KR102672799B1 - System and method for matching and analyzing real-time process data of semiconductor equipment - Google Patents

Abstract

The present invention relates to a system for matching and analyzing real-time process data of semiconductor equipment and, more specifically, to a system and method for matching and analyzing real-time process data of semiconductor equipment for pre-response to prediction before a failure. According to the present invention, real-time process data can be collected from semiconductor equipment, analyzed through data mining, and appropriate process parameters can be derived, and data visualization before and after the occurrence of an abnormality of the semiconductor equipment can be provided. When an abnormality of semiconductor/FPD equipment occurs, the data before and after the occurrence of the abnormality can be visualized by using the process data collected in real time.

Description

System and method for matching and analyzing real-time process data of semiconductor equipment}

The present invention relates to a system for matching and analyzing real-time process data of semiconductor equipment, and to a system and method for matching and analyzing real-time process data of semiconductor equipment to predict and respond in advance before failure occurs.

Currently, the field operation of semiconductor/FPD (Flat Panel Display) equipment by domestic companies is operated by monitoring and controlling the process status in front of the equipment in the field using a process control program installed on each equipment. According to this method, when an abnormality occurs in the semiconductor/FPD equipment, the abnormality can be determined through the semiconductor/FPD manufacturer's MES (Manufacturing Execution System) or warning light/buzzer.

In the process of semiconductor/FPD equipment, the process is carried out step by step, and then the measurement process determines whether the product is defective. When it is confirmed that the product has a defect, various log files (Log files) stored in the process control program equipment are processed. File), so a lot of time is wasted in analyzing the cause of an error and coming up with a solution to the problem. In addition, we are faced with the reality that yield and productivity depend on the capabilities of field engineers.

Meanwhile, domestic semiconductor/FPD equipment is currently maintained only with process control and FDC (Fault Detection and Classification) functions within the equipment.

Therefore, in a situation where semiconductor/FPD equipment is highly dependent on foreign countries and mostly relies on imports, the cost of maintaining equipment from major overseas companies is also high for semiconductor manufacturers. Overseas manufacturers provide and operate an Equipment Engineering System (EES) for equipment maintenance. The main functions of this system are process control, equipment productivity, and equipment abnormality monitoring and control, and maintenance costs are high.

Public patent 1999-0004365 (1999.01.15)

The task to be solved is to collect real-time process data from semiconductor equipment, analyze it through data mining, derive appropriate process parameters, and provide a system and method that provides data visualization before and after the occurrence of an abnormality in semiconductor equipment. .

According to one feature, it is a system for matching and analyzing real-time process data of semiconductor equipment, collecting a plurality of semiconductor equipment, real-time process data from the plurality of semiconductor equipment, and matching and comparing the collected real-time process data for each semiconductor equipment. It includes an analysis server that analyzes and a plurality of GUI (Graphic User Interface) devices that visualize and output matching/comparative analysis data of process data received from the analysis server, wherein the analysis server purifies the collected real-time process data, A preprocessing module for preprocessing to remove unnecessary data, input data including failure occurrence records and process data of the plurality of semiconductor equipment based on the preprocessed data, and a binary value indicating whether a failure of the equipment has occurred in the previous process data. A machine learning module that learns output data and a second input data including a second semiconductor equipment failure occurrence record and process data are compared with previous data of the results learned through the machine learning module to evaluate the possibility of equipment failure. Includes predictive maintenance module.

According to the embodiment, process data of each semiconductor equipment is collected through real-time process data matching and analysis between semiconductor/FPD equipment, and data trends are analyzed for each process and time series to identify problems in a short time and apply the optimal recipe. possible.

In addition, when an error occurs in semiconductor/FPD equipment, process data collected in real time can be used to visualize data before and after the error occurs.

In addition, failure factors are derived through data analysis based on facility master data, process data, and equipment log data already held by semiconductor/FPD equipment companies, and rules related to mass production yield are selected and applied. This makes it possible to simulate basic empirical results.

1 is a configuration diagram of a real-time process data matching/analysis system according to an embodiment.
Figure 2 is a block diagram showing the internal configuration of a semiconductor device (EQP) according to an embodiment.
Figure 3 is a block diagram showing the detailed configuration of EQP according to an embodiment.
Figure 4 is a block diagram showing the detailed configuration of an analysis server according to an embodiment.
Figure 5 shows the configuration of a simulation system for deriving and applying EQP optimal process conditions according to an embodiment.
Figure 6 shows a process for detecting errors and reference types through application of EQP process data rules according to an embodiment.
Figure 7 is an example diagram showing visualization of data characteristic analysis results of process parameter/process data matching between EQPs according to an embodiment.
Figure 8 is a flowchart showing a method of matching and analyzing real-time process data of semiconductor equipment according to an embodiment of the present invention.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in various different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts unrelated to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification, when a part “includes” a certain element, this means that it may further include other elements rather than excluding other elements, unless specifically stated to the contrary.

