Factor analysis support device and factor analysis support method

Abstract

To provide a factor analysis support device and a factor analysis support method which support analysis of defect factors in a power semiconductor module inspection.SOLUTION: A factor analysis support device is configured to: extract, as interlocked manufacture process information, manufacture process information in which switching of a production lots of products is interlocked with switching of individual identification numbers of the manufacture process information; calculate an average defect rate of a production lot number corresponding to the individual identification number of the manufacture process information associated with a production lot number extracted in the manufacture process information extraction step and a preceding individual identification number in a production order of individual identification numbers of the manufacture process information associated with the production lot number; calculate, as an influence score, a degree of influences upon the defect rate caused by switching the individual identification numbers of the manufacture process information; and align, on a table, the interlocked manufacture process information, the average defect rate and the influence score, and rearrange them in a descending order according to a magnitude of the influence score, and display them.SELECTED DRAWING: Figure 13

Description

[Technical field]

[0001] The present invention relates to a factor analysis support device and a factor analysis support method in power semiconductor module inspection. [Background technology]

[0002] Power semiconductor modules are widely used in various fields, such as electric vehicles, electric railways, and renewable energy power generation systems, and there is a demand for them to be even more durable and reliable.

[0003] Meanwhile, insulation breakdown due to partial discharge is known to be one of the causes of failure of power semiconductor modules. Partial discharge is a weak discharge (corona discharge) that occurs at defective parts of the insulator, and accelerates the deterioration of the insulator. If partial discharge occurs at the driving voltage, it shortens the product life and leads to sudden failure and destruction. Therefore, it is very important to use partial discharge testing to detect latent defects that cannot be found by other withstand voltage tests, and it is extremely effective in evaluating the insulation reliability of power semiconductor modules.

[0004] Incidentally, there are several possible causes of partial discharge in power semiconductor modules, including voids in the resin and cracks on the surface of the substrate. In addition, partial discharge inspections are carried out after the packaging process, in which sealing is performed using gel or adhesive. For this reason, in order to clarify the root cause of partial discharge, it is necessary to dismantle the power semiconductor module, which takes a considerable amount of time and cost to identify the cause. Therefore, there is a demand for more efficient cause analysis.

[0005] To improve the efficiency of factor analysis, it is necessary to grasp the change points in the materials, workers, and equipment used when a defect occurs. The present invention automatically extracts the material lot that has changed in conjunction with the change when the defect rate or the discharge amount of partial discharge inspection increases, and visually presents it to the analyst, thereby speeding up the analyst's defect factor analysis. Here, the defect rate represents the ratio of the number of defects to the number of inspections in a given unit that represents a production unit, such as a production lot.

[0006] Known prior art related to defect factor analysis is, for example, that disclosed in Patent Document 1. Patent Document 1 discloses a manufacturing defect factor analysis support device that accumulates manufacturing history information extracted from a plurality of work processes before a product is manufactured through the plurality of work processes, and displays a matrix of a specified information item from the accumulated manufacturing history information with a specified product's manufacturing lot number or individual identification number, thereby enabling the analysis of manufacturing defect factors. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2010-182284 A Summary of the Invention [Problem to be solved by the invention]

[0008] In the above conventional technology, the manufacturing lot or individual identification number of the product arranged in the manufacturing order by product unit and the process information related to the manufacturing such as the part lot are displayed in a matrix table, so that the analyst can visually confirm the process information linked to the increase in the defect rate and the discharge amount of the partial discharge inspection. However, since the power semiconductor module is a module in which multiple parts are mounted, and the partial discharge inspection is performed after packaging, multiple discharge factors are considered. For this reason, when the conventional technology described in the above Patent Document 1 is used for the cause analysis of the discharge in the partial discharge inspection of the power semiconductor module, there may be multiple candidate factors of the process information linked to the increase in the defect rate and the discharge amount of the partial discharge inspection, making it difficult to visually identify them. Here, if the contribution rate of each candidate factor to the defect can be confirmed when multiple candidate factors are extracted, the in-depth analysis can be carried out in descending order of the contribution rate. For example, efficiency can be improved by dismantling only a part corresponding to the candidate factor with a high contribution rate, rather than dismantling the entire module. However, in the conventional technology described in Patent Document 1, there is no description equivalent to the contribution rate, so the order in which in-depth analysis of candidate factors is performed depends on the analyst, and there are cases where it becomes necessary to completely dismantle a module, making it difficult to speed up factor analysis.

[0009] The present invention has been made in view of the above, and has an object to provide a factor analysis support device and a factor analysis support method for supporting the analysis of defect factors in power semiconductor module inspection. [Means for solving the problem]

