KR101998972B1 - Method of analyzing and visualizing the cause of process failure by deriving the defect occurrence index by variable sections - Google Patents

Abstract

A failure occurrence index for each variable section is derived, and a cause of the process failure is identified and visualized. The method includes the steps of obtaining first data including quality inspection result data measured by at least one sensor installed in each of a plurality of processes and second data including process data, A step of preprocessing individual data of each of the first data and the second data, generating analysis target data by integrating the preprocessed individual data, and analyzing the poor data based on univariate analysis Analyzing the cause according to individual variables, and deriving the failure cause index according to the analysis result based on the univariate analysis.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and a device for visualizing a cause of an inferiority of a process,

An embodiment of the present invention relates to a process failure analysis technique, and more particularly, to a method of determining a cause of a process failure by visualizing a failure occurrence index for each variable section and visualizing the cause.

Recently, the manufacturing process line for producing products has become complicated in accordance with the trend of intelligent and high performance of electric / electronic products and communication products. In particular, there is a need to increase the reliability of products and reduce costs by predicting defects in manufacturing lines including various processes.

As described above, in order to cope with quality problems after shipment of products, it is necessary to track manufacturing history information on manufactured goods, effectively perform quality cause analysis or simulation, and to allow end users to visually or intuitively grasp the results of the performance .

Korean Patent Laid-Open No. 10-2016-0076646 (Jul. Korean Patent Registration No. 10-1530848 (June 17, 2015)

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the related art, and it is an object of the present invention to provide a method and apparatus for tracking manufacturing history information on a manufactured product, effectively performing quality cause analysis or simulation and visually or intuitively And to provide a method for providing such information.

It is an object of the present invention to provide a method of determining a cause of a process failure by visualizing a defect occurrence index for each variable section and visualizing the cause.

In order to accomplish the above object, there is provided a method for analyzing and visualizing a cause of a process defect by deriving a defect occurrence index for each variable section according to an aspect of the present invention (hereinafter, briefly referred to as a process defect cause analysis method) The method according to any one of claims 1 to 3, wherein the first defect data includes at least one or more sensors that are installed in a plurality of processes, Obtaining second data including data; Pre-processing individual data of each of the first data and the second data; Integrating the preprocessed individual data to generate analysis target data; Analyzing the cause of the failure based on univariate analysis expressing information affecting the final quality by numerical value for each individual variable; And deriving a failure cause index according to the analysis result based on the univariate analysis.

In one embodiment, the step of preprocessing the individual data may comprise a variable generating process, a numeric variable transforming process, a meaningful variable generating process, or a combination thereof.

In one embodiment, the variable generation process may include converting the matrix-shaped data into column data on a separate product basis.

In one embodiment, the step of converting the numeric type variable comprises setting a nominal variable as a new item in the individual data including the nominal variable and processing the value of the new item for the corresponding product of the nominal variable as a binary can do.

In one embodiment, the generating of the meaningful variable may include: obtaining a rate of occurrence of a failure in each process based on an average, a maximum value, or a minimum value of the sensed value in the selected process; May be set or operated to be at least 50% of the total analysis data for the selected process.

In one embodiment, the step of generating the analysis target data may include dividing individual variables in one of the plurality of processes into a plurality of sections for a product manufactured through a plurality of processes, It is possible to generate an index based on the number of defects in each section of the sections and the number of defects or defects and non-defects.

In one embodiment, the process failure cause analysis method may further include outputting information for visualizing the quality cause based on the failure cause index after the deriving step.

According to the present invention, it is possible to effectively derive a cause of a quality problem, a process variable, and an influence index for analyzing a cause of a final quality defect of a product produced on a manufacturing line based on univariate analysis.

In addition, according to the present invention, it is possible to increase the reliability of processes and products and to reduce the occurrence of defects by controlling the process parameters according to the result of the cause analysis based on the derived quality problem causal factors and the influence index, By providing a visual or intuitive user interface for the index, it is possible to maximize user convenience for process monitoring and process control.