In addition, terms such as “…unit”, “…unit”, and “…module” used in the specification refer to a unit that processes at least one function or operation, which may be implemented through hardware or software or a combination of hardware and software. You can.

The devices described in the present invention are composed of hardware including at least one processor, a memory device, a communication device, etc., and a program that is executed in conjunction with the hardware is stored in a designated location. The hardware has a configuration and performance capable of executing the method of the present invention. The program includes instructions that implement the operating method of the present invention described with reference to the drawings, and executes the present invention by combining it with hardware such as a processor and memory device.

In this specification, “transmission or provision” may include not only direct transmission or provision, but also indirect transmission or provision through another device or by using a circuitous route.

In this specification, expressions described as singular may be interpreted as singular or plural, unless explicit expressions such as “one” or “single” are used.

In this specification, the same reference numbers refer to the same elements regardless of the drawings, and “and/or” includes each and all combinations of one or more of the mentioned elements.

In this specification, terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but the components are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, a first component may be referred to as a second component, and similarly, the second component may be referred to as a first component without departing from the scope of the present disclosure.

In the flowchart described herein with reference to the drawings, the order of operations may be changed, several operations may be merged, certain operations may be divided, and certain operations may not be performed.

FIG. 1 is a configuration diagram of a real-time process data matching/analysis system according to an embodiment, and FIG. 2 is a block diagram showing the internal configuration of a semiconductor equipment (EQP) according to an embodiment.

Referring to Figure 1, the real-time process data matching/analysis system matches and analyzes real-time process data between semiconductor/FPD (Flat Panel Display) equipment.

Here, semiconductor/FPD equipment can be collectively referred to as semiconductor equipment (Equipment, hereinafter referred to as 'EQP'). Additionally, in this specification, equipment may be referred to as semiconductor equipment or EQP.

EQP (110) is connected to the analysis server (130) through the switch hub (120).

The monitoring system 200 includes an operation detection server (Fault Detection and Classification, hereinafter referred to as 'FDC') 210 and a plurality of Graphic User Interface (GUI) devices 220. The FDC server 210 is connected to the analysis server 130 and connected to a plurality of GUI devices 200.

Here, the analysis server 130 and the GUI device 220 may be computing devices equipped with at least one processor, memory, and storage.

The EQP 110 is a device that performs a semiconductor manufacturing process and may be a semiconductor device or an FPD. The analysis server 130 may receive and store real-time process data from the EQP 110 and perform matching and comparative analysis. The GUI device 220 visualizes and outputs the matching/comparison analysis data of the process data collected from the analysis server 130.

The EQP 110 may include an EQP interface module 111, a Cluster Tool Controller (CTC) 112, a Transfer Module Controller (TMC) 113, and a Process Module Controller (PMC) 114.

The EQP interface module 111 is connected to the analysis server 130 through the switch hub 120. The EQP interface module 111 transmits process data to the analysis server 130 through the switch hub 120.

The EQP interface module 111 is installed in the EQP 110 and is linked to the process control program.

The EQP control device 115 transmits the processing/table history/recipe/variable/DCOP (Data Collection of Parameter) data of the EQP (110) to the analysis server 130. You can.

The CTC (112) outputs control data to operate each component of the semiconductor manufacturing equipment according to the operator's operation command input, and receives, samples and stores operation status data measured from each semiconductor manufacturing equipment, and allows the operator to monitor it. Display it on the screen.

TMC (113) is an abbreviation for Transfer Module Controller and is responsible for transporting wafers.

PMC 114 is an abbreviation for Process Module Controller, and is responsible for overall control of vacuum, plasma, temperature, gas flow, etc. to control the wafer process.

The analysis server 130 collects process data including status and process speed data, alarm data, process parameters, etc. from the EQPs 110 in real time, and matches and compares each process data by detailed process by time. can do.

The analysis server 130 is a real-time process data matching and comparative analysis technology that can present appropriate recipes or process parameters to improve production yield of semiconductor/FPD equipment and apply them to each EQP 110.

In FIG. 1, the system for matching and analyzing real-time process data of semiconductor equipment according to an embodiment of the present invention includes a preprocessing module 1391, a machine learning module 1392, and a predictive maintenance module 1393. Includes.

The system for matching and analyzing real-time process data of semiconductor equipment according to this embodiment collects real-time process data from a plurality of semiconductor equipment, and matches and compares and analyzes the collected real-time process data for each semiconductor equipment. It includes an analysis server and a plurality of GUI (Graphic User Interface) devices that visualize and output matching/comparison analysis data of process data received from the analysis server.