[0010] The present application includes a plurality of means for solving the above-mentioned problems. One example of such a means is a cause analysis support device for supporting analysis of causes of manufacturing defects in a power semiconductor module, the cause analysis support device comprising: a manufacturing process information storage means for linking and collecting information on an individual identification number of a product, a manufacturing lot number, and a manufacturing sequence for manufacturing process information, which is information on components, processing equipment, workers, etc., related to the manufacture of a power semiconductor module; and when a manufacturing lot number of a product whose defect rate exceeds a predetermined threshold is specified, the manufacturing process information storage means stores, in the manufacturing process information storage means, the manufacturing lot number immediately preceding the specified manufacturing lot number in the manufacturing sequence, the individual identification number of the manufacturing process information linked to the manufacturing lot number, and the individual identification number immediately preceding the individual identification number of the manufacturing process information linked to the manufacturing lot number in the manufacturing sequence. The method comprises a manufacturing process information extraction means for extracting information from the collected information, and an analysis result display means for extracting manufacturing process information in which a change in a product manufacturing lot and a change in an individual identification number of the manufacturing process information are linked as linked manufacturing process information, calculating an average defect rate of the manufacturing lot number corresponding to the individual identification number of the manufacturing process information linked to the manufacturing lot number extracted by the manufacturing process information extraction means and the individual identification number immediately preceding in the manufacturing order of the individual identification number of the manufacturing process information linked to the manufacturing lot number, calculating the degree of influence on the defect rate due to the change in the individual identification number of the manufacturing process information as an impact score, organizing the linked manufacturing process information, the average defect rate, and the impact score in a table, and displaying the table in descending order of the magnitude of the impact score. Effect of the Invention

[0011] According to the present invention, it is possible to appropriately assist in the analysis of defect factors in the inspection of a power semiconductor module. [Brief description of the drawings]

[0012] [Figure 1] FIG. 2 is a functional block diagram illustrating an example of a functional configuration of the cause analysis support device. [Diagram 2] FIG. 2 is a perspective view showing a state of the power semiconductor module during partial discharge inspection. [Diagram 3] FIG. 11 is a diagram showing a typical example of partial discharge inspection conditions. [Figure 4] FIG. 13 is a diagram showing an example of a discharge pattern during partial discharge inspection. [Diagram 5] 13 is a diagram illustrating an example of a configuration of a data table stored in a manufacturing performance data storage unit; FIG. [Figure 6] 13 is a diagram showing an example of the configuration of a data table stored in a manufacturing process data storage unit; FIG. [Figure 7] 13 is a diagram showing an example of the configuration of a data table combined by a data combination execution unit; FIG. [Figure 8] FIG. 13 is a diagram showing an example of the configuration of a data table in which data is grouped by lot number. [Figure 9] FIG. 1 is a chart showing the relationship between manufacturing lot numbers and defect rates. [Figure 10] FIG. 13 is a diagram illustrating a method for calculating an influence score. [Figure 11] FIG. 13 is a diagram illustrating a method for calculating an influence score. [Figure 12] 11 is a chart showing the relationship between the manufacturing lot number and the defect rate and the impact score, respectively. [Figure 13] 13 is a flowchart showing the contents of a cause analysis support process. [Figure 14] FIG. 13 is a diagram illustrating an example of a GUI of the cause analysis support device. [Figure 15] 1 is a vertical cross-sectional view showing an example of the structure of a power semiconductor module. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0014] This embodiment proposes a factor analysis support device and a factor analysis support method for supporting the factor analysis of defects in the inspection of a power semiconductor module. In this embodiment, a partial discharge inspection is exemplified as an inspection of a power semiconductor module, and a factor analysis support device and a factor analysis support method are proposed, which identify individual identification numbers of process information such as component lots linked to a manufacturing lot in which the charge amount or defect rate has increased, extract process information in which the individual identification numbers of the process information have changed in conjunction with an abnormality in the defect rate as factor candidates, calculate an impact score indicating the possibility of contribution to the defect for each factor candidate, and display information that can be compared between the factor candidates.

[0015] First, partial discharge in a power semiconductor module will be described.

[0016] FIG. 15 is a vertical cross-sectional view showing an example of the structure of a power semiconductor module.

[0017] 15, a power semiconductor module 31 is mounted with a semiconductor chip 81 such as an IGBT (Insulated Gate Bipolar Transistor) or a diode. The back surface of the semiconductor chip 81 is joined to metal wiring 83 on the surface of an insulating substrate 84 by solder 82. The surface of the insulating substrate 84 may contain cracks 95 such as splits or scratches at the time of delivery. A bonding wire 93 is connected to the surface of the semiconductor chip 81, and the semiconductor chip 81 is connected to other metal wiring 83 on the surface of the insulating substrate 84 via the bonding wire 93.

[0018] Metal wiring 85 on the back surface of insulating substrate 84 is joined to heat dissipating base plate 87 by solder 86. A case 88 and a lid 89 are adhered and fixed to heat dissipating base plate 87 by adhesive 90 so as to cover semiconductor chip 81, metal wiring 83, insulating substrate 84, and metal wiring 85. An auxiliary terminal 92 is arranged on case 88.

[0019] The gaps between the heat dissipating base plate 87, the case 88, and the lid 89, and between the semiconductor chip 81, the metal wiring 83, the insulating substrate 84, and the metal wiring 85 are filled with soft resin 91 in order to insulate the inside of the power semiconductor module 31. Furthermore, hard resin 94 is disposed at the ends of the metal wiring 83 on the surface of the insulating substrate 84 where the electric field is particularly likely to concentrate, in order to strengthen the insulation.

[0020] Heat dissipating base plate 87 is at, for example, ground potential, and when a drive voltage is applied to power semiconductor module 31, an electric field is generated in the direction indicated by arrows EF1 to EF5. At this time, for example, if voids (air bubbles) exist in hard resin 94, soft resin 91, or adhesive 90, partial discharges P1 to P3 may occur. Furthermore, if a void or gap exists between metal wiring 83 and insulating substrate 84, partial discharge P4 may occur. Furthermore, even in cracks 95 on the surface of insulating substrate 84, partial discharge P5 may occur due to concentration of the electric field at the cracks.

