Data analysis system, data analysis device, data analysis method, and computer program
Patent Information
- Application Number
- JP2024507780
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Filing Date
- 2024-09-03
- Publication Date
- 2024-11-26
AI Technical Summary
Existing data analysis systems struggle to identify the cause of defects across multiple organizations in complex supply chains, where data sharing and anonymization are necessary to trace issues upstream without compromising sensitive information.
A data analysis system comprising a first data conversion device that anonymizes data from a first product, an analysis device that evaluates the degree of influence between anonymized and non-anonymized data from subsequent products, and an optimization device that optimizes manufacturing parameters based on selected influential elements, allowing for defect identification across multiple manufacturers without exposing raw data.
Enables effective identification of defect causes across multiple manufacturers, optimizing manufacturing processes by selecting and anonymizing relevant data elements, thus providing feedback without sharing raw data, enhancing supply chain efficiency and product quality.
Abstract
Description
Data analysis system, data analysis device, data analysis method, and computer program
[0001] The present disclosure relates to a data analysis system, a data analysis device, a data analysis method, and a computer program.
[0002] When companies conduct business, they may form inter-company networks. A typical example of such an inter-company network is the supply chain.
[0003] A supply chain is a network made up of multiple organizations such as suppliers, assemblers, manufacturers, makers, and buyers, in which procurement and shipping from each organization are continuously linked, progressing from the procurement of raw materials to the sale of the final product.
[0004] In a supply chain, identifiers are assigned to distributed parts, and when a defect occurs in a product downstream in the supply chain, the ID of the part that caused the defect is provided to the upstream parts manufacturer. The parts manufacturer then compares the ID with data it holds and attempts to determine the cause of the defect.
[0005] Japanese Patent Application Laid-Open No. 2007-213506
[0006] The present disclosure provides a data analysis system, a data analysis device, a data analysis method, and a computer program that can identify the cause of defects that occur due to multiple causes across multiple organizations.
[0007] A data analysis system according to one embodiment of the present disclosure includes a first data conversion device that anonymizes first data related to a first product manufactured by a first manufacturing system and outputs the anonymized data as first anonymized data, and a data analysis device that selects elements of the first anonymized data that affect second data related to a second product manufactured using the first product by evaluating a degree of influence between the first anonymized data and the second data.
[0008] According to the present disclosure, it is possible to identify the cause of defects that occur due to multiple causes across multiple manufacturers.
[0009] FIG. 1 is an explanatory diagram illustrating an example of the configuration of a data analysis system according to a first embodiment. FIG. 2 is a block diagram illustrating the configuration of an analysis device. FIG. 3 is an explanatory diagram illustrating an example of creating concealed data. FIG. 4 is an explanatory diagram illustrating element selection by an analysis device. FIG. 5 is a schematic diagram illustrating an example of display in an analysis device. FIG. 6 is a schematic diagram illustrating an example of display in a first data conversion device. FIG. 7 is a schematic diagram illustrating an example of display in a first data conversion device. FIG. 8 is an explanatory diagram illustrating an optimization technique by an optimization device. FIG. 9 is an explanatory diagram illustrating an optimization technique by an optimization device. FIG. 10 is a flowchart illustrating the procedure of processing executed in a data analysis system according to a first embodiment. FIG. 11 is an explanatory diagram illustrating an example of the configuration of a data analysis system according to a second embodiment. FIG. 12 is an explanatory diagram illustrating an example of the configuration of a data analysis system according to a third embodiment. FIG. 13 is an explanatory diagram illustrating an example of the configuration of a data analysis system according to a fourth embodiment.
[0010] (First Embodiment) FIG. 1 is an explanatory diagram illustrating an example of the configuration of a data analysis system according to a first embodiment. The data analysis system according to the first embodiment is a system for realizing data distribution, integration, analysis, and feedback across a supply chain. The example of FIG. 1 illustrates, among multiple organizations constituting the supply chain, a first manufacturer M1 that procures parts and materials from upstream to manufacture a first product, and a second manufacturer M2 that procures the first product manufactured by the first manufacturer M1 to manufacture a second product. Other organizations, such as other manufacturers, suppliers, assemblers, manufacturers, and buyers, may be located upstream and downstream of the first manufacturer M1, as well as upstream and downstream of the second manufacturer M1.
[0011] In the following, an application example of the present invention will be described assuming that the first manufacturer M1 is a manufacturer that procures parts from upstream and manufactures semiconductor manufacturing equipment, and the second manufacturer M2 is a manufacturer that manufactures semiconductor devices by introducing semiconductor manufacturing equipment manufactured by the first manufacturer M1.
[0012] The first manufacturer M1 includes a first manufacturing system 100, a first database 120, a first data conversion device 140, and a first optimization device 160.