At this time, the analysis server according to this embodiment includes a preprocessing module 1391, a machine learning module 1392, and a predictive maintenance module 1393.

The preprocessing module 1391 purifies the collected real-time process data and performs preprocessing to remove unnecessary data.

Here, the preprocessing module 1391 is a component that purifies the collected real-time process data and removes unnecessary data.

Cleaning is necessary to keep the format of the input data consistent, remove noise or errors, and obtain better learning results. For example, you can perform tasks such as identifying and removing outliers or missing values in data, or processing duplicate data. Additionally, the process of removing unnecessary data may be included as part of preprocessing. For example, you can reduce the size of the data and increase the efficiency of the analysis by performing tasks such as removing variables that are not necessary for analysis or consolidating similar variables into one. This preprocessing is an important step that affects the performance of machine learning models and must be performed carefully to ensure that the data can be processed reliably.

The machine learning module 1392 learns input data including failure occurrence records and process data of the plurality of semiconductor equipment based on preprocessed data, and output data having a binary value indicating whether a failure of the equipment has occurred from previous process data. do.

The machine learning module 1392 according to this embodiment performs the function of learning input data and output data. The input data is failure occurrence records and process data of multiple semiconductor equipment. Previously occurring data is also used to identify the cause or pattern of previously occurring failures.

The output data indicates whether a failure has occurred in the equipment expressed as a binary value. The learned model identifies patterns between input data and output data, and based on this, predicts the likelihood of equipment failure when new input data is given.

Through this learning, the predictive maintenance module 1393 can evaluate the possibility of equipment failure by comparing the second input data, including the second semiconductor equipment failure occurrence record and process data, with previous data.

Therefore, the machine learning module 1392 is one of the core modules of the predictive maintenance system that analyzes real-time process data of semiconductor equipment and predicts the possibility of failure.

In machine learning, a model refers to a mathematical modeling of the relationship between input data and output data. Input data is a data set composed of several variables, and output data is the prediction result for the input data, expressed as one variable.

Therefore, identifying patterns between input data and output data means finding interactions, patterns, rules, etc. between variables in these data sets. To achieve this, the model uses various algorithms and techniques to analyze input data and mathematically models the relationship between input and output data.

Through this learning, the model can predict output data with patterns similar to the input data when given new input data. Therefore, if a model is trained using real-time process data of semiconductor equipment as learning data, the model can predict the possibility of failure of the equipment when given new real-time process data.

The predictive maintenance module 1393 evaluates the possibility of equipment failure by comparing the second input data, including the second semiconductor equipment failure occurrence record and process data, with previous data of the results learned through the machine learning module.

This predictive maintenance module is a module that evaluates the possibility of equipment failure by comparing new input data, including secondary semiconductor equipment failure occurrence records and process data, with previous data.

This predictive maintenance module first uses the model learned in the machine learning module to identify the relationship between input data and output data. The model analyzes input data and compares it to previously learned patterns to evaluate the likelihood of equipment failure.

For example, given new input data, the predictive maintenance module compares it to previous data to find patterns similar to previously learned patterns. Based on this pattern, the possibility of failure of the equipment is predicted, visualized, and provided to the user. This allows users to monitor the status of equipment in real time and take necessary actions.

Here, the learned pattern refers to the pattern of data learned from the machine learning model. A machine learning model is trained using previously collected semiconductor equipment failure records and process data to learn the correlation between input data and output data. At this time, the model analyzes various characteristics of the input data and provides the results of a comprehensive evaluation as output data.

For example, the learned model analyzes various factors related to failure of semiconductor equipment, such as operating temperature, pressure, vibration, and power. At this time, the model determines the pattern in which these factors are related from previously learned data. Therefore, the pattern learned from previous data refers to the pattern of factors associated with the possibility of failure of semiconductor equipment.

Similar patterns refer to patterns that the predictive maintenance module finds by comparing it to previous data when analyzing new input data. If a pattern similar to a previously learned pattern is found, the input data is judged to be in a situation similar to a past failure case, and the possibility of failure of the equipment is predicted. By finding similar patterns, the predictive maintenance module can identify problems with the equipment based on previous failure cases and suggest necessary preventive measures.

When new input data is given, the predictive maintenance module according to this embodiment finds a pattern similar to the previously learned first pattern by comparing it with previous data, and the predictive maintenance module transfers the first pattern through big data analysis. A machine learning model is trained using the failure occurrence records and process data of semiconductor equipment collected in to find similar patterns of factors related to the possibility of failure of semiconductor equipment in the data through which the correlation between input data and output data is learned.