[0021] That is, since there are multiple factors that cause partial discharge across components in the power semiconductor module 31, it may be difficult to isolate the cause of a defect in an electrical characteristic test performed after the packaging process, such as a partial discharge test.

[0022] In this embodiment, therefore, a case where a defect cause analysis in a partial discharge inspection of a power semiconductor module is supported will be described as an example. Note that, although a partial discharge inspection will be described as an example of the inspection of a power semiconductor in this embodiment, the inspection is not limited thereto, and can be applied to other inspections performed after packaging.

[0023] FIG. 1 is a functional block diagram showing an example of a functional configuration of a factor analysis support device according to the present embodiment.

[0024] As shown in FIG. 1, the power semiconductor module factor analysis support device mainly comprises a discharge test facility 11, a computer server 12, and a communication unit 13 connecting the discharge test facility 11 and the computer server 12.

[0025] The discharge test equipment 11 includes a current measurement unit 1101 and a voltage application unit 1102. During partial discharge inspection, an electronic device 1103 to be subjected to a discharge test is connected to the current measurement unit 1101 and the voltage application unit 1102, a test voltage is applied from the voltage application unit 1102 to the electronic device 1103, and the current measurement unit 1101 measures the value of a current (amount of discharged charge) flowing from the electronic device 1103, thereby measuring the amount of charge caused by partial discharge. The electronic device 1103 corresponds to the power semiconductor module 31.

[0026] The computer server 12 comprises a memory unit 14, a calculation unit 15, and a user interface unit 16. Although Fig. 1 illustrates an example in which the memory unit 14, the calculation unit 15, and the user interface unit 16 are installed in the same place as functional units of the computer server 12, they may be installed in different places and interconnected by a network.

[0027] The storage unit 14 includes a partial discharge inspection program 1401, a result display program 1402, a data table combining program 1403, an interlocking lot extraction program 1404, a factor analysis start definition program 1405, an influence score calculation program 1406, an influence score anomaly detection program 1407, a manufacturing performance data storage unit 1408, and a manufacturing process data storage unit 1409. Note that, in FIG. 1, the manufacturing performance data storage unit 1408 is included in the computer server 12, but the manufacturing performance data storage unit 1408 may be installed in a location separate from the computer server 12 and the manufacturing performance data storage unit 1408 and the computer server 12 may be interconnected via a network, or a plurality of manufacturing performance data storage units may be prepared, such as the first and second manufacturing performance data storage units, and the first manufacturing performance data storage unit may be installed in the computer server 12 while the second manufacturing performance data storage unit may be distributed and installed in other locations that can be connected to the network.

[0028] The calculation unit 15 includes an inspection execution unit 1501, a signal processing unit 1502, a data combination execution unit 1503, a linked lot extraction execution unit 1504, an influence score calculation execution unit 1505, an influence score anomaly detection execution unit 1506, and a result display execution unit 1507.

[0029] The user interface unit 16 (GUI: Graphical User Interface) includes an input unit 1601 , a computer communication unit 1602 , and a display unit 1603 .

[0030] Next, the basic principle of the cause analysis support process in this embodiment will be described.

[0031] Fig. 2 is a perspective view showing a state of a power semiconductor module during a partial discharge inspection, Fig. 3 is a diagram showing a typical example of partial discharge inspection conditions, and Fig. 4 is a diagram showing an example of a discharge pattern during a partial discharge inspection.

[0032] As shown in FIG. 2, during partial discharge inspection, an AC voltage is applied from a voltage application unit 1102 to a power semiconductor module 31 under test, and a current measurement unit 1101 measures a current value (amount of discharged charge).

[0033] As a test condition for the partial discharge inspection, for example, a test voltage is applied to the power semiconductor module 31 using a test voltage pattern in which the voltage pattern changes with time as shown in Fig. 3. In Fig. 3, the vertical axis indicates the applied voltage AV, and the horizontal axis indicates the test time TT.

[0034] In the example of Figure 3, the voltage is increased from the start of the test until it reaches a specified voltage V1, at which point the applied voltage V1 is maintained for a fixed period of time (approximately 60 seconds in Figure 3). After that, the voltage is decreased until it reaches a specified voltage V2, at which point the applied voltage V2 is maintained for a fixed period of time (approximately 25 seconds in Figure 3). Furthermore, the specified voltage V3 is maintained for a fixed period of time (approximately 5 seconds in Figure 3), and the voltage is then decreased.

[0035] 3 illustrates an example in which the voltages V2 and V3 are set to the same voltage condition, but the voltages V2 and V3 may be set to different voltage values. Also, the test conditions (test voltage pattern) differ for each type of power semiconductor module 31.

[0036] Fig. 4 is a diagram showing the test results using the test voltage pattern shown in Fig. 3. In Fig. 4, the vertical axis indicates the charge amount CA, and the horizontal axis indicates the test time TT.

