Method and computer program for generating explanatory variables used in product factor analysis

Abstract

To provide a technique that enables easy generation of explanatory variables for time-series data.SOLUTION: A method disclosed herein comprises (a) acquiring time-series data in product manufacturing step, (b) extracting a simplified waveform obtained by simplifying the time-series data, and (c) generating waveform symbols expressing the simplified waveform with one or more characters as explanatory variables.SELECTED DRAWING: Figure 5

Description

Technical Field

[0001] The present disclosure relates to a method for creating explanatory variables used in factor analysis of products and a computer program.

Background Art

[0002] Conventionally, in factor analysis of products, a mechanism is known for creating an analysis model from data such as manufacturing conditions and quality characteristics and specifying explanatory variables that are the causes of defects (for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, conventionally, when using the features of time-series data as explanatory variables, it has been necessary to calculate as necessary data represented by numerical values such as the average value, maximum value, minimum value, and frequency of the time-series data. For this reason, there has been a problem that optimal analysis cannot be performed unless appropriate features are extracted well, and proficiency is required for the extraction. Therefore, a technique for easily creating explanatory variables related to time-series data has been desired.

Means for Solving the Problems

[0005] According to a first aspect of the present disclosure, a method for creating explanatory variables used in factor analysis of products is provided. This method includes: (a) a step of acquiring time-series data in the manufacturing process of the product; (b) a step of extracting a simplified waveform obtained by simplifying the time-series data; and (c) a step of creating, as the explanatory variable, a waveform symbol representing the simplified waveform with one or more characters.

[0006] According to a second aspect of the present disclosure, there is provided a computer program for creating explanatory variables to be used in factor analysis of a product. This computer program causes a computer to execute: (a) a process of acquiring time-series data in the manufacturing process of the product; (b) a process of extracting a simplified waveform obtained by simplifying the time-series data; and (c) a process of creating, as the explanatory variable, a waveform symbol representing the simplified waveform with one or more characters.

Brief Description of the Drawings

[0007]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Embodiments for Carrying Out the Invention

[0008] FIG. 1 is a block diagram showing the configuration of an analyzer 100 in an embodiment. The analyzer 100 is a factor analyzer that analyzes the cause of the quality of a specific product. The analyzer 100 can be realized by, for example, a personal computer.

[0009] The analysis device 100 includes a processor 110, a memory 120, an interface circuit 130, and an input device 140 and a display device 150 connected to the interface circuit 130. The processor 110 not only has the function of executing the processes detailed below, but also has the function of displaying the data obtained by the processes and the data generated during the process on the display device 150.

[0010] The processor 110 realizes the functions of a time-series data acquisition unit 112 that acquires time-series data of a product, a waveform extraction unit 114 that extracts a simplified waveform obtained by simplifying the time-series data, a waveform symbolization unit 116 that creates a waveform symbol representing the simplified waveform, and an analysis unit 118 that performs factor analysis of the product. The functions of these units are realized by the processor 110 executing a computer program stored in the memory 120. However, the functions of each unit may be realized by a hardware circuit. The processor in this specification is a term that also includes such a hardware circuit. Also, the processor that executes the processes of each unit may be a processor included in a remote computer connected to the analysis device 100 via a network. Further, the processes of each unit may be executed by a plurality of processors. The memory 120 stores a product data mart DM and time-series data TD. These data will be described below. Note that the data included in the memory 120 may be stored in an external storage device connected to the analysis device 100 via a network.

[0011] FIG. 2 is an explanatory diagram showing an example of the product data mart DM. The product data mart DM includes a product ID of the product, a pass / fail determination result, a lot ID, and attribute data of the product. The attribute data is measurement data related to the pass / fail of the product. In this example, as the attribute data, three dimensions Da, Db, and Dc of the product are registered. These attribute data are used as explanatory variables in factor analysis. Note that some of the plurality of items included in the product data mart DM may be omitted.

[0012] FIG. 3 is an explanatory diagram showing an example of time-series data TD. The time-series data TD is measurement data of manufacturing parameters in the manufacturing process of a product. As the time-series data TD, for example, measured values such as temperature and pressure are used. In the example of FIG. 3, the temperature change during processing is measured as the time-series data TD. The time-series data TD is obtained as data per product ID when performing factor analysis for each product ID, and is obtained as data per lot ID when performing factor analysis for each lot ID.

[0013] FIG. 4 is a flowchart showing the procedure of the factor analysis process. In step S10, the user selects the product ID of the target product to be subject to factor analysis. In step S20, the time-series data acquisition unit 112 determines whether time-series data TD is available for the target product. If there is no time-series data TD, the process proceeds to step S80, and the analysis unit 118 performs factor analysis of the product using the product data mart DM as it is. On the other hand, if there is time-series data TD, the process proceeds to step S30, and the time-series data acquisition unit 112 acquires the time-series data TD from the memory 120. In step S40, the waveform extraction unit 114 extracts a simplified waveform from the time-series data TD.