The analysis server 130 according to this embodiment can collect various data generated from equipment using IoT sensors to collect and analyze data in real time, and transmit and process the data in real time. This enables rapid collection and processing of data and improves system performance.

Referring to FIG. 2, the analysis server 130 includes a plurality of data collection modules 131, a plurality of data analysis modules 132, and a plurality of graphic user interface (GUI) interface modules. (133), a data storage module (Database Module) (134), and an analysis database (135).

The plurality of EQPs 110 include each EQP interface module 111.

The plurality of GUI devices 220 each includes a communication module 221 and a screen viewer 222.

A plurality of data collection modules 131 receive real-time process data from a plurality of EQPs 110 through Ethernet communication and transmit it to a plurality of data analysis modules 132. At this time, the plurality of data collection modules 131 may collect real-time process data collection management and process result data, that is, process speed data, process parameters, detailed process data, and recipes, and transmit them to the plurality of data analysis modules 132. .

The plurality of data analysis modules 132 may generate data by matching and comparatively analyzing the received real-time process data. A plurality of data analysis modules 132 analyze real-time process data and process result data through data mining to derive appropriate process parameters, and analyze failure factors through data visualization before and after the occurrence of an error in the EQP (110). You can.

Matching/comparative analysis data generated by the plurality of data analysis modules 132 may be stored in the analysis database 135 through the plurality of data storage modules 134.

Matching/comparative analysis data generated by the plurality of data analysis modules 132 may be transmitted to the plurality of GUI devices 220 through the plurality of GUI interface modules 133.

At this time, the analysis server 130 includes a plurality of data collection modules 131, a plurality of data analysis modules 132, and a plurality of GUI interface modules for processing each process data for each EQP 110 and GUI device 220. It may include (133).

A plurality of data collection modules 131 communicate with a plurality of EQP interface modules 111 over Ethernet. A plurality of data collection modules 131 may be provided for each EQP (110). That is, data collection module #1 (131) is dedicated to Ethernet communication with EQP interface module #1 (111) of EQP #1 (110), and data collection module #2 (131) is responsible for EQP interface module #1 (111) of EQP #2 (110). It is dedicated to Ethernet communication with interface module #2 (111), and data collection module #n (131) is dedicated to Ethernet communication with EQP interface module # n (111) of EQP # n (110).

A plurality of data analysis modules 132 receive real-time process data from a plurality of EQP interface modules 111, and transmit the received real-time process data to a plurality of data analysis modules 132 and a plurality of data storage modules 134. You can.

The plurality of data analysis modules 132 collect real-time process data for each EQP from the plurality of data collection modules 131. That is, data analysis module #1 (132) receives real-time process data of EQP #1 (110) from data collection module #1 (131) and transfers it to GUI interface module #1 (133). Data analysis module #2 (132) receives real-time process data of EQP #2 (110) from data collection module #2 (131) and transmits it to GUI interface module #2 (133). The data analysis module #n (132) receives real-time process data of the EQP #n (110) from the data collection module #n (131) and transmits it to the GUI interface module #n (133).

According to the embodiment, the plurality of data analysis modules 132 may use TTTM (Tool To Tool Match) technology. TTTM is a technology developed to analyze logs and data created in equipment based on .NET Frameworks 4.0. The plurality of data analysis modules 132 execute the TTTM program and each of the EQPs 110 uses them. It can provide a 1:1 comparison function between recipes, FA Variables, and process data.

A plurality of data analysis modules 132 collect and store real-time process data and recipes for each EQP 110, and use real-time or stored data to simultaneously produce data time series of multiple EQPs 110. Comparative analysis can be performed. Conventionally, it was difficult to collect real-time process data, so log files and recipes created in semiconductor/FPD equipment were read from the semiconductor/FPD equipment and the data was compared and analyzed, so only at sites where semiconductor/FPD equipment is installed. It is possible to check. To use it externally, the log file and recipe could be copied from the semiconductor/FPD equipment and saved to a separate computer with the TTTM program installed. Therefore, only 1:1 comparison between semiconductor/FPD equipment is possible. However, in an embodiment of the present invention, data time series comparative analysis of several EQPs 110 can be performed simultaneously.

In this way, the plurality of data analysis modules 132 compare/match the main parameters of process data collected from multiple EQPs 110, recipes, and log files recorded in the equipment between devices. and can be analyzed. Conventionally, EQP only provides information managed by the manufacturer's MES or FDC, so collecting accurate process data requires linking with process control programs installed in the equipment and the contents of the log file. However, the embodiment of the present invention can be applied to the EQP (110) by deriving optimal process conditions through process data comparison/matching analysis between the EQP (110).