[0037] In Fig. 4, the application period (about 60 s) of voltage V1 of the test voltage pattern in Fig. 3 corresponds to period t1. The application period (about 25 s) of voltage V2 corresponds to period t2, and the application period (about 5 s) of voltage V3 corresponds to period t3. The dashed-dotted line TH is a threshold for determining whether a product is normal or defective in the partial discharge inspection. If the charge amount CA is equal to or greater than the threshold TH, which indicates a threshold abnormality, the product is determined to be defective, and if the charge amount CA is less than the threshold TH, the product is determined to be non-defective. The value of the threshold TH may be different for each product type.

[0038] As shown in Fig. 4, during a period t1 when a relatively high voltage V1 is applied, the charge CA gradually increases while fluctuating up and down. On the other hand, during a period t2 when a voltage V2 lower than the voltage V1 is applied, the charge CA gradually decreases while fluctuating up and down.

[0039] In this way, by applying a test voltage to the power semiconductor module 31 using a test voltage pattern (see FIG. 3) whose voltage pattern changes over time, it is possible to obtain time-series data (discharge pattern) of the discharge charge amount during partial discharge inspection, and whether the module is normal or defective is determined based on whether the threshold value TH is exceeded.

[0040] In this way, the results of the partial discharge inspection for each module are stored in a manufacturing performance data storage unit 1408 of the storage unit 14 of the computer server 12 via the communication unit 13. In addition, process information relating to the manufacturing of the power semiconductor module, such as various parts used in the power semiconductor module shown in Fig. 15, is stored as so-called "manufacturing process information" (such as complaints) in a manufacturing process data storage unit 1409 of the storage unit 14.

[0041] Fig. 5 is a diagram showing an example of the configuration of a data table stored in the manufacturing performance data storage unit, and Fig. 6 is a diagram showing an example of the configuration of a data table stored in the manufacturing process data storage unit.

[0042] 5, the data table stored in the manufacturing performance data storage unit 1408 records information on each item of the "product name" of the inspection target, the lot number ("production lot number"), the "sublot number", the "inspection process", the "characteristic value" obtained by the inspection, and the judgment result ("judgment") that is judged as normal (OK) or defective (NG) based on the characteristic value, in association with the inspection information linked to the individual identifier of the electronic device 1103 that is subjected to partial discharge inspection by executing the partial discharge inspection program 1401. Note that, although the inspection process is limited to the partial discharge inspection in this embodiment, information on other inspection processes may be recorded. Also, the product name may be a product other than PRODUCT_A.

[0043] The lot number is a unit for managing individual modules collectively, and the sublot number is an identifier for identifying individual modules in a lot. In other words, the combination of the lot number and the sublot number corresponds to the serial number (identification number) of an individual module. For example, in the example shown in FIG. 5, the sublot numbers (01 to 100) are collectively represented as lot number L001. In the case of partial discharge inspection, the characteristic value indicates a representative value (minimum value, etc.) of the discharge charge amount as shown in FIG. 4, but other items may be used as long as the value corresponds to the inspection process. In the case of partial discharge inspection, the judgment value is recorded as a normal (OK) judgment if the discharge charge amount as shown in FIG. 4 is less than TH, and as an abnormal (NG) judgment if it is TH or more.

[0044] FIG. 6 is a diagram showing an example of the configuration of a data table stored in the manufacturing process data storage unit.

[0045] As shown in Fig. 6, in the data table stored in the manufacturing process data storage unit 1409, information related to the manufacturing process such as materials, workers, and devices when manufacturing a power semiconductor module is recorded in association with the lot number of the product. In Fig. 6, for example, information on each item of "product name", lot number ("product"), "manufacturing process", "material name" corresponding to the manufacturing process, "material lot number" of the material used, and "worker" is recorded. Note that, although the lot number is used as the product identifier in Fig. 6, an identifier combining a lot number and a sublot number, or another identifier associated with the product identifier in the manufacturing performance data storage unit 1408 may be used.

[0046] The manufacturing process is a process identifier linked to a process for manufacturing a product, the component name is the name of the component used in the manufacturing process, the component lot number is an identifier for the component of the linked component name, and the worker is an identifier for the worker who performed the work in the manufacturing process. The manufacturing performance data storage unit 1408 may store information related to the manufacturing process other than the components and workers (e.g., the equipment used). In the following description, such information related to the manufacturing process may be collectively referred to as manufacturing process data or manufacturing process information.

[0047] FIG. 7 is a diagram showing an example of the configuration of a data table combined by the data combination execution unit.

[0048] The data combination execution unit 1503 combines the data table of the manufacturing performance data storage unit 1408 shown in FIG. 5 with the data table of the manufacturing process data storage unit 1409 shown in FIG. 6 using a combination of lot numbers and sublot numbers as an identification key (primary key), and generates a data table in which the data are rearranged in the manufacturing order. In FIG. 7, the first to fifth columns ("Product Name" to "Decision") show a part of the data table stored in the manufacturing performance data storage unit 1408, and the sixth and subsequent columns ("001_PARTS_A" to) show a part of the data table stored in the manufacturing process data storage unit 1409. For example, in the column 001_PARTS_A, the part lot number of the part PARTS_A of the manufacturing process 001 is recorded in association with the lot number. In FIG. 7, only the part lot number is described as an example, but process information such as workers and devices can also be recorded in the same manner. 7, it is possible to process changes in lot numbers and sublot numbers in association with changes in characteristic values ​​and changes in component lot numbers. Therefore, for example, when a characteristic value changes from a normal value to an abnormal value, it is possible to extract process information such as the component lot number that is linked to the change.