[0013] The first manufacturing system 100 is a system for manufacturing a first product (semiconductor manufacturing equipment). The first product manufactured by the first manufacturing system 100 includes equipment such as an exposure apparatus, a resist processing apparatus, an etching apparatus, a cleaning / drying apparatus, a film forming apparatus, an ion implantation apparatus, and an inspection apparatus. The manufacturing process includes steps such as material procurement, storage, processing, assembly, and inspection.
[0014] The first database 120 stores data related to the first product manufactured by the first manufacturing system 100. Here, the data related to the first product is data set or measured during the manufacturing process of the first product. This data may include, for example, identification information related to the device that outputs the data, such as a serial number assigned to each product or a device model number indicating the type of device. The data may also include real number information representing conditions in the manufacturing process, such as temperature and pressure. Furthermore, the data may also include real number information output from the inspection process as manufacturing results, such as film thickness, shape, and characteristic values.
[0015] The first data conversion device 140 has a function of anonymizing data and a function of decrypting the anonymized data. For example, when providing data stored in the first database 120 to an external party, the first data conversion device 140 can read the target data from the first database 120, anonymize the read data, and provide it to the external party. The anonymization method will be described in detail later. The data is provided to, for example, the analysis device 500. Furthermore, when the first data conversion device 140 obtains anonymized data from an external party, it decrypts the obtained data and sends the decrypted data to, for example, the first optimization device 160.
[0016] The first optimization device 160 refers to the data stored in the first database 120 and the data input through the first data conversion device 140 to optimize the manufacturing parameters in the first manufacturing system 100 .
[0017] The second manufacturer M2 includes a second manufacturing system 200 and a second database 220 .
[0018] The second manufacturing system 200 is a system for manufacturing a second product (semiconductor device), and includes semiconductor manufacturing equipment introduced from the first manufacturing system 100. The second products manufactured by the second manufacturing system 200 are semiconductor devices such as diodes, transistors, light-emitting devices, light-receiving devices, ICs, microcomputers, CPUs, and memories. The manufacturing process includes an exposure process, a resist process, an etching process, a cleaning and drying process, a film formation process, an ion implantation process, and an inspection process.
[0019] Second database 220 stores data related to the second product manufactured by second manufacturing system 200. Here, the data related to the second product is data set and data measured in the manufacturing process of the second product. This data includes, for example, serial numbers assigned to individual products, product model numbers indicating the product type, and measurement values such as current, voltage, temperature, pressure, and film thickness measured in an inspection process or the like. In the first embodiment, the data stored in second database 220 is provided to analysis device 500 without being anonymized.
[0020] The data analysis system according to this embodiment includes an analysis device 500 for acquiring data from manufacturers and the like that make up a supply chain and analyzing the data across the supply chain. To perform the analysis, the analysis device 500 acquires data (first anonymized data) that is anonymized by a first data conversion device 140 and relates to a first product manufactured in a first manufacturing system 100. The analysis device 500 also acquires data (second data) that is not anonymized and that is provided by a second manufacturing system 200 and relates to a second product manufactured in the second manufacturing system 200. The analysis device 500 evaluates the degree of influence between the acquired first anonymized data and the second data, thereby selecting elements of the first anonymized data that affect the second data, and outputs the selection result.
[0021] 2 is a block diagram showing the configuration of analysis device 500. Analysis device 500 is a dedicated or general-purpose computer such as a server computer, and includes control unit 501, storage unit 502, communication unit 503, operation unit 504, display unit 505, etc.
[0022] The control unit 501 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The ROM included in the control unit 501 stores control programs and the like that control the operation of each of the above hardware components. The CPU in the control unit 501 reads and executes the control programs stored in the ROM and various computer programs stored in the storage unit 502, and controls the operation of each of the hardware components, thereby causing the entire device to function as the data analysis device of the present disclosure. The RAM included in the control unit 501 temporarily stores data used during execution of calculations.
[0023] In the embodiment, the control unit 501 is configured to include a CPU, a ROM, and a RAM, but the configuration of the control unit 501 is not limited to the above. The control unit 501 may be, for example, one or more control circuits or arithmetic circuits including a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), a DSP (Digital Signal Processor), a quantum processor, volatile or non-volatile memory, etc. The control unit 501 may also have functions such as a clock that outputs date and time information, a timer that measures the elapsed time from when a measurement start instruction is given until when a measurement end instruction is given, and a counter that counts numbers.
[0024] The storage unit 502 includes a storage device such as a hard disk drive (HDD), a solid state drive (SSD), an electronically erasable programmable read-only memory (EEPROM), etc. The storage unit 502 stores various computer programs executed by the control unit 501 and various data used by the control unit 501.
[0025] The computer programs stored in the storage unit 502 include an analysis program PG for analyzing data acquired from a first manufacturer M1, a second manufacturer M2, .... The analysis program PG may be a single computer program or may be composed of multiple computer programs. The analysis program PG may also partially use an existing library.