The plurality of data analysis modules 132 generate a specification model by applying statistical techniques based on process reference data and process information, and can detect changes in process information of each single EQP (110) among the EQPs (110). .

The plurality of data analysis modules 132 may monitor data status and provide a graph showing comparison of key status values using process data.

In addition, the data analysis module 132 derives failure factors through data analysis based on the equipment master data, process data, and equipment log data of the EQP (110) and selects rules related to mass production yield. By applying the basic verification results, the modules 221 receive data obtained by matching/comparing and analyzing real-time process data from each GUI interface module 133.

Based on the matching/comparison analysis data received from the GUI interface modules 133, the communication modules 221 visualize the entire EQP operation status of the EQPs 110 through each connected screen viewer 222. Can be printed. The entire EQP operation situation can be implemented in the form of a graph optimized for time series matching data between EQPs (110).

The communication modules 221 may provide an alarm function when an event occurs based on matching/comparison analysis data received from the GUI interface modules 133. The communication modules 221 access the analysis database 135 through the GUI interface modules 133 and provide search and search functions by conditions such as process data, process parameters, and recipes for each stored EQP 110. can do.

FIG. 3 is a block diagram showing the detailed configuration of the EQP according to an embodiment, and shows the detailed configuration of the EQP 110 described in FIGS. 1 and 2.

Referring to FIG. 3, the EQP 110 may include an EQP interface module 111, a CTC 112, a TMC 113, a PMC 114, and an EQP control device 115. Since the EQP interface module 111, CTC 112, TMC 113, and PMC 114 are described in FIGS. 1 and 2, their description will be omitted.

The EQP control device 115 includes a viewer (115-1), an analyzer (115-2), a local connector (115-3), a remote connector (115-4), a remote receiver (115-5), and an equipment database (115-5). 6) may be included.

The viewer 115-1 outputs processing data, table history data, recipes, variables, and DCOP collected from the semiconductor process modules CTC 112, TMC 113, and PMC114 to the analyzer 115-2.

The analyzer (115-2) connects the processing data, table history data, recipes, variables, and DCOP collected from the viewer (115-1) to the local connector (115-3), remote connector (115-4), and remote receiver (115-115-2). It is output to the EQP interface module 111 through 5).

These processing data, table history data, recipes, variables, and DCOP are transmitted to the analysis server 130 through the EQP interface module 111.

The analyzer 115-2 stores processing data, table history data, recipes, variables, and DCOP collected from the viewer 115-1 in the equipment database 115-6.

Figure 4 is a block diagram showing the detailed configuration of an analysis server according to an embodiment.

At this time, Figure 4 shows a detailed embodiment configuration of the analysis server 130 described in Figures 1 and 2.

Referring to FIG. 4, the analysis server 130 includes a plurality of data collection modules 131, a plurality of data analysis modules 132, a plurality of GUI interface modules 133, a data storage module 134, and a Relational Data Base (RDBMS). Management System) (135a), NoSQLDB (135b), main processor 136, batch processor 137, and message queue handler (138).

As described in FIG. 2, a plurality of data collection modules 131, a plurality of data analysis modules 132, and a plurality of GUI interface modules 133 are respectively connected to a plurality of EQPs 110 and a plurality of GUI devices 220. Link.

At this time, in order to prevent message bottlenecks, an independent dedicated interface for each EQP is linked, that is, a plurality of data collection modules 131, and the plurality of data collection modules 131 and the plurality of data analysis modules 132 are connected to a message queue (Message Queue type internal communication can be performed.

The plurality of data analysis modules 132 can compare and analyze real-time large-scale data processing techniques, such as distributed processing methods or streaming methods, and apply data processing techniques suitable for operation for each EQP.

The data storage module 134 stores standardized data collected from the EQPs 110 by a plurality of data collection modules 131, such as equipment information, process parameters, recipes, etc., through the message queue handler 138. It is transmitted by the main processor 136 and stores this standardized data in the RDBMS 135a, a relational database system.

The RDBMS 135a stores information that requires a relational structure, such as user information, EQP information, process parameters, equipment setting information, recipe information, events, and alarms.

The plurality of data analysis modules 132 may store unstructured data, such as process data and variables, in the NoSQLDB 135b that does not define a schema.

The NoSQLDB 135b stores data that must be managed as real-time time series information, such as process data and process element values, and can be stored in key value format or column family format.

The plurality of data analysis modules 132 may simultaneously perform matching of process parameters and/or process data of several EQPs 110 based on TTTM (Tool To Tool Match) technology.

The plurality of data analysis modules 132 may select process data matching elements. At this time, data to be matched may include Processing, Table History, Recipe, Variable, DCOP (Data Collection of Parameter), etc., and process data elements to be added may be selected.