[0049] Fig. 8 is a diagram showing an example of the configuration of a data table in which data is grouped by lot number for the data table shown in Fig. 7. In Fig. 8, information on the lot number ("production lot number"), the "characteristic value" of the discharge test, the "defective rate", and the component lot number linked to the manufacturing process and the component used (component lot number for each component used) is recorded. In Fig. 7, as an example, only 001_PARTS_A, 001_PARTS_B, and 001_PARTS_C are listed as columns showing the manufacturing process and the component lot number linked to the component used, but similar columns related to other processes, components, and process information may be added.

[0050] The defect rate is expressed as a percentage of the number of modules determined to be defective among the number of inspected modules for multiple sublot numbers corresponding to one lot number in Fig. 7. In Fig. 8, the average characteristic value obtained by partial discharge inspection of modules corresponding to the lot number is recorded as the characteristic value, but other representative values ​​may be recorded. By using a data table such as that shown in Fig. 8, it is possible to extract the manufacturing lot number with an increased defect rate as shown in the seventh row of Fig. 8 and the component lot number corresponding to the manufacturing lot number immediately before that, and therefore it is possible to extract the part name linked to the increase in the defect rate.

[0051] For example, in the case of FIG. 8, the defect rate increases between the 6th row and the 7th row. When a factor analysis focusing on this change is to be performed, 001_PARTS_A changes from LOT111 to LOT112, and 001_PARTS_B changes from LOT114 to LOT115. On the other hand, 001_PARTS_C does not change and remains at LOT191. Therefore, 001_PARTS_A and 001_PARTS_B are extracted as factor terms that have changed in conjunction with the defect rate. This extraction process is executed by the result display execution unit 1507 in FIG. 1, and the factor candidates are presented to the user through the display unit 1603 by the result display execution unit 1507 in FIG. 1. An example of presentation to the user will be described in detail later in FIG. 14. In this way, the user can know the change points of the process information, and the factor analysis can be made more efficient.

[0052] Next, a method for calculating an impact score that quantitatively represents the possibility of contribution of each of the candidate factors extracted from the data table of FIG. 8 to an increase in the defect rate will be described.

[0053] Fig. 9 is a chart showing the relationship between the manufacturing lot number and the defect rate. In Fig. 9, the horizontal axis shows the manufacturing lot and the vertical axis shows the defect rate for the data table shown in Fig. 8. In Fig. 9, the manufacturing lot numbers on the horizontal axis are arranged in the order of manufacture.

[0054] The threshold value 501 shown in FIG. 9 represents a threshold value for the defect rate. When the threshold value 501 is exceeded, first, the interlocking lot extraction execution unit 1504 in FIG. 1 is executed, and the process of the data combination execution unit 1503 shown in FIG. 4, the influence score calculation execution unit 1505, the influence score abnormality detection execution unit 1506, and the result display execution unit 1507 are executed. The threshold value 501 is stored in the definition program 1405 for starting factor analysis in FIG. 1. In addition to the threshold value 501 for the defect rate, for example, a threshold value for a characteristic value may be recorded in the definition program 1405 for starting factor analysis as a trigger condition for starting the factor analysis. In FIG. 9, the manufacturing lot number L007 exceeds the threshold value. In the following description, a method for calculating the influence score will be described for the factor analysis of the abnormal increase in the defect rate in the manufacturing lot number L007.

[0055] Fig. 10 is a diagram for explaining a method of calculating an influence score by taking 001_PARTS_A as an example, and Fig. 11 is a diagram for explaining a method of calculating an influence score by taking 001_PARTS_B as an example.

[0056] In Fig. 10, the defect rate is shown by dividing it into parts by the part lot numbers (LOT110, LOT111, LOT112) of 001_PARTS_A in comparison with Fig. 9. In this case, the influence score S is calculated based on the following (Equation 1).

[0057]

number

[0058] In the above (Equation 1), q represents the defect rate of the factor analysis target (in the case of FIG. 10, the defect rate of the manufacturing lot number L007), μ represents the average value of the defect rate corresponding to the component lot immediately before the factor analysis target (in the case of FIG. 10, the component lot is LOT111, and the manufacturing lots are L004 to L006), and σ represents the standard deviation of the defect rate corresponding to the component lot immediately before the factor analysis target. In this way, by dividing the squared value of the difference between the defect rate after the component lot change and the average defect rate before the component lot change by the standard deviation of the defect rate before the defect rate increase, it is possible to evaluate how much the defect rate after the component lot change has increased, including the variation, compared to the defect rate before the component lot change, and the effect of the defect rate increase due to the component lot change can be calculated. An impact score calculation example 502 shows the calculation process and result of the impact score of 001_PARTS_A according to the above (Equation 1).

[0059] In Fig. 11, similarly to Fig. 10, the defect rate is shown by marking the part lot numbers (LOT111, LOT112) of 001_PARTS_B in Fig. 9. Here, an influence score calculation example 503 shows the calculation process and results of the influence score of 001_PARTS_B according to the above (Equation 1).