[0026] Computer programs such as the analysis program PG stored in the storage unit 502 are provided by a non-transitory storage medium RM on which the computer programs are readably recorded. The storage medium RM is a portable memory such as a CD-ROM, USB memory, a Secure Digital (SD) card, a microSD card, or a CompactFlash (registered trademark). The control unit 501 reads various computer programs from the storage medium RM using a reading device (not shown) and stores the read computer programs in the storage unit 502. The computer programs stored in the storage unit 502 may also be provided via communication. In this case, the control unit 501 acquires the computer programs via the communication unit 503 and stores the acquired computer programs in the storage unit 502.
[0027] The communication unit 503 includes a communication interface for transmitting and receiving various types of data. The communication interface included in the communication unit 503 is, for example, a communication interface conforming to the LAN communication standards used in Wi-Fi (registered trademark) and Ethernet (registered trademark). When data to be transmitted is input from the control unit 501, the communication unit 503 transmits the data to be transmitted to a specified destination. Furthermore, when the communication unit 503 receives data transmitted from an external device, it outputs the received data to the control unit 501.
[0028] The operation unit 504 includes operation devices such as a touch panel, a keyboard, and switches, and receives various operations and data input by the user. The control unit 501 performs appropriate control based on various pieces of operation information provided by the operation unit 504, and stores the input data in the storage unit 502 as necessary.
[0029] The display unit 505 includes a display device such as a liquid crystal display or an organic EL (Electro-Luminescence) display, and displays information to be notified to the user in response to an instruction from the control unit 501 .
[0030] In this embodiment, analysis device 500 may be a single computer, or may be a computer system configured with multiple computers and peripheral devices, etc. Furthermore, analysis device 500 may be a virtual machine whose entity is virtualized, or may be a cloud.
[0031] The anonymized data to be analyzed will be described below. FIG. 3 is an explanatory diagram illustrating an example of creating anonymized data. As described above, the first database 120 stores data related to a first product manufactured in the first manufacturing system 100. For example, if the first product is a film forming apparatus, the data stored in the first database 120 includes data such as a serial number assigned to each product (film forming apparatus), an apparatus model number, and temperature, pressure, and film thickness measured in an inspection process. The first database 120 shown as an example in FIG. 3 has five columns (elements) including a serial number, a temperature, a pressure, an apparatus model number, and a film thickness, and shows a state in which values obtained for each product are stored in each column.
[0032] The first data conversion device 140 reads target data from the first database 120 and conceals the read data. Specifically, the first data conversion device 140 conceals the data using at least one of masking column names, swapping columns, standardizing values, and dimensionality compression.
[0033] The example in Figure 3 shows the state after anonymization by changing the original column names of "S / N", "Temperature", "Pressure", "Device Model Number", and "Film Thickness" to "Param0", "Param2", "Param1", "Param4", and "Param3", respectively, in order to mask these column names.
[0034] 3 shows a state in which data has been concealed by swapping the columns of "temperature" and "pressure" and swapping the columns of "device model number" and "film thickness." Note that the first data conversion device 140 preferably stores the correspondence between column names before and after the mask and the correspondence between column names before and after the swapping.
[0035] 3 shows a state in which the data has been concealed by standardizing the values of the "temperature," "pressure," and "film thickness" columns. Specifically, the first data conversion device 140 calculates the mean value and variance for each column and standardizes each value based on the calculated mean value and variance. It is preferable that the first data conversion device 140 store the mean value and variance values used for standardization.
[0036] Furthermore, data commonly used by each organization may be encoded using a method common to the organizations. For example, among the columns shown in FIG. 3, a common value is used for the product serial number (S / N) between the first manufacturer M1 and the second manufacturer M2. Therefore, the first data conversion device 140 may use a predetermined hash function to replace the product serial number with a different character string. Any function common to the organizations may be used for the conversion.
[0037] Furthermore, for data in which the types of values that appear are fixed, predetermined numbers or character strings may be assigned. For example, among the columns shown in FIG. 3, only three types of values, P001 to P003, are used for the device model number. Therefore, the first data conversion device 140 can assign the numbers "0," "1," and "2" to P001 to P003, respectively, in that order. Alternatively, known character strings such as "A," "B," and "C" may be assigned.
[0038] 3 is merely an example. The first data conversion device 140 may perform appropriate anonymization processing in accordance with the attributes of the columns prepared in the first database 120. For example, the first data conversion device 140 may perform dimensional compression of the data stored in the first database 120 using an algorithm such as principal component analysis or independent component analysis.
[0039] The analysis device 500 acquires the first anonymized data that has been anonymized by the first data conversion device 140 and the second data related to the second product obtained in the second manufacturing system 200, and selects elements (columns) of the first anonymized data that affect the second data by evaluating the degree of influence of both.
[0040] FIG. 4 is an explanatory diagram illustrating element selection by the analysis device 500. In FIG. 4, the first anonymized data generated by the first data conversion device 140 is denoted as X, and the second data generated by the second manufacturing system 200 is denoted as Y. The first anonymized data X is assumed to be data consisting of n elements (columns), where n is an integer equal to or greater than 2. The first anonymized data shown as an example in FIG. 3 has five columns, so the case of n=5 is illustrated.