In addition, according to the present invention, it is possible to effectively grasp the cause of defects in the manufacturing process by analyzing the influence of individual factors based on the statistical test methodology.

FIG. 1 is a flowchart of a method of deriving and visualizing a cause of a process failure using a machine learning model in a data unbalance environment according to an embodiment of the present invention (hereinafter, simply referred to as a "process failure cause analysis method").
FIG. 2 is a flowchart of a manufacturing process to which the process failure cause analysis method of FIG. 1 can be applied.
FIGS. 3 to 6 are diagrams for explaining the individual data preprocessing process among the processes of the process failure cause analysis method of FIG.
FIG. 7 is a diagram for explaining a data integration process in the process failure cause analysis method of FIG. 1. FIG.
FIGS. 8 to 10 are diagrams for explaining a univariate analysis process among the processes of the process failure cause analysis method of FIG.
FIGS. 11 and 12 are views for explaining a process of deriving and visualizing a failure cause factor among the processes of the process failure cause analysis method of FIG.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention is capable of various modifications and various embodiments, and specific embodiments are illustrated in the drawings and described in detail in the detailed description. It is to be understood, however, that the invention is not to be limited to the specific embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. It is also to be understood that the terms " comprising, "" including," or "having ", when used in this specification, specify the presence of stated features, integers, But do not preclude the presence or addition of one or more other features, numbers, steps, operations, elements, parts, or combinations thereof.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a flowchart of a method of deriving and visualizing a cause of a process failure using a machine learning model in a data unbalance environment according to an embodiment of the present invention (hereinafter, simply referred to as a "process failure cause analysis method").

Referring to FIG. 1, the process failure cause analysis method according to the present embodiment includes a data collection step S11, an individual data preprocessing step S12, a data integration step S13, an analysis object data preprocessing step S14, A variance analysis based failure factor analysis step (S15), a failure cause index derivation step (S16), and a failure cause index visualization step (S17).

The process failure cause analysis method of this embodiment can be used to analyze the cause of quality defect using multivariate analysis in a manufacturing process including various various processes.

In the data collecting step S11, the individual data preprocessing step S12 and the data merging step S13, undersampling is performed on the good data to adjust the ratio of the good product and the bad to a predetermined ratio, Can be adjusted.

In the analysis target data preprocessing step (S14) and the univariate analysis based failure cause factor analysis step (S15), the failure cause factor can be grasped based on the learning data set.

Then, the process failure cause derivation method can derive the failure cause index after univariate analysis and output information for visualizing the quality cause on the screen of the display device based on the derived failure cause index.

FIG. 2 is a flowchart of a manufacturing process to which the process failure cause analysis method of FIG. 1 can be applied.

The process failure cause analysis method according to the present embodiment is characterized in that the bar code is printed, the application cream is printed, the lead application (SPI) inspection is performed as indicated by various steps S21 to S30 in Fig. 2, (AOI) inspection is carried out, and the current flow inspection (in) is performed by inserting the device, placing the device by a substrate mounting (SMD) method, performing reflow, soldering and baking, (vectorization), data integration, and data collection in a manufacturing process of at least one specific product including various processes including performing a circuit tester (ICT) and performing an in-process tester (IFT) We perform a series of processes such as preprocessing of data to be analyzed, univariate analysis (failure factor analysis), derivation of failure factor index and visualization of failure factor index, and manufacturing information It can be tracked and visualized.

The process failure cause analysis method according to the present embodiment can be applied to the above-described manufacturing process.

Data collected in the SPI inspection process can be generated in a total of 16,177 columns (e.g., 14,847 * 16,177 matrices). The matrix may include an identifier (ID), a processing machine, a date, a time, and a number of counts.

Data preprocessing in the SPI inspection process can be quantified in the number of good / bad, good / bad, and reflected in the overall inspection results for each inspection item. Duplicate data may be selected as the last collected or measured value.

The data preprocessing in the reflow process can be performed to include the results and transform values for each measurement using the 13 highest and 13 lowest values for the two orders (1, 2), respectively. have. For example, by preprocessing the reflow process data, a table of 1,130,863 columns and 606,451 columns can be created.