[0060] Here, when comparing the impact score 502 for 001_PARTS_A shown in FIG. 10 with the impact score 502 for 001_PARTS_B shown in FIG. 11, the impact score 502 for 001_PARTS_B shows a higher value. Since the impact score quantitatively indicates the possibility of contribution to the increase in the defect rate, it indicates that 001_PARTS_B, which corresponds to the impact score 502 showing a higher value, is more likely to contribute to the increase in the defect rate (than 001_PARTS_A). In this way, the possibility of contribution to defects is quantitatively evaluated for the candidate factors of the manufacturing process data that have changed in conjunction with the increase in the defect rate, and the result is presented to the user. Therefore, even if there are multiple candidate factors, the user can narrow down the candidate factors to be investigated by looking at the impact score, and the user's factor analysis can be made more efficient and rational.

[0061] Next, a method of evaluating whether a certain influence score is abnormal or normal in consideration of influence scores in past material lot changes will be described.

[0062] Fig. 12 is a chart showing the relationship between the manufacturing lot number and the defect rate and the impact score. In Fig. 12, a chart 601 shows the manufacturing lot number on the horizontal axis and the defect rate on the vertical axis. Also, a chart 602 shows the manufacturing lot number on the horizontal axis and the impact score on the vertical axis.

[0063] As shown in Fig. 12, when the defect rate increases in manufacturing lot number L017, the component PARTS_X of manufacturing process 005 changes accordingly. In chart 602, the impact score is calculated not only for manufacturing lot number L017 but also each time the component lot number changes. The degree of deviation of the impact score of manufacturing lot number L017 from the impact scores other than manufacturing lot number L017 can be calculated by the following (Equation 2). By using the degree of deviation of the impact scores calculated by the following (Equation 2), it can be determined whether the impact score of manufacturing lot number L017 is the same as in the past or is significantly different.

[0064]

number

[0065] In the above (Equation 2), SL represents the impact score of L002, which is the subject of factor analysis, SM represents the average value of past impact scores, and SV represents the standard deviation of past impact scores.

[0066] The user can freely set how far back in the past the lot should be to calculate the impact score, for example, up to the time before the most recent maintenance of the manufacturing equipment was performed, or the most recent time when the manufacturing recipe for the equipment was stable.

[0067] The advantage of determining whether the impact score is abnormal or normal will be explained. If the impact score of the factor analysis target deviates significantly from the trend of past impact scores, it may be that the manufacturing lot number LOT017 is caused by a greater fluctuation than the usual material properties of the component 005_PARTS_X, and it is considered effective to analyze the difference between the manufacturing lot number LOT017 and the component lot when the impact score was stable before that, as a deep analysis. Conversely, if the impact score of the factor analysis target is about the same as the past trend, the characteristic fluctuation of the manufacturing lot number LOT017 may be about the same as the past characteristic fluctuation, suggesting that the defective rate is not robust to the characteristic fluctuation of the component 005_PARTS_X. In other words, design changes to the component 005_PARTS_X itself may be effective in reducing the defective rate.

[0068] As described above, it is determined whether the impact score of the factor analysis target is abnormal with respect to past trends, and the result is presented to the user. Abnormality is determined based on whether it exceeds the impact score threshold 603 set by the user in advance. The process of determining abnormal values ​​for the impact score is executed by the impact score anomaly detection execution unit 1506 in FIG. 1, and presented to the user via the display unit 1603 in FIG. 1.

[0069] FIG. 13 is a flowchart showing the details of the cause analysis support process.

[0070] 13, first, a manufacturing lot to be subjected to factor analysis is designated by a user through the user interface unit 16 in Fig. 1 (step S100). Note that instead of user designation, the manufacturing lot may be automatically determined by using the factor analysis start definition program 1405 in Fig. 1 as a trigger.

[0071] Next, the data join execution unit 1503 in FIG. 1 executes the data table join program 1403 to join the data table in the manufacturing performance data storage unit and the data table in the manufacturing process data storage unit (step S110: see FIGS. 5, 6, 7, and 8).

[0072] Next, the interlocking lot extraction execution unit 1504 in FIG. 1 executes the interlocking lot extraction program 1404 to extract, as a factor candidate, manufacturing process information that has changed in conjunction with the change in the manufacturing lot that is the subject of the factor analysis (step S120: see FIG. 8).

[0073] Next, the influence score calculation execution unit 1505 in FIG. 1 executes the influence score calculation program 1406, and calculates influence scores for the factor candidates extracted in step S120 based on the above (Equation 1) (step S130: see FIGS. 9, 10, and 11).

[0074] Next, the impact score anomaly detection execution unit 1506 in Figure 1 executes the impact score anomaly detection program 1407, and determines whether the impact score of the candidate factor calculated in step S130 is abnormal or normal based on the above (Equation 2) (step S140: see Figure 12).

[0075] Next, the result display execution unit 1507 of FIG. 1 executes the result display program of FIG. 1 to display on the display unit 1603 at least table information linking the candidate causes, the impact scores, and the abnormality judgment of the impact scores, and information including a chart of the defect rate sorted by manufacturing process information and arranged in order of manufacturing lot, and terminates the process (Step S150: see FIG. 14 below).

[0076] FIG. 14 is a diagram illustrating an example of a GUI of the cause analysis support device.

[0077] The GUI (Graphical User Interface) of the factor analysis support device shown in FIG. 14 is output and displayed on the display unit 1603 by the result display execution unit 1507 in FIG. 1 executing the result display program in FIG. 1 during power semiconductor module testing.