[0014] FIG. 5 is an explanatory diagram showing the content of the process of step S40 for extracting the simplified waveform SWF from the time series data TD and the process of step S50 for symbolizing the simplified waveform SWF. The time series data TD generally has a shape in which a high-frequency vibration component is superimposed on a low-frequency waveform. The simplified waveform SWF is a waveform obtained by simplifying the changes in the time series data TD. The simplified waveform SWF can be extracted, for example, as the trend waveform of the time series data TD. Specifically, by using an analysis method called STL decomposition (Seasonal and Trend decomposition using Loess), the time series data TD can be decomposed into a trend component, a seasonal variation, and a residual, and the trend component can be extracted as the simplified waveform SWF. Alternatively, the simplified waveform SWF may be extracted by performing a low-pass filtering process for removing the high-frequency vibration component from the time series data TD. The low-pass filtering process can be realized, for example, by an operation of taking a moving average.

[0015] In step S50, the waveform symbolization unit 116 generates a waveform symbol of the simplified waveform SWF by executing the symbolization of the simplified waveform SWF. As shown in FIG. 5, in step S50, first, the entire section of the simplified waveform SWF is divided into a plurality of symbolization sections Ej. Then, an interval waveform symbol representing the waveform in each symbolization section Ej is determined. Further, by integrating these interval waveform symbols, the waveform symbol Cw of the simplified waveform SWF is determined. Note that "j" is an ordinal number for distinguishing the plurality of symbolization sections Ej, and in the example of FIG. 5, j = 1 to 3.

[0016] The plurality of symbolization sections Ej are divided according to a preset rule. As this rule, for example, one of the following can be adopted. <Rule DR1> The entire section of the simplified waveform SWF is divided into a plurality of symbolization sections Ej by using the timing of the extreme value as a boundary. <Rule DR2> The entire section of the simplified waveform SWF is evenly divided into n symbolization sections E1 to En. Here, n is an integer of 2 or more. <Rule DR3>Divide the simplified waveform SWF into a plurality of symbolized intervals Ej by dividing it at regular intervals from the starting point of the simplified waveform SWF. In this case, the length of the last symbolized interval is variable.

[0017] In Rule DR1, the waveform symbol Cw can be generated so as to more clearly represent the increasing and decreasing tendencies of the simplified waveform SWF. In Rules DR2 and DR3, the entire section of the simplified waveform SWF can be easily divided into a plurality of symbolized intervals Ej by simple operations. In this embodiment, Rule DR1 is used.

[0018] Each of the plurality of symbolized intervals Ej set by Rule DR1 may be a monotonically increasing interval or a monotonically decreasing interval in which the data value of the simplified waveform SWF changes monotonically. The "monotonically increasing interval" means a continuous interval in which the data value changes without decreasing. The "monotonically decreasing interval" means a continuous interval in which the data value changes without increasing.

[0019] Note that the waveform symbol Cw of the simplified waveform SWF may be determined without dividing the entire section of the simplified waveform SWF into a plurality of symbolized intervals Ej. For example, the waveform symbol Cw of the simplified waveform SWF may be determined using pattern matching. In pattern matching, matching is performed between the simplified waveform SWF and a plurality of template waveforms created in advance, and the waveform symbol preset for the template waveform with the highest matching degree is adopted as the waveform symbol Cw of the simplified waveform SWF. However, if symbolization is performed using a plurality of symbolized intervals Ej, a waveform symbol Cw that clearly represents the increasing and decreasing tendencies of the simplified waveform SWF can be obtained regardless of the length of the simplified waveform SWF.

[0020] FIG. 6 is a flowchart showing the detailed procedure of step S50. FIG. 7 is an explanatory diagram showing the extreme value calculation section used in step S50. FIG. 8 is an explanatory diagram showing an example of the symbolized interval Ej and its symbolization process.

[0021] In step S51, the waveform encoding unit 116 calculates the value range Δmax (= max - min) from the maximum value max and the minimum value min of the simplified waveform SWF. As shown in FIG. 7, in the simplified waveform SWF, between the start point P1 and the end point P6, there are points P2 and P4 indicating extreme values. Among these points, the end point P6 indicates the maximum value max, and the point P4 indicates the minimum value min. Therefore, the difference between the values of these points P6 and P4 is calculated as the value range Δmax. If there are no extreme values other than the end points P1 and P6, one of the end points P1 and P6 becomes the point indicating the maximum value max, and the other becomes the point indicating the minimum value min.