[0041] The analysis device 500 acquires the first anonymized data X from the first data conversion device 140 and acquires the second data Y from the second manufacturing system 200. The analysis device 500 evaluates the degree of influence between the first anonymized data X and the second data Y, thereby selecting elements of the first anonymized data X that affect the second data Y. Specifically, the analysis device 500 sets a model function f (i.e., Y = f(X)) for predicting the objective variable from the explanatory variables, with the first anonymized data X as an explanatory variable and the second data Y as an objective variable, and selects a subset x of X necessary for predicting Y based on the set model function f. In other words, the analysis device 500 excludes columns of X that are irrelevant to the prediction of Y.
[0042] In this embodiment, in order to exclude elements that are irrelevant to the prediction of Y, the degree of influence of each element included in X on Y is evaluated. For example, the analysis device 500 can use a statistical method (filtering method) to find an element x of X that has a strong influence on Y.
[0043] The analysis device 500 can also use a machine learning model (embedding method) to find an element x of X that has a strong influence on Y. For example, when Y = f(X) is a linear model, if the scale of the explanatory variable X is appropriately adjusted, the regression coefficient (absolute value) will directly represent the degree of influence. In particular, when it is desired to remove irrelevant elements, a method such as Lasso (L1 regularization) can be used. Furthermore, when Y = f(x) is a nonlinear model, the degree of influence can be evaluated using an index such as the Gini coefficient in a random forest. The evaluation index for the degree of influence is not limited to the Gini coefficient, and evaluation indexes such as Permutation Feature Importance, LIME, and SHAP may also be used.
[0044] Furthermore, analysis device 500 may use a wrapper method in which feature values are input and output, and elements with a high degree of influence are selected while checking the prediction accuracy. In the wrapper method, machine learning models are repeatedly created to evaluate the degree of influence and select feature values, which takes a long calculation time but allows the degree of influence to be evaluated with high accuracy.
[0045] If the analysis device 500 determines, based on the evaluation of the degree of influence, that a specific element (referred to as element p) among the elements of the first anonymized data X has a particular influence on the second data Y, the analysis device 500 selects the element p and notifies the first data conversion device 140 of information regarding the selected element p.
[0046] The number of elements p selected by the analysis device 500 is not limited to 1, and may be multiple. That is, the analysis device 500 may exclude elements of the first anonymized data X that are irrelevant to the prediction of the second data Y, and select a subset x of the first data X that is necessary for the prediction of the second data Y.
[0047] 4, the element selection process between two organizations has been described, but the same element selection process can be applied to any number of organizations connected in series. For example, in a supply chain with n organizations connected in series, let X1 be the anonymized data generated in the most upstream organization, X2 be the anonymized data generated in the downstream organization, ..., and X3 be the anonymized data generated in the upstream organization of the most downstream organization. n-1 , and data generated by the most downstream organization is denoted as Y. The analysis device 500 acquires these data from each organization. The analysis device 500 analyzes the anonymized data X n-1 By evaluating the degree of influence between data Y and anonymized data X, the degree of influence of data Y is calculated. n-1 Element p of n-1 Next, the analysis device 500 selects the anonymized data X n-2 and anonymized data X n-1 By evaluating the degree of influence between n-1 Element p of n-1 Anonymous data X that affects n-2 Element p of n-2 In this way, analysis device 500 can select elements that affect each data by analyzing sequentially from the most downstream to the most downstream.
[0048] In addition, the element of the anonymized data that has the greatest effect on data Y is anonymized data X. n-3 Element p of n-3 Generally speaking, the analysis device 500 calculates Y=f(x⊂X⊂, x⊂X⊂, ..., x n-1 ⊂X n-1 ) and perform factor analysis. n-1 are concatenated and regarded as one explanatory variable X, and the subset required for predicting the response variable Y can be selected.
[0049] The first data conversion device 140 stores a correspondence relationship between elements of data before anonymization (first data) and data after anonymization (first anonymized data). Therefore, when the first data conversion device 140 is notified of information about element p from the analysis device 500, the first data conversion device 140 can identify an element of the first data that affects the second data based on the information about element p.
[0050] During the element selection process, the analysis device 500 may display information on the influence degree of each element on the display unit 505. FIG. 5 is a schematic diagram showing an example of a display in the analysis device 500. Assume that during the element selection process, a model Y = f(x1, x2, ..., xn) = a1 * x1 + a2 * x2 + ... + an * xn is obtained. In this case, the larger the absolute value of the coefficients a1, a2, ..., an, the greater the influence of the element on data Y. The sign of the coefficient changes depending on whether the influence on data Y is positive or negative. The analysis device 500 can display the influence degree of each element in a diagram by, for example, plotting the influence degree on the horizontal axis and the element on the vertical axis as a bar graph showing the influence degree of each element. FIG. 5 shows an example in which elements are sorted in descending order of influence degree. Note that, because each element (column name) is anonymized in the analysis device 500, it is impossible to recognize the actual column names (column names before anonymization).