[0078] 14, the GUI includes a table display button 801, an automatic mode selection button 802 for selecting the automatic mode, a manual mode selection button 803 for selecting the manual mode, a product name and manufacturing lot selection section 804, a graph display button 805, a data table section 806, and a graph section 807. A user can input data to the factor analysis support device for power semiconductor module inspection using the GUI shown in FIG.

[0079] For example, when a user wishes to use the factor analysis support device for power semiconductor module inspection using factor analysis start definition program 1405 in Fig. 1, the user presses automatic mode selection button 802. Fig. 14 shows an example in which the manual mode is selected. When the manual mode is selected, an arbitrary product name and manufacturing lot are selected in product name / manufacturing lot selection section 804, and table display button 801 is pressed. Fig. 14 shows an example in which "PRODUCT_A" is selected as the product name and "L017" is selected as the manufacturing lot number.

[0080] The data table section 806 displays at least the part name linked to the defect rate change linked to the product name and manufacturing lot selected in the product name / manufacturing lot selection section 804, the material lot number and defect rate, the material lot number of the candidate factor described above, the representative value of the defect rate when the material lot number is used, the influence score, the result of the abnormality judgment of the influence score, and a check box for selecting whether or not to display a graph. The data table section 806 is sorted so that the influence score is in descending order. In FIG. 14, the material lot is displayed as an example of manufacturing process data, but other manufacturing process data such as workers and equipment may be displayed.

[0081] Next, check any one of the check boxes in the data table section 806 and press the graph display button 805, and at least a chart of the defect rate and the impact score as shown in Fig. 12 will be displayed in the graph section 807. Note that Fig. 14 shows an example of the result when 005_PARTS_X is checked.

[0082] The operation and effects of the semiconductor device according to the present embodiment configured as above will be described.

[0083] In the above conventional technology, the manufacturing lot or individual identification number of the product arranged in the manufacturing order by product unit and the process information related to the manufacturing such as the part lot are displayed in a matrix table, so that the analyst can visually confirm the process information linked to the increase in the defect rate and the discharge amount of the partial discharge inspection. However, since the power semiconductor module is a module in which multiple parts are mounted, and the partial discharge inspection is performed after packaging, multiple discharge factors are considered. For this reason, when the conventional technology described in the above Patent Document 1 is used for the cause analysis of the discharge in the partial discharge inspection of the power semiconductor module, there may be multiple candidate factors of the process information linked to the increase in the defect rate and the discharge amount of the partial discharge inspection, making it difficult to visually identify them. Here, if the contribution rate of each candidate factor to the defect can be confirmed when multiple candidate factors are extracted, the in-depth analysis can be carried out in descending order of the contribution rate. For example, efficiency can be improved by dismantling only a part corresponding to the candidate factor with a high contribution rate, rather than dismantling the entire module. However, in the conventional technology described in Patent Document 1, there is no description equivalent to the contribution rate, so the order in which in-depth analysis of candidate factors is performed depends on the analyst, and there are cases where it becomes necessary to completely dismantle a module, making it difficult to speed up factor analysis.

[0084] In contrast, in the present embodiment, a cause analysis support device for supporting analysis of causes of manufacturing defects in a power semiconductor module includes a manufacturing process information accumulation means for collecting information on manufacturing process, which is information on members, processing devices, workers, etc., relating to the manufacture of a power semiconductor module, by linking the individual identification number of a product, a manufacturing lot number, and information on a manufacturing sequence, and a manufacturing process information extraction means for, when a manufacturing lot number of a product whose defect rate exceeds a predetermined threshold is specified, extracting from the information collected by the manufacturing process information accumulation means the immediately preceding manufacturing lot number in the manufacturing sequence of the specified manufacturing lot number, the individual identification number of the manufacturing process information linked to the manufacturing lot number, and the individual identification number immediately preceding the individual identification number of the manufacturing process information linked to the manufacturing lot number, in the manufacturing sequence, and a process for switching the manufacturing lot of the product and The present invention is configured to include an analysis result display means for extracting, as linked manufacturing process information, manufacturing process information that is linked to a switch of an individual identification number in process information, extracted by the manufacturing process information extraction means, calculating an average defect rate of a manufacturing lot number corresponding to the individual identification number of the manufacturing process information linked to the manufacturing lot number and the individual identification number that precedes the individual identification number in the manufacturing order of the individual identification number of the manufacturing process information linked to the manufacturing lot number, calculated as an impact score, and organizing the linked manufacturing process information, the average defect rate, and the impact score in a table and displaying them in descending order by the magnitude of the impact score, thereby making it possible to appropriately support analysis of defect factors in power semiconductor module inspection, and to improve the accuracy and speed of the cause analysis.