Unlike the prior art, the plurality of data analysis modules 132 may store data to be matched in a separate database table.

The plurality of data analysis modules 132 may display different data and display detailed item graphs when analyzing matching data. The plurality of data analysis modules 132 can display data that differs as a result of data matching with separate shading and select items that require detailed analysis among the matched data to display them in a graph. The plurality of data analysis modules 132 may perform correlation analysis between EQPs through matching data mining.

The plurality of data analysis modules 132 can detect errors and reference types and model failure types through application of EQP process data rules.

The plurality of data analysis modules 132 may provide a rule service for determining defects. A plurality of data analysis modules 132 subdivide the process data of each EQP 110 into lot/substrate step units to analyze trends and patterns, and serve as a standard for determining defects. By reflecting the (Nelson Rules), errors and standard types of problematic sections can be detected.

The plurality of data analysis modules 132 may detect errors that do not meet predetermined standards for determining defects based on analyzed trends and patterns. By repeating this process, the plurality of data analysis modules 132 can collect various types of errors and standards and model types of failures in the process that may occur in various situations.

The plurality of data analysis modules 132 may analyze the characteristics of process parameters and/or process data matching result data between the EQPs 110.

The plurality of data analysis modules 132 may perform statistical logic suitable for EQP matching data mining, for example, statistical logic for data characteristic analysis and test verification of at least one of univariate statistics, model-based selection, and iterative selection.

The plurality of data analysis modules 132 may apply statistical logic to process parameters and/or process data matching data between the EQPs 110 to calculate a failure threshold standard according to process data characteristic analysis.

The plurality of data analysis modules 132 may perform correlation analysis between the EQP 110 through process failure type modeling and process data characteristic analysis and visualization. The plurality of data analysis modules 132 can derive standard conditions for predicting situations before equipment failure occurs through failure type modeling.

The plurality of data analysis modules 132 can compare production efficiency between equipment by analyzing and visualizing process data characteristics. By identifying the correlation between production efficiency and process conditions, optimal process conditions (Recipe, Variables, etc.) can be suggested.

The plurality of data analysis modules 132 can confirm and verify the production process results by applying optimal process conditions to the EQP (110).

The plurality of data analysis modules 132 may perform failure factor analysis through data visualization before and after the occurrence of an error between the EQPs 110. At this time, the plurality of data analysis modules 132 can visualize data before and after the EQP abnormality occurs.

A plurality of data analysis modules 132 collect real-time EQP status data and process data, and when an EQP abnormality occurs, the data is stored in a separate alarm table with the alarm occurrence time and corresponding alarm information. Together, real-time status data and process data can be stored. Based on the stored data, the plurality of data analysis modules 132 can query the EQP abnormality occurrence data history under conditions for each period (month/week/day).

The plurality of data analysis modules 132 can select search result data to check visualization contents. The plurality of data analysis modules 132 can perform failure factor analysis through data visualization before and after the EQP error occurs.

According to one embodiment, the plurality of data analysis modules 132 may represent the value of each element of the process data stored under the time conditions before and after the occurrence of the abnormality determined by the user and the time of occurrence of the abnormality in a graph.

According to another embodiment, the plurality of data analysis modules 132 convert the process data received and stored in real time from the EQP (110) into packet format so that the user can check the detailed status of the actual EQP (110). It can be expressed on the screen in the same way.

Analysis and correlation of process data and influence between elements can be derived based on advisory data input from EQP experts.

The plurality of data analysis modules 132 can derive EQP optimal process conditions and perform application simulation.

The main processor 136 operates in conjunction with the message queue handler 138 and the GUI interface module 133, and in conjunction with the data storage module 134.

The main processor 136 may transfer the structured data collected from the message queue handler 138 to the data storage module 134.

The main processor 136 may transmit structured data and unstructured data collected from the message queue handler 138 to the GUI device 220 through the GUI interface module 133.

The batch processor 137 may perform a batch function of data stored in the NoSQLDB 135b.

Figure 5 shows the configuration of a simulation system for deriving and applying EQP optimal process conditions according to an embodiment.

Referring to FIG. 5, a plurality of EQP simulators 300 are connected to the analysis server 130, and the analysis server 130 is connected to the GUI device 220.

The EQP simulator 300 is an EQP for test verification and performs a simulation function by the analysis server 130.

The data collection module 131 of the analysis server 130 is a plurality of data collection modules that collect real-time process data from a plurality of semiconductor devices and are dedicated interfaces for Ethernet communication with the plurality of semiconductor devices, including an EQP interface and status data. Performs collection, process data collection, EQP event collection, and setup parameter collection.