Also, data preprocessing in the reflow process can process each order data into a single row (14,847 * 63 matrix). Results for each order can be shown. No bad joins are allowed in the table, and only OK can be left. Then, the worker information can be displayed on the table.

Also, the data preprocessing in the AOI process can use OK or bad (NG) information in the header table. The AOI process is usually checked once in each of two facilities, and in case of failure, it can be inspected several times in each facility. Data preprocessing can index good and bad information for each inspection facility in AOI process as 0 and 1. If there is no specific inspection facility data value, a dummy value such as -9999 can be generated. In addition, in the data preprocessing, if there are many times of inspection, the possibility of a problem is implied, so that a column of the number of times of inspection can be used.

Data preprocessing in the ICT inspection process expresses measurement values (measure_value) and deviation for each ICT inspection step (for example, 1,533 pieces), and outputs final results (good / bad or OK / NG ) And the number of test variables. According to this data preprocessing, for example, 3068 variables can be generated, and a table in which the OK / NG information and the number of tests are generated in the last two columns can be provided.

As described above, according to the present embodiment, the data is collected in the reflow / soldering (burning) process, and the SPI inspection (coating cream), the AOI inspection (correct position inspection), the ICT inspection (current flow) Motion), and collect other process environmental data through static electricity, vibration sensors, and external weather. Then, after individual data preprocessing is performed on the collected data, the data is integrated, and the integrated data is preprocessed as the analysis target data, and the unfavorable factor of the cause can be analyzed by univariate analysis. Using the results of the univariate analysis, the process failure cause index can be derived and visualized, intuitively displayed to a user, a manager, etc., or signals or data necessary for process analysis, process control, process control, and the like can be transmitted to a preset control terminal.

FIGS. 3 to 6 are diagrams for explaining the individual data preprocessing process among the processes of the process failure cause analysis method of FIG.

The process failure cause analysis method according to the present embodiment may include at least one of a process of generating a variable, a process of converting a numerical variable, and a process of generating a meaningful variable, as individual data preprocessing processes.

Specifically, as shown in FIG. 3 (a), the variable generating process is a process of generating a variable such as an identifier (ID) of a specific product A, an item corresponding to a sensor number, a value 1, a value 2 corresponding to a sensing value If the measured values from several sensors in various processes for the product are in several rows in the collection data, the column (s) to make it exist in one row per product ) Can be extended to perform data vectorization. As described above, in this embodiment, data of 4 rows and 4 columns can be vectorized into data of 2 rows and 7 columns, and individual data can be preprocessed.

As shown in FIG. 3 (b), when the result and cause of the products A, B, C, and D are stored as nominal variables, You can create and process it as a binary. Thus, in this embodiment, the nominal variables of the result item, OK and NG, and the nominal variables of causal items EE, AA, and DD are classified into the cause items AA, cause BB, cause CC, cause DD, EE binary value (0 or 1) can be processed.

4, when there are a plurality of raw data for each pad identifier (PADID) in one product (see FIG. 3 (a)), It is possible to generate preprocessing data generated by individual variables (see FIG. 3 (b)). In the preprocessing data, an abnormal number of abnormal types can be additionally generated as a variable.

As shown in FIG. 5, when the weather information (a) exists in a nominal form, the process of converting the numeric variables described above can perform individual data preprocessing for converting each weather condition into a numeric variable. FIG. 5A shows raw data and FIG. 5B shows an example of preprocessed data.

On the other hand, the process of generating meaningful variables is to generate meaningful variables such as individual weather and whether the maximum temperature, humidity, and rain have ever reached during the manufacturing process (IS_RAIN) if the weather is an important factor in the manufacturing process . In the meaningful variable generating process, as shown in FIG. 6, an average, maximum or minimum value variable of the process progress can be generated from the weather information.

FIG. 7 is a diagram for explaining a data integration process in the process failure cause analysis method of FIG. 1. FIG.