[0085] <Additional Notes> The present invention is not limited to the above-described embodiments, and includes various modifications and combinations within the scope of the gist of the present invention. The present invention is not limited to those having all the configurations described in the above-described embodiments, and includes those in which some of the configurations are deleted. The above-described configurations, functions, etc. may be realized by designing some or all of them as an integrated circuit, for example. The above-described configurations, functions, etc. may be realized by software, in which a processor interprets and executes a program that realizes each function. [Explanation of symbols]

[0086] 12...computer server, 13...communication unit, 14...storage unit, 15...arithmetic unit, 16...user interface unit, 31...power semiconductor module, 81...semiconductor chip, 82...solder, 83...metal wiring, 84...insulating substrate, 85...metal wiring, 86...solder, 87...heat dissipation base plate, 88...case, 89...lid, 90...adhesive, 91...soft resin, 92...auxiliary terminal, 93...bonding wire, 94...hard resin, 95...crack, 100...sublot number, 501...threshold, 502...impact score, 601, 602...chart, 603...threshold, 801...table display button, 802...automatic mode selection button, 803...manual mode selection button, 804...product name / production lot selection unit, 805...graph display button, 806...data table unit, 807...graph unit, 1101...current measurement unit, 1102...voltage application unit, 1103...electronic device, 1401...partial discharge inspection program, 1402...result display program, 1403...data table combination program, 1404...linked lot extraction program, 1405...definition program, 1406...influence score calculation program, 1407...abnormality detection program, 1408...manufacturing performance data storage unit, 1409...manufacturing process data storage unit, 1501...inspection execution unit, 1502...signal processing unit, 1503...data combination execution unit, 1504...linked lot extraction execution unit, 1505...influence score calculation execution unit, 1506...abnormality detection execution unit, 1507...result display execution unit, 1601...input unit, 1602...computer communication unit, 1603...display unit

Claims

1. A factor analysis support device that supports analysis of manufacturing defect factors of a power semiconductor module, A manufacturing process information storage means for collecting manufacturing process information, which is information on components, processing equipment, workers, etc., relating to the manufacturing of the power semiconductor module, by linking the individual identification number of the product, the manufacturing lot number, and information on the manufacturing sequence; When a manufacturing lot number of a product whose defect rate exceeds a predetermined threshold is specified, A manufacturing lot number immediately preceding the designated manufacturing lot number in the manufacturing order; An individual identification number of manufacturing process information linked to the manufacturing lot number; a manufacturing process information extracting means for extracting, from the information collected by the manufacturing process information storing means, an individual identification number immediately preceding the individual identification number of the manufacturing process information linked to the manufacturing lot number; Extracting, as linked manufacturing process information, manufacturing process information in which the change in the manufacturing lot of the product and the change in the individual identification number of the manufacturing process information are linked; Calculating an average defect rate of a manufacturing lot number corresponding to an individual identification number of the manufacturing process information linked to the manufacturing lot number extracted by the manufacturing process information extraction means and an individual identification number immediately preceding in the manufacturing order the individual identification number of the manufacturing process information linked to the manufacturing lot number; Calculating the degree of influence on the defect rate due to the change in the individual identification number of the manufacturing process information as an influence score; an analysis result display means for sorting and displaying the linked manufacturing process information, the average defect rate, and the influence score in a table in descending order of the magnitude of the influence score; A factor analysis support device comprising:

2. 2. The factor analysis support device according to claim 1, The impact score of the degree of impact on the defect rate due to the change in the individual identification number of the manufacturing process information is: A defect rate corresponding to a manufacturing lot number of a product when the defect rate exceeds a pre-specified threshold value; and A factor analysis support device characterized in that the defect rate is calculated based on the average value and standard deviation of the defect rate of the manufacturing lot number linked to the individual identification number that is one step before in the manufacturing order of the individual identification number of the manufacturing process information linked to the manufacturing lot number.

3. 3. The factor analysis support device according to claim 2, The impact score for the interlocking manufacturing process information; A group of individual identification numbers obtained by tracing back an arbitrary period in the manufacturing order of the individual identification numbers of the manufacturing process information linked to the manufacturing lot number is compared with a past trend of an impact score calculated every time the individual identification number is changed, A factor analysis support device, characterized in that when the impact score for the interlocking manufacturing process information deviates from a past trend, it is determined to be an abnormal value.

4. A method for supporting analysis of manufacturing defects in a power semiconductor module, comprising: a manufacturing process information accumulation process for collecting manufacturing process information, which is information on components, processing equipment, workers, etc., related to the manufacturing of the power semiconductor module, by linking the individual identification number of the product, the manufacturing lot number, and information on the manufacturing sequence; When a manufacturing lot number of a product whose defect rate exceeds a predetermined threshold is specified, A manufacturing lot number immediately preceding the designated manufacturing lot number in the manufacturing order; An individual identification number of manufacturing process information linked to the manufacturing lot number; a manufacturing process information extraction step of extracting, from the information collected in the manufacturing process information accumulation step, an individual identification number immediately preceding the individual identification number of the manufacturing process information linked to the manufacturing lot number; Extracting, as linked manufacturing process information, manufacturing process information in which the change in the manufacturing lot of the product and the change in the individual identification number of the manufacturing process information are linked; Calculating an average defect rate of a manufacturing lot number corresponding to an individual identification number of the manufacturing process information linked to the manufacturing lot number extracted in the manufacturing process information extraction step and an individual identification number immediately preceding in a manufacturing order the individual identification number of the manufacturing process information linked to the manufacturing lot number; Calculating the degree of influence on the defect rate due to the change in the individual identification number of the manufacturing process information as an influence score; an analysis result display step of sorting the linked manufacturing process information, the average defect rate, and the impact score in a table and displaying the table in descending order of the magnitude of the impact score; A factor analysis support method comprising the steps of:

Images