The data analysis module 132 of the analysis server 130 can perform EQP data trend analysis, alarm/state relationship analysis, and Job/process/recipe relationship analysis. The data analysis module 132 matches and compares real-time process data and is a plurality of data analysis modules that each correspond to the plurality of data collection modules.

The data storage module 133 of the analysis server 130 stores EQP standard information, EQP operation information, status data, process data, EQP events, setting parameters, and various histories generated by the data analysis module 132, and Information management and periodic data backup can be performed.

The data analysis module 132 of the analysis server 130 may perform a simulation function. At this time, the data analysis module 132 can collect data transmitted/received in real time while the EQP 110 process is in progress. The data analysis module 132 may generate situation analysis data that may occur during the process of the EQP (110). The data analysis module 132 may generate situation analysis data such as process parameters in normal operation/error occurrence situations.

The data analysis module 132 can create scenarios for each situation based on data collected during the process and situation analysis data that may occur. The data analysis module 132 can implement a settings editor for changing scenarios and recipes, and changing process data values.

The data analysis module 132 may perform system verification using the EQP simulator 300. The data analysis module 132 can vary the data transmission/reception speed of the EQP simulator 300 to verify the system data processing speed.

The data analysis module 132 can generate a failure situation using the EQP simulator 300 to check system failure recognition function, analyze failure factors, analyze correlation between EQPs, and predict production volume using the EQP simulator 300. . At this time, the data analysis module 132 can transmit normal process data and discontinuous event data from the EQP simulator 300 to predict production volume according to maximum production volume and number of failures.

Figure 6 shows a process for detecting errors and reference types through application of EQP process data rules according to an embodiment, and Figure 7 is an example diagram showing visualization of data characteristic analysis results of process parameter/process data matching between EQPs according to an embodiment. .

Referring to FIG. 6, the data analysis module 132 analyzes raw data collected from the EQPs 110 (S101) and analyzes trends/patterns of process data based on this (S102).

The data analysis module 132 can apply a predefined rule service to trend/pattern analysis data (S103) and perform error detection and confirmation (S104).

Referring to FIG. 7, the data analysis module 132 can express trend/pattern analysis in the form of a graph and apply a rule service that sets a threshold according to statistical logic.

The embodiments of the present invention described above are not only implemented through devices and methods, but can also be implemented through a program that implements functions corresponding to the configuration of the embodiment of the present invention or a recording medium on which the program is recorded.

Figure 8 is a flowchart showing a method of matching and analyzing real-time process data of semiconductor equipment according to an embodiment of the present invention.

In FIG. 8, a plurality of semiconductor devices, an analysis server that collects real-time process data from the plurality of semiconductor devices, matches and compares and analyzes the collected real-time process data for each semiconductor device, and matching/matching process data received from the analysis server. In a method using a system (hereinafter referred to as the system) that matches and analyzes real-time process data of semiconductor equipment including a plurality of GUI (Graphic User Interface) devices that visualize and output comparative analysis data, (a) the system The collected real-time process data is refined and pre-processed to remove unnecessary data.

(b) Based on the preprocessed data, the system learns input data including failure occurrence records and process data of the plurality of semiconductor equipment, and output data having a binary value indicating whether a failure has occurred in the equipment from previous process data.

(c) Next, the system evaluates the possibility of equipment failure by comparing the second input data, including the second semiconductor equipment failure occurrence record and process data, with previous data of the results learned through the machine learning module.

(d) Next, the system visualizes and outputs the matching/comparison analysis data of the process data received from the analysis server and the results of step (c) above.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention defined in the following claims are also possible. It falls within the scope of rights.

110:EQP
111: EQP interface module
112:CTC
113:TMC
114:TMC
115-1: Viewer
115-2: Analyzer
115-3: Local connector
115-4: Remote connector
115-5: Remote receiver
115-6: Equipment database
120: switch hub
130: analysis server
131: Data collection module
132: Data analysis module
133: GUI interface module
134: data storage module
135: Analysis database
136: main processor
137: batch processor
138: Message Queen Handler
1391: Preprocessing module
1392: Machine learning module
1393: Predictive maintenance module
220: GUI device
221: Communication module
222: Screen viewers
300: EQP simulator

Claims (3)