[0051] Similarly, the analysis device 500 may notify the first data conversion device 140 of the calculated influence degree information for each element and cause the first data conversion device 140 to display the influence degree information for each element. FIGS. 6A and 6B are schematic diagrams showing display examples in the first data conversion device 140. Similar to the analysis device 500, the first data conversion device 140 can display the influence degree of each element by plotting a bar graph with the influence degree on the horizontal axis and the element on the vertical axis, thereby indicating the influence degree of each element. FIG. 6A shows an example in which elements are sorted in descending order of influence degree. The first data conversion device 140 can present the pre-anonymization column name as the name of the element by converting the element by referring to the correspondence between the data before anonymization (first data) and the data after anonymization (first anonymized data). Therefore, in the example of FIG. 6A, it can be seen that the "temperature" of "process 2" has the greatest impact on the quality of the second product manufactured by the second manufacturing system 200.
[0052] Furthermore, the first data conversion device 140 may display a graph showing the relationship between the control range of each element and the variation range of the performance Y. In the example of FIG. 6B , a graph shows the relationship between the control range of "temperature" in "process 2" and the variation range of the performance. The variation range of the performance is a value specified by, for example, the second manufacturing system 200 and notified via the analysis device 500. By viewing this graph, the user can understand how much the variation range of the performance of the second product can be controlled by changing the control range of "temperature" in "process 2".
[0053] The first optimization device 160 refers to the information on the elements selected by the first data conversion device 140 and optimizes the parameters in the manufacturing process of the first product.
[0054] 7A and 7B are explanatory diagrams illustrating the optimization technique by the first optimization device 160. Assume that the element xp identified by the first data conversion device 140 is the temperature T of a certain process of a certain part. In this case, the first optimization device 160 optimizes the control value of the temperature condition in that process.
[0055] For example, suppose that the analysis device 500 obtains a prediction formula Y = a*xp + ... for the element xp, and the first data conversion device 140 obtains a corresponding formula T = b*xp + c for the temperature T. Here, Y is data related to the performance of the second product, and a, b, and c are coefficients optimized when creating the model. In this case, the first optimization device 160 determines how to set the temperature T based on the coefficients a, b, and c, as well as the range of Y that is desired to be achieved.
[0056] For example, the first optimization device 160 may previously have managed the temperature within the temperature range of T0±ΔT, but may now manage it within T0±ΔT' to control the variation in the quality of the second product manufactured by the second manufacturing system 200 (see FIG. 7A). Also, as shown in FIG. 7B, the first optimization device 160 may previously have managed the temperature within the temperature range of T0±ΔT, but may now manage it within T1±ΔT to control the offset in the quality of the second product manufactured by the second manufacturing system 200 (see FIG. 7B).
[0057] If the first optimization device 160 cannot change the control value of the selected element xp, it may change the control value of the element with the second greatest influence. Furthermore, the first optimization device 160 may change the control values of multiple elements.
[0058] FIG. 8 is a flowchart illustrating the processing steps executed in the data analysis system according to the first embodiment. When a first product is manufactured in the first manufacturing system 100, data related to the first product (first data) is stored in the first database 120. The first data conversion device 140 acquires the first data from the first database 120 (step S101) and anonymizes the acquired first data (step S102). The anonymization of the first data uses at least one of column name masking, column swapping, value standardization, and dimensionality reduction. The first data conversion device 140 stores the element correspondence between the data before anonymization (first data) and the data after anonymization (first anonymized data) in an internal memory (not shown). The first data conversion device 140 transmits the first anonymized data obtained by anonymizing the first data to the analysis device 500 (step S103).
[0059] The second manufacturing system 200 procures the first product manufactured by the first manufacturing system and manufactures the second product. When manufacturing the second product, the second manufacturing system 200 generates data (second data) related to the second product (step S105), stores the data in the second database 220, and transmits the generated second data to the analysis device 500 (step S106).
[0060] The analysis device 500 receives the first anonymized data transmitted from the first data conversion device 140 (step S104), and receives the second data transmitted from the second manufacturing system 200 (step S107).
[0061] The analysis device 500 evaluates the degree of influence between the acquired first anonymized data and the second data (step S108), and selects elements of the first anonymized data that affect the second data (step S109). The analysis device 500 can evaluate the degree of influence of each element of the first anonymized data on the second data using the above-mentioned filter method, embedding method, wrapper method, etc., and select elements with a high degree of influence.
[0062] The analysis device 500 notifies the first data conversion device 140 of information about the selected element (step S110). The information notified to the first data conversion device 140 includes the column name of the element (column) selected by the analysis device 500. Note that the analysis device 500 does not need to understand the meaning of the selected element, so it is sufficient to notify the first data conversion device 140 of, for example, the column name in a masked state.