After the individual data preprocessing process described above, the data integration process can be performed. In the data integration process, as shown in FIG. 7, the entire data is extracted in a small number of bad data sets, and only a small amount of data is extracted in a relatively large good data set to optimize the configuration of the integrated data set Can be designed.

FIGS. 8 to 10 are diagrams for explaining a univariate analysis process among the processes of the process failure cause analysis method of FIG.

Univariate analysis that can be employed in the process failure cause analysis method according to the present embodiment may include a process of expressing the degree of influence on the final quality for each individual variable by an index. That is, in order to consider the influence of the variable interval, as shown in FIG. 8, an index may be generated based on the number of defects and the number of good items for each interval by dividing the individual variable into n equals (7 equals in this embodiment) . The number of defects or good products can be expressed in terms of density for each sub-section of the process for that variable.

Also, the univariate analysis process can analyze the influence of sub-interval or variable interval. That is, the failure occurrence rate can be corresponded to an index for finding an interval in which most of defects occur. Here, the number of defective chambers (precision) can correspond to a value obtained by dividing the defective number in the section by the total number of defective parts. The number of good product rooms can correspond to a value obtained by dividing the number of good products in the section by the total number of good products. Also, the occurrence of defects can correspond to the value obtained by dividing the defective accuracy by the good product accuracy. That is, the defect occurrence rate is a value obtained by dividing the defective number in the section by the total defective number and denoting a second value (good product ratio) obtained by dividing the number of good articles in the relevant section by the total number of good articles . According to this calculation method, it is possible to accurately and efficiently find the section in which the ratio of defects exceeds the average.

For example, in a manufacturing process with 10 defects and 100 good defects, a section with a defective ratio of 0.1 or more can be effectively found. More specifically, for example, as shown in FIG. 9, if the number of defects is 5 and the number of defective products is 50 in the section A (the leftmost to the second and the first of the six bars), the number of defective rooms or the percentage defective Is 1/2 and the good product ratio is 1/2, the defect occurrence rate in the corresponding section A is one. Also, if there is one defect in section B (fourth and third) and the number of defective parts is 30, the defective ratio is 1/10 and the good part ratio is 30/100, so that the defect occurrence rate in the corresponding section B is 1/3 do. And, if there are 4 defects in section C (sixth and fifth) and 20 defective parts, the defective ratio is 4/10 and the good part ratio is 20/100, so that the defective occurrence rate in the corresponding section C becomes 2. As described above, according to the present embodiment, it is possible to effectively find a section of a plurality of processes in which the ratio of defects relative to the good among the sub-sections in a specific process or a specific process is larger than a predetermined value.

In addition, the univariate analysis process in the method according to the present embodiment can grasp the bad association relationship in addition to the analysis of the influence of the sub-interval or the variable interval. A bad association may correspond to a representation of an index to find the association of the interval IF (Then). The numerical expression may include poor support, bad confidence, poor lift, and the like.

Here, the defective support ratio can correspond to a value obtained by dividing the defective number in the section by the total number of data. The reliability confidence rate may correspond to a value obtained by dividing the number of defective sections in the section by the total number of data and denominator a second value obtained by dividing the number of data sections in the section by the total number of data. That is, the defective confidence rate may correspond to a value obtained by dividing a defective number in a section by the number of data in the section. The defective lift ratio may correspond to a value obtained by dividing the defective number in the section by the number of data in the section and a second value obtained by dividing the total defective number by the total number of data. That is, the defective lift ratio can be used to find a defective occurrence frequency of a specific section higher than a total defective occurrence frequency.

For example, in a manufacturing process in which defects are 10 and good defects are 100 defects, it is possible to effectively find a section in which a defect occurrence frequency in a specific section is higher than a total defect occurrence occurrence frequency. More specifically, for example, as shown in FIG. 9, if there are five defects in section A and 50 defective parts, the defective support ratio is 5/110 and the defective confidence ratio is 5/55, The bad lift ratio is 1. Also, if there is one defect in section B and 30 defective parts, the defective support ratio is 1/110 and the defective confidence factor is 1/31, so that the defective lift ratio in the corresponding section B is 11/31. If there are 4 defects in section C and 20 defective parts, the defective approval ratio is 4/110 and the defective confidence rate is 5/24, so that the defective lift ratio in the corresponding section C becomes 55/25.

