard shape is applied to the curve showing the accumulated cumulative etching time deviation Sr, the character string representing the transition pattern at the current lot number 13 is compared with the past character string, and a series of comparisons are made. Is 4
Assuming that it is a lot, the transition pattern in lot number 5, It matched with the character string. Next, by searching the maintenance history for the maintenance performed immediately after the processing of lot number 5, it was found that overetching, that is, excessive etching, occurred as a process abnormality. Therefore, using the maintenance history tabulation function, equipment failure items that cause over-etching are displayed in order of occurrence frequency as shown in FIG. 7, and the displayed equipment failure items are inspected in order. The vacuum degree was found to have deteriorated. As a result, abnormal factors in the process were identified early, and product defects could be avoided in advance.

In this way, it is possible to prevent in advance an extreme abnormality of equipment called a failure from the time-series transition of the process or product state.

Finally, the processing procedure of the diagnostic apparatus for manufacturing equipment according to the present invention will be described with reference to FIG.
It consists of (VIII), and the details of blocks (III) and (IV) are shown in FIG.

The flow of processing will be described below mainly with reference to FIG.

(I): First, a parameter for monitoring a change in the state of equipment is selected. This parameter is quality control data such as processing dimensions and accuracy, or data related to equipment operation such as equipment condition setting values and the number of processed wafers. It should be collected online from manufacturing equipment or inspection equipment and then stored in a storage device. It has become. The selection of parameters is performed by the operator via the input / output device for each equipment to be diagnosed.

(II): In order to observe the variation of the average value of the parameter selected in (I), the deviation amount from the standard value of the parameter is integrated. This amount is referred to as a deviation accumulated amount Sr, and is obtained by the above-mentioned equation (4).

(III): A double V mask having angles θ ₁ and θ ₂ and a preceding distance d is applied to a cumulative sum curve obtained by plotting the deviation cumulative value Sr calculated in (II) on a graph. To do. With this double V mask, the accumulated deviation Sr
And the control limit value are compared and patterned as shown in FIG. With the length of the patterning target being i (Ci2, C; constant), the V mask is set with the preceding distance d for each point on the deviation cumulative sum curve at the time of diagnosis. However, the jth (j = 2,3, ...
, I) is associated with a symbol depending on which of the five areas divided by the double V mask belongs to. That is, the correspondence is based on the following judgment logic.

In this way, the character string in which the symbols associated with each point are arranged in order (j = 1,2, ..., i) is defined as Str (0).

(IV): Next, the character string Str (0) is compared with the character string Str (k) (k2) of the past length i. If they do not match in this comparison, the length i of the string to be patterned is updated to be smaller.

i ← i-1 (V): If the character strings match, the history relating to maintenance at that time point k is retrieved from the maintenance database on the recording device.

(VI): If the maintenance history extracted in (V) indicates an abnormality (for example, failure) of the equipment, the location and state of the equipment that causes the abnormality are estimated in the processing device, and as shown in FIG. Display on the I / O device.

(VII): Inspect the locations of the equipment displayed in (VI) in order, and take action if any abnormality is found.

(VIII): Confirm the effect of the treatment performed in (VII), and if there is no effect, return to (I) and reconsider the selection of parameters. If there is an effect, the process ends.

〔The invention's effect〕

As described above, according to the present invention, the state of the manufacturing equipment in the manufacturing line is estimated from the time series transition of the product processing,
Since it is diagnosed and further countermeasure instructions can be given, even non-experts can prevent equipment failures and product processing troubles in advance, reducing equipment downtime and improving product quality. is there.

[Brief description of drawings]

FIG. 1 is a diagram showing a schematic configuration of an example of a diagnostic apparatus for manufacturing equipment according to the present invention together with a structural line, and FIG. 2 is a diagram showing a conceptual configuration of a plasma etching apparatus as an example of a diagnostic object according to the present invention. FIG. 3 is a diagram showing how the time required for etching a certain amount changes depending on the adjustment of the degree of vacuum, and FIG. 4 is a state transition similar pattern search method using a deviation cumulative sum method and a V-shaped mask. Figure to explain,
FIGS. 5 and 6 are diagrams showing a flow of processing in the diagnostic apparatus according to the present invention and a partial detailed flow thereof, and FIG. 7 is an example in which failure items causing overetching are displayed in order of occurrence frequency. FIG. A _{1 to} An …… Manufacturing equipment, B …… Workpiece material, C _{1 to} Cn …… Inspection device, D …… Storage device, E _{1 to} En …… Input / output device, F…
… Processor, H ₁₁ ,, H _{12 to} Hn ₁ , Hn ₂ …… Interface,
I ₁₁ , I ₁₂ , I _{13 to} In ₁ , In ₂ In ₃ ... Local area network node, j ... Termination control device, K ... Modem.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takemasa Iwasaki 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Inside the Institute of Industrial Science, Hitachi, Ltd. (56) Reference JP-A-62-264854 (JP, A) Kai 61-112214 (JP, A) JP 62-228357 (JP, A) JP 62-20008 (JP, A)

Claims (1)

[Claims] 1. A manufacturing line having a plurality of manufacturing facilities for manufacturing products and an inspection device provided for each manufacturing facility and inspecting the processing contents of each product processed by each manufacturing facility in time series. Is a diagnostic device for manufacturing equipment in the past.
A storage means in which a series of causal maintenance history data including causes and countermeasures, quality control data such as product processing dimensions and accuracy, and equipment operation data such as processing condition setting values and number of processing of manufacturing equipment are stored in advance. , A processing means for extracting a related failure point based on inspection data of time-series transition regarding the processing state of the product, and a parameter provided in advance for each manufacturing facility and preset based on quality control data and facility operation data While inputting a record of the state of the failure, the cause of the failure, the processing size and accuracy of the product, the failure point extracted by the processing device can be displayed as a necessary countermeasure point, and the processing means. , Storage means, the input / output means, data transfer means for transmitting and receiving data between the manufacturing equipment and the inspection means, and the processing means At the time when the desired parameter related to the processing state of the product is selected via the input / output means, the inspection data of the time-series transition of each product obtained by the inspection means corresponding to each manufacturing facility is obtained by the cumulative sum method. , The means for plotting the obtained values in time series, and the plotted time series data and the control limit values corresponding to the above selected parameters and predetermined by the type of process or product are compared and plotted. Means for patterning time-series data, and searching the patterned time-series data from various past time-series transition patterns, and the patterned time-series data is a specific time series in the past. When the pattern substantially coincides with the pattern of the statistical transition, a procedure for retrieving causal maintenance history data based on the matched time-series inspection data from the recording means. If, diagnostic apparatus of a manufacturing facility, characterized in that it comprises a means for displaying a failure location associated location via the input means based on maintenance history data the search in descending order of frequency of occurrence.