[0063] The first data conversion device 140 receives element information notified from the analysis device 500 (step S111) and decodes the received element information (step S112). The first data conversion device 140 refers to the correspondence stored in memory and identifies the first data element corresponding to the element selected by the analysis device 500 (step S113). The first data conversion device 140 notifies the first optimization device 160 of the information on the identified first data element (step S114).
[0064] The first optimization device 160 optimizes the control value of the corresponding element based on the information notified from the first data conversion device 140 (step S115). For example, the first optimization device 160 may set the range of the control value of the element from the range of the performance Y that is desired to be achieved for the second product.
[0065] As described above, the data analysis system according to the first embodiment evaluates the degree of influence between the first anonymized data obtained by the first manufacturer M1 and the second data obtained by the second manufacturer M2, and selects elements of the first anonymized data that affect the second data. Therefore, even when multiple manufacturers are involved, it is possible to identify upstream data factors that affect downstream products and provide feedback to each manufacturer without providing raw data (data before anonymization).
[0066] Second Embodiment In a second embodiment, a configuration in which a second manufacturer M2 includes a data conversion device will be described.
[0067] 9 is an explanatory diagram showing an example of the configuration of a data analysis system according to embodiment 2. The first manufacturer M1 is the same as that shown in embodiment 1, and includes a first manufacturing system 100, a first database 120, a first data conversion device 140, and a first optimization device 160.
[0068] The second manufacturer M2 includes a second manufacturing system 200, a second database 220, a second data conversion device 240, and a second optimization device 260.
[0069] Like the first data conversion device 140, the second data conversion device 240 has a function of anonymizing data and a function of decrypting the anonymized data. The anonymization method used is the same as the anonymization method used in the first data conversion device 140. That is, the second data conversion device 240 may generate second anonymized data by performing at least one of masking column names, swapping columns, standardizing values, and dimensionality reduction on data related to the second product (second data) read from the second database 220. The second anonymized data obtained by the second data conversion device 240 is provided to, for example, the analysis device 500.
[0070] The second optimization device 260 refers to the data stored in the second database 220 and the data input through the second data conversion device 240 to optimize the manufacturing parameters in the second manufacturing system 200 .
[0071] The analysis device 500 in the second embodiment acquires first anonymized data from the first data conversion device 140 and acquires second anonymized data from the second data conversion device 240. The analysis device 500 evaluates the degree of influence of the acquired anonymized data, thereby selecting elements (columns) of the first anonymized data that affect the second anonymized data. The method of selecting elements is the same as in the first embodiment.
[0072] The analysis device 500 notifies the first data conversion device 140 of information on the selected elements. The first data conversion device 140 decodes the notified information to identify elements of the first data corresponding to the elements selected by the analysis device 500. The first optimization device 160 refers to the information on the elements identified by the first data conversion device 140 and optimizes parameters in the manufacturing process of the first product.
[0073] As described above, the data analysis system according to the second embodiment evaluates the degree of influence between the first anonymized data obtained by the first manufacturer M1 and the second anonymized data obtained by the second manufacturer M2, and selects elements of the first anonymized data that affect the second anonymized data. Therefore, even in cases involving multiple manufacturers, it is possible to identify upstream data factors that affect downstream products and provide feedback to each manufacturer without providing raw data (data before anonymization).
[0074] (Embodiment 3) A supply chain in embodiment 3 includes a third manufacturer M3 arranged in parallel with a first manufacturer M1. A second manufacturer M2 procures a first product manufactured by the first manufacturer M1 and a third product manufactured by the third manufacturer M3, and manufactures a second product.
[0075] 10 is an explanatory diagram showing an example of the configuration of a data analysis system according to embodiment 3. The configuration of the first manufacturer M1 is the same as that of embodiment 1. That is, a first product manufactured by a first manufacturing system 100 is supplied to the first manufacturer, and first anonymized data obtained by anonymizing first data related to the first product is provided to an analysis device 500.
[0076] The second manufacturer M2 procures a first product manufactured by the first manufacturer M1 and a third product manufactured by a third manufacturer M3, and is provided with a second manufacturing system 200 that manufactures a second product. The second manufacturing system 200 stores data relating to the second product (second data) in a second database 220, and provides the second data to the analysis device 500.
[0077] The third manufacturer M3 is located upstream of the second manufacturer M2 and includes a third manufacturing system 300, a third database 320, a third data conversion device 340, and a third optimization device 360.
[0078] The third manufacturing system 300 is a system for manufacturing a third product (e.g., semiconductor manufacturing equipment). The third database 320 stores data related to the third product manufactured by the third manufacturing system 300. The third data conversion device 340 has a function of anonymizing data and a function of decrypting the anonymized data. The anonymization method is the same as the anonymization method used by the first data conversion device 140. That is, the third data conversion device 340 generates third anonymized data by performing at least one of column name masking, column swapping, value standardization, and dimensionality reduction on data related to the third product (third data) read from the third database 320. The third anonymized data obtained by the third data conversion device 340 is provided to, for example, the analysis device 500. The third optimization device 360 refers to the data stored in the third database 320 and the data input through the third data conversion device 340 , and optimizes the manufacturing parameters in the third manufacturing system 300 .