On the other hand, the univariate analysis described above can calculate the index of influence of the variable section to determine how many good products or defects in the section are mixed. Such a value may be referred to as entropy. The entropy can be expressed by the following equation (1).

Entropy calculates the impurity of good or bad in the whole section and can be proportional to the ratio of good / bad in the generally used section.

For example, in a manufacturing process in which defects are 10 and good products are 100, entropy can be expressed as shown in FIG. More specifically, if there are 5 defects in section A and 50 good parts, the entropy becomes 0.4394. Also, if there is one defect in section B and 30 good parts, the entropy becomes 0.2055. And if there are 4 defects in section C and 20 good parts, the entropy becomes 0.6500.

FIGS. 11 and 12 are views for explaining a process of deriving and visualizing a failure cause factor among the processes of the process failure cause analysis method of FIG.

The process failure cause analysis method according to the present embodiment can derive and visualize numerical values of univariate analysis results obtained through the above-described series of processes in the form of a table as shown in FIG. The result of univariate analysis of the table type can be filtered and sorted, so it is very effective in eliminating meaningless result by adjusting the minimum occurrence rate (defective number).

Also, as shown in FIG. 12, when each variable name is clicked in the univariate analysis result table, a distribution graph of the good and / or failure of the variable can be automatically generated and displayed on the screen.

According to the embodiments described above, it is possible to effectively identify and visualize causes of abnormalities in process quality by utilizing the process failure analysis technique provided in this specification in connection with smart manufacturing technology in the main industries of the industry such as semiconductor, automobile, and steel have.

On the other hand, the method for analyzing the cause of the process failure may be implemented by a computing device including a control unit, a storage unit, and an interface. The control unit may include a processor or a microprocessor, and the storage unit may include a memory, and the interface may include a communication subsystem for supporting connection and management with the network and the input / output device, an input / output interface, and the like. The input / output device may include a display device.

The control unit may perform a series of procedures of the method for analyzing the cause of the process failure by executing a software module or a program stored in the storage unit.

The control unit may include at least one central processing unit (CPU) or a core. The central processing unit or core includes a register for storing instructions to be processed, an arithmetic logical unit (ALU) for performing comparison, determination, and operation, and an arithmetic logic unit A control unit for controlling the control unit, and an internal bus for connecting the control unit and the control unit. The central processing unit or core may be implemented as a system on chip (SOC) in which a micro control unit (MCU) and a peripheral device (integrated circuit for external expansion device) are disposed together, but the present invention is not limited thereto.

In addition, the control unit may include, but is not limited to, one or more data processors, image processors, or CODECs. The control unit may include a peripheral device interface and a memory interface. The peripheral device interface connects the control unit to the input / output device or other peripheral device, and the memory interface can connect the control unit and the memory.

The storage may store one or more software modules for implementing the processes of the process failure cause analysis method. The software module may include modules for performing the respective steps of Fig.

The storage unit may be a nonvolatile random access memory (NVRAM), a semiconductor memory such as a dynamic random access memory (DRAM) as a typical volatile memory, a hard disk drive (HDD), an optical storage device, a flash memory Lt; / RTI > The storage unit may store an operating system, a program, and a command set for a control unit, a computing device, and the like in addition to a software module for performing a method of analyzing the cause of the process failure.

Meanwhile, in the above-described embodiments, the processes of the cause failure analysis method may be stored in a computer-readable medium (recording medium) in the form of software for implementing a series of functions in a predetermined computing device, And may be downloaded through a network from a remote computing device to perform the corresponding function. The computer-readable medium includes, but is not limited to, a device for storing software readable by a computing device, and may include a storage device or a cloud system of a plurality of computer devices connected via a network.

The computer readable medium may further include a program command, a data file, a data structure, and the like that are directly or indirectly associated with the method for analyzing the cause failure of a process, singly or in combination. Programs recorded on a computer-readable medium may include those specifically designed and constructed for the present invention or those known and available to those skilled in the computer software arts.