A system that matches and analyzes real-time process data of semiconductor equipment,
plural semiconductor equipment,
Collect real-time process data from the plurality of semiconductor devices, match and compare the collected real-time process data for each semiconductor device, and subdivide the process data of each of the plurality of semiconductor devices into lot/substrate step units to determine trends and patterns. Analysis server that analyzes and
It includes a plurality of GUI (Graphic User Interface) devices that visualize and output matching/comparative analysis data of the process data received from the analysis server,
The analysis server includes a preprocessing module that purifies the collected real-time process data and preprocesses it to remove unnecessary data;
A machine learning module that learns input data including failure occurrence records and process data of the plurality of semiconductor equipment based on the preprocessed data, and output data having a binary value indicating whether a failure of the equipment has occurred in the previous process data, and
It includes a predictive maintenance module that evaluates the possibility of equipment failure by comparing second input data including second semiconductor equipment failure occurrence records and process data with previous data of the results learned through the machine learning module, wherein the prediction The maintenance module is characterized by learning the correlation between input data and output data by learning a machine learning model using collected failure occurrence records and process data of semiconductor equipment to find similar patterns of factors related to the possibility of failure of semiconductor equipment in the learned data. and
The plurality of GUI (Graphic User Interface) devices provide an alarm function when an event occurs based on matching/comparison data received from the analysis server, and provide process data, process parameters, and recipes for each of the plurality of semiconductor devices stored by accessing the analysis server. A communication module that provides search and search functions by condition; and
A screen viewer that connects and outputs the entire semiconductor equipment operation status visualizing the status of the plurality of semiconductor devices based on the matching/comparison analysis data received from the analysis server, but displays the entire semiconductor equipment operation status between semiconductor devices. Implemented in graph form according to time series matching data,
The analysis server analyzes real-time process data and process result data through data mining to derive process parameters, and analyzes failure factors through data visualization before and after the occurrence of an error in semiconductor equipment (EQP).
Through analysis of logs and data created from each semiconductor equipment (EQP), each recipe, FA Variable, and process data used in multiple EQPs are compared 1:1, allowing data from multiple EQPs to be collected at the same time. A system for matching and analyzing real-time process data of semiconductor equipment, characterized by performing time series comparative analysis. According to paragraph 1,
When given new input data, the predictive maintenance module compares previous data to find a pattern similar to the previously learned first pattern,
The predictive maintenance module learns the first pattern through big data analysis using previously collected failure occurrence records and process data of semiconductor equipment to learn the correlation between input data and output data. A system that matches and analyzes real-time process data of semiconductor equipment, characterized by finding similar patterns of factors related to the possibility of equipment failure. A plurality of semiconductor devices, an analysis server that collects real-time process data from the plurality of semiconductor devices, matches and compares and analyzes the collected real-time process data for each semiconductor device, and matching/comparison analysis data of the process data received from the analysis server. In a method using a system for matching and analyzing real-time process data of semiconductor equipment including a plurality of graphic user interface (GUI) devices that visualize and output,
(a) Preprocessing to purify the real-time process data collected by the system and remove unnecessary data,
(b) The system learns input data including failure occurrence records and process data of the plurality of semiconductor equipment based on the preprocessed data, and output data having a binary value indicating whether a failure of the equipment has occurred from previous process data. steps to do,
(c) the system evaluates the possibility of equipment failure by comparing second input data including second semiconductor equipment failure occurrence records and process data with previous data of results learned through a machine learning module; and
(d) comprising the step of visualizing and outputting the matching/comparison analysis data of the process data received by the system from the analysis server and the results of step (c),
In step (b), real-time process data is collected from the plurality of semiconductor devices, the collected real-time process data is matched and compared for each semiconductor device, and the process data of each of the plurality of semiconductor devices is analyzed on a lot/substrate step basis. Analyze trends and patterns by segmenting them into
In step (c), a machine learning model is trained using the collected failure occurrence records and process data of semiconductor equipment to identify similar patterns of factors related to the possibility of failure of semiconductor equipment in the learned data and the correlation between input data and output data. It is characterized by finding,
Step (d) provides an alarm function when an event occurs based on matching/comparison data received from the analysis server, and accesses the analysis database through the GUI interface module to provide process data, process parameters, and recipes for each plurality of semiconductor devices stored. Provides inquiry and search functions by condition, and outputs the entire semiconductor equipment operation status visualizing the status of the plurality of semiconductor equipment based on the matching/comparison analysis data received from the analysis server by connecting to each screen viewer, The semiconductor equipment operation status is implemented in the form of a graph based on time series matching data between semiconductor equipment.
In the analysis server, real-time process data and process result data are analyzed through data mining to derive process parameters, and failure factors are analyzed through data visualization before and after the occurrence of an error in the semiconductor equipment (EQP).
Through analysis of logs and data created from each semiconductor equipment (EQP), each recipe, FA Variable, and process data used in multiple EQPs are compared 1:1, allowing data from multiple EQPs to be collected at the same time. A method of matching and analyzing real-time process data of semiconductor equipment, characterized by performing time series comparative analysis.