[0079] The analysis device 500 acquires the first anonymized data provided from the first data conversion device 140, the third anonymized data provided from the third data conversion device 340, and the second data provided from the second manufacturing system 200. The analysis device 500 evaluates the degree of influence between the acquired data including the first anonymized data and the third anonymized data and the second data, and thereby selects elements of the first anonymized data and the third anonymized data that affect the second data. The method of selecting the elements is the same as in the first embodiment.
[0080] For example, if the first anonymized data is data (x1, x2, ..., xn) consisting of n elements and the third anonymized data is data (z1, z2, ..., zm) consisting of m elements, the analysis device 500 evaluates Y = f(x1, x2, ..., xn, z1, z2, ..., zm) and selects elements that have a high degree of influence on fluctuations in Y. In this case, the analysis device 500 may select elements from only one of the first anonymized data and the third anonymized data, or may select elements from both the first anonymized data and the third anonymized data.
[0081] The analysis device 500 notifies the first data conversion device 140 and the third data conversion device 340 of information on the selected elements. The first data conversion device 140 decodes the notified information to identify elements of the first data corresponding to the elements selected by the analysis device 500. The first optimization device 160 refers to the information on the elements identified by the first data conversion device 140 and optimizes parameters in the manufacturing process of the first product.
[0082] Similarly, third data conversion device 340 decodes the notified information to identify the third data element corresponding to the element selected by analysis device 500. Third optimization device 360 refers to the information on the element identified by third data conversion device 340 to optimize parameters in the manufacturing process of the third product.
[0083] As described above, the data analysis system according to the third embodiment evaluates the degree of influence between data including the first anonymized data obtained by the first manufacturer M1 and the third anonymized data obtained by the third manufacturer M3, and the second data obtained by the second manufacturer M2, and selects elements of the first anonymized data and the third anonymized data that affect the second data. Thus, even when data spans multiple manufacturers, it is possible to identify upstream data factors that affect downstream products and provide feedback to each manufacturer without providing raw data (data before anonymization).
[0084] (Fourth Embodiment) A supply chain according to a fourth embodiment includes a fourth manufacturer M4 arranged in parallel with a second manufacturer M2. The second manufacturer M2 procures a first product manufactured by a first manufacturer M1 to manufacture a second product, and the fourth manufacturer M4 procures a first product manufactured by the first manufacturer M1 to manufacture a fourth product.
[0085] 11 is an explanatory diagram showing an example of the configuration of a data analysis system according to embodiment 4. The configuration of the first manufacturer M1 is the same as that of embodiment 1. In embodiment 4, a first product manufactured by a first manufacturing system 100 is supplied to the first manufacturer M1 and a fourth manufacturer M4. In addition, first anonymized data obtained by anonymizing first data related to the first product is provided to an analysis device 500.
[0086] The second manufacturer M2 has the same configuration as in embodiment 2. It includes a second manufacturing system 200 that procures the first product manufactured by the first manufacturer M1 and manufactures the second product. Data (second data) related to the second product manufactured by the second manufacturing system 200 is stored in a second database 220, anonymized, and provided to the analysis device 500 as second anonymized data.
[0087] The fourth manufacturer M4 is arranged downstream of the first manufacturer M1 in parallel with the second manufacturer M2. The fourth manufacturer M4 includes a fourth manufacturing system 400, a fourth database 420, a fourth data conversion device 440, and a fourth optimization device 460.
[0088] The fourth manufacturing system 400 is a system for manufacturing a fourth product (e.g., a semiconductor device). The fourth database 420 stores data related to the fourth product manufactured by the fourth manufacturing system 400. The fourth data conversion device 440 has a function of anonymizing data and a function of decrypting the anonymized data. The anonymization method is the same as the anonymization method used by the first data conversion device 140. That is, the fourth data conversion device 440 generates fourth anonymized data by performing at least one of column name masking, column swapping, value standardization, and dimensionality reduction on data related to the fourth product (fourth data) read from the fourth database 420. The fourth anonymized data obtained by the fourth data conversion device 440 is provided to, for example, the analysis device 500. The fourth optimization device 460 refers to the data stored in the fourth database 420 and the data input through the fourth data conversion device 440 , and optimizes the manufacturing parameters in the fourth manufacturing system 400 .