The computer-readable medium may also include a hardware device specifically configured to store and execute program instructions, such as a ROM, a RAM, a flash memory, and the like. The program instructions may include machine language code such as those generated by a compiler, as well as high-level language code that may be executed by a computer using an interpreter or the like. The hardware device may be configured to operate by at least one software module to perform the process failure cause analysis method of the present embodiment, and vice versa.

The interface may include a device for transmitting and receiving signals and data with a plurality of sensors in a manufacturing process and a means for controlling or managing the same or a component thereof. The interface may include an input port for connection with an intranet or the Internet, a mobile communication network, a satellite network, etc., a wired or wireless communication line, and the like. Such an interface may correspond to at least some of the functional units or components of the input device, the data collection unit, or the data acquisition unit among the components of the computing device that implement the process defect cause analysis method. That is, such an interface may comprise one or more wired and / or wireless communication subsystems that support one or more communication protocols as communication means or communication devices.

The wired communication subsystem may be connected to a wireless communication subsystem and may be a public switched telephone network (PSTN), an asymmetric digital subscriber line (ADSL) or a very high-data rate digital subscriber line (VDSL) network, a PSTN Emulation Service (PESN) , An internet protocol (IP) multimedia subsystem (IMS), and the like. The wireless communication subsystem may include a radio frequency (RF) receiver for a wireless network connection, an RF transmitter, an RF transceiver, an optical (e.g., infrared) receiver, an optical transmitter, an optical transceiver, or a combination thereof.

A wireless network basically refers to a local area network such as an intranet or Wi-Fi (Wireless Fidelity), but is not limited thereto. In this embodiment, the interface may be implemented to support various wireless networks. The wireless network may be a wireless communication system such as a Global System for Mobile Communications (GSM), Enhanced Data GSM Environment (EDGE), Code Division Multiple Access (CDMA), W-Code Division Multiple Access (W-CDMA), LET- OFDMA (Orthogonal Frequency Division Multiple Access), WiMax, Bluetooth, and the like.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. I will understand. Therefore, the technical scope of the present invention should be defined by the claims.

Claims (7)

A computer program for controlling a computer system having a storage unit for storing a software module and a control unit for executing the software module, the method comprising: As a result,
Acquiring first data including quality inspection result data measured in at least one sensor installed in each of a plurality of processes and second data including process data;
Pre-processing individual data of each of the first data and the second data;
Integrating the preprocessed individual data to generate analysis target data;
Analyzing the cause of the process failure based on a univariate analysis expressing information affecting the final quality by numerical value for each individual variable; And
And deriving a failure cause index according to an analysis result based on the univariate analysis,
The step of merging the individual data in the step of generating the analysis target data may include extracting all the data in the bad data set and extracting only a part of the data in the good data set to generate the analysis target data of the integrated data set , Method of analysis of cause of process failure. The method according to claim 1,
Wherein the step of preprocessing the individual data comprises a variable generation process, a numeric variable conversion process, a meaningful variable generation process, or a combination thereof. The method of claim 2,
Wherein the variable generating step includes converting the matrix-form data into column data on an individual product basis. The method of claim 2,
Wherein the numeric variable conversion process sets the nominal variable as a new item in the individual data including the nominal variable and processes the value of the new item for the corresponding product of the nominal variable as a binary. The method of claim 2,
The meaningful variable generating process includes:
The rate of occurrence of defects in each process is obtained based on the average, maximum value or minimum value of the sensed values in the selected process,
Wherein the amount of data of interest selected for use as analysis data is greater than or equal to 50% of the total analysis data for the selected process. The method according to claim 1,
Wherein the step of generating the analysis target data comprises the steps of: dividing individual variables in one of the plurality of processes into a plurality of sections for a product manufactured through a plurality of processes, And an index based on the number of defective or non-defective products or defective products. The method according to claim 1,
And outputting information for visualizing the cause of quality defect based on the defect cause index after the deriving step.

Images