[0089] The analysis device 500 acquires the first anonymized data provided from the first data conversion device 140, the second anonymized data provided from the second data conversion device 240, and the fourth anonymized data provided from the fourth data conversion device 440. The analysis device 500 evaluates the degree of influence between the acquired first anonymized data and data including the second anonymized data and the fourth anonymized data, thereby selecting elements of the first anonymized data that affect the second anonymized data and the fourth anonymized data. The method of selecting elements is the same as in the first embodiment. For example, assume that the first anonymized data includes, in addition to n elements x1, x2, ..., xn, i elements v1, v2, ..., vi that affect only the second manufacturing system 200, and j elements w1, w2, ..., wj that affect only the fourth manufacturing system 400. In this case, the analysis device 500 identifies elements that have a high degree of influence on fluctuations in the second anonymized data Y1 using Y1=f(x1, x2, ..., xn, v1, v2, ..., vi) for the second anonymized data Y1. Also, the analysis device 500 identifies elements that have a high degree of influence on fluctuations in the fourth anonymized data Y2 using Y2=g(x1, x2, ..., xn, w1, w2, ..., wj) for the fourth anonymized data Y2. Here, f and g are independent functions, and elements can be selected independently for Y1 and Y2.
[0090] The analysis device 500 notifies the first data conversion device 140 of information on the selected elements. The first data conversion device 140 decodes the notified information to identify elements of the first data corresponding to the elements selected by the analysis device 500. The first optimization device 160 refers to the information on the elements identified by the first data conversion device 140 and optimizes parameters in the manufacturing process of the first product.
[0091] As described above, the data analysis system according to the fourth embodiment evaluates the degree of influence between the first anonymized data obtained by the first manufacturer M1, the second anonymized data obtained by the second manufacturer M2, and the fourth anonymized data obtained by the fourth manufacturer, and selects elements of the first anonymized data that affect the second anonymized data or the fourth anonymized data. Thus, even in cases involving multiple manufacturers, it is possible to identify upstream data factors that affect downstream products and provide feedback to each manufacturer without providing raw data (data before anonymization).
[0092] The embodiments disclosed herein are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims.
[0093] M1 First manufacturer 100 First manufacturing system 120 First database 140 First data conversion device 160 First optimization device M2 Second manufacturer 200 Second manufacturing system 220 Second database 500 Analysis device 501 Control unit 502 Storage unit 503 Communication unit 504 Operation unit 505 Display unit
Claims
1. a first data conversion device that conceals first data related to a first product manufactured by a first manufacturing system and outputs the concealed data; a data analysis device that selects an element of the first anonymized data that influences the second data by evaluating a degree of influence between the first anonymized data and second data related to a second product manufactured using the first product; A data analysis system comprising:
2. The second data is data that has been concealed by a second data conversion device. The data analysis system according to claim 1 .
3. a third data conversion device that conceals third data related to a third product manufactured by the third manufacturing system and outputs the third concealed data; The data analysis device evaluates a degree of influence of the data including the first anonymized data and the third anonymized data on the second data, and selects elements of the first anonymized data and the third anonymized data that affect the second data. The data analysis system according to claim 1 .
4. a fourth data conversion device that conceals fourth data related to a fourth product manufactured by a fourth manufacturing system using the first product and outputs the fourth concealed data; The data analysis device selects an element of the first anonymized data that affects the second data and the fourth anonymized data by evaluating a degree of influence between the first anonymized data and the second data and a degree of influence between the first anonymized data and the fourth anonymized data. The data analysis system according to claim 1 .
5. a first optimization device that adjusts a parameter in a manufacturing process of the first product based on an element of the first data corresponding to the selected element of the first concealed data; The data analysis system according to claim 1 .
6. and a third optimization device that adjusts parameters in a manufacturing process of the third product based on the element of the third data corresponding to the selected element of the third concealed data. The data analysis system according to claim 3 .
7. The data for each product includes a number of elements and a number of values for each element; The data concealment includes at least one of masking element names, swapping elements, standardizing values of each element, and dimensionality reduction. The data analysis system according to claim 1 .
8. a first acquisition unit that acquires first anonymized data obtained by anonymizing first data related to a first product manufactured by a first manufacturing system; a second acquisition unit that acquires second data related to a second product manufactured by a second manufacturing system using the first product; a selection unit that selects an element of the first anonymized data that affects the second data by evaluating a degree of influence between the first anonymized data and the second data; An output section that outputs the selection result; A data analysis device comprising:
9. the first data includes a plurality of elements and a plurality of values for each element; The anonymization of the first data includes at least one of masking element names, swapping elements, standardizing values of each element, and dimensionality reduction; The selection unit calculates a degree of influence of each element included in the first anonymized data on the second data, and selects an element of the first anonymized data according to the calculated degree of influence. The data analysis device according to claim 8.
10. acquiring first anonymized data obtained by anonymizing first data related to a first product manufactured by a first manufacturing system; acquiring second data relating to a second product manufactured by a second manufacturing system using the first product; selecting an element of the first anonymized data that influences the second data by evaluating a degree of influence between the first anonymized data and the second data; Output the selection results A data analysis method in which processing is carried out by a computer.
11. acquiring first anonymized data obtained by anonymizing first data related to a first product manufactured by a first manufacturing system; acquiring second data relating to a second product manufactured by a second manufacturing system using the first product; selecting an element of the first anonymized data that influences the second data by evaluating a degree of influence between the first anonymized data and the second data; Output the selection results A computer program for causing a computer to execute a process.