[0022] In step S52, the waveform encoding unit 116 determines the intersection points P3 and P5 between the line indicating the median value Vc of the value range Δmax and the simplified waveform SWF, and also determines the extreme value calculation intervals C1, C2, and C3. As shown in FIG. 7, the median value Vc is the average value of the maximum value max and the minimum value min. Each of the individual extreme value calculation intervals C1, C2, and C3 is an interval set between two adjacent points among the intersection points P3 and P5 of the start point P1 and the end point P6 of the simplified waveform SWF and the median value Vc. Specifically, the first extreme value calculation interval C1 is the interval from the start point P1 to the first intersection point P3. The second extreme value calculation interval C2 is the interval from the intersection point P3 to the next intersection point P5. The third extreme value calculation interval C3 is the interval from the intersection point P5 to the end point P6. Here, the intersection points P3 and P5 are determined using the median value Vc, but the intersection points with the simplified waveform SWF may be determined using a reference value other than the median value Vc.

[0023] In step S53, the waveform encoding unit 116 searches for extreme values in each of the extreme value calculation intervals C1 to C3. As shown in FIG. 7, in the first extreme value calculation interval C1, between its both end points P1 and P3, a point P2 indicating a maximum value which is the maximum value of the interval is detected. In the second extreme value calculation interval C2, between its both end points P3 and P5, a point P5 indicating a minimum value which is the minimum value of the interval is detected. In the third extreme value calculation interval C3, there is no extreme value between its both end points P5 and P6. The timing indicating the extreme value detected in the extreme value calculation interval is used as a boundary for dividing the symbolization interval Ej.

[0024] In addition, when the maximum value is less than or equal to the median value Vc, the maximum value may not be used as a boundary for dividing the symbolization interval Ej. Similarly, when the minimum value is greater than or equal to the median value Vc, the minimum value may not be used as a boundary for dividing the symbolization interval Ej. By doing so, since a small variation part showing an increasing and decreasing tendency opposite to that of the entire simplified waveform SWF is not set as the symbolization interval Ej, the waveform symbol Cw can be generated so as to represent only the increasing and decreasing tendency of the entire simplified waveform SWF. That is, the symbolization interval Ej may be divided so as to include a small variation having an increasing and decreasing tendency opposite to that of the entire symbolization interval Ej, rather than a monotonically increasing interval or a monotonically decreasing interval. The maximum value and the minimum value not used as boundaries are excluded from the extreme values detected in step S53.

[0025] As can be understood from these explanations, when dividing the symbolization interval Ej using extreme values, for example, the extreme values can be determined using any of the following methods. <Extreme value determination method M1> In each extreme value calculation period, all maximum values and minimum values are detected as extreme values, and all of those extreme values are adopted as boundaries for dividing the symbolization interval Ej. <Extreme value determination method M2> In each extreme value calculation period, when the maximum maximum value is greater than the median value Vc, that maximum value is adopted as a boundary for dividing the symbolization interval Ej. Also, in each extreme value calculation period, when the minimum minimum value is less than the median value Vc, that minimum value is adopted as a boundary for dividing the symbolization interval. Note that the median value Vc may be used as another reference value.

[0026] In the above method M1, each of the plurality of symbolization intervals Ej divided at the timing of extreme values becomes a monotonically increasing interval or a monotonically decreasing interval. By doing so, since even a small variation part becomes a target for symbolization as one symbolization interval Ej, there is an advantage that a waveform symbol Cw that represents the shape of the simplified waveform SWF in more detail can be obtained.

[0027] On the other hand, in the above method M2, each of a plurality of symbolization intervals Ej divided by the timing of extreme values may include a small variable portion having an increasing / decreasing tendency opposite to that of the entire symbolization interval Ej. When it is assumed that such a small variable portion does not significantly affect the manufacturing result of the product, it is preferable to use this method M2. In this method M2, since the number of determined extreme values is small, the number of characters of the waveform symbol Cw representing the entire simplified waveform SWF is reduced, and there is an advantage that only larger changes can be captured.

[0028] In step S54, the waveform symbolization unit 116 determines whether or not the extreme value detected in step S53 exists in one or more of the extreme value calculation intervals C1 to C3. If there is an extreme value, the process proceeds to step S55 described later, and if there is no extreme value, the process proceeds to step S59. In step S59, the waveform symbolization unit 116 linearly approximates the time series data and symbolizes the slope of the approximation formula. For example, when the slope of the linear approximation formula is positive, the waveform symbol Cw of the entire simplified waveform SWF is determined to be "U". When the slope of the linear approximation formula is negative, the waveform symbol Cw of the entire simplified waveform SWF is determined to be "D". The symbol "U" means that the entire simplified waveform SWF has a shape indicating a tendency for the value to increase. The symbol "D" means that the entire simplified waveform SWF has a shape indicating a tendency for the value to decrease. Thus, when there is no extreme value in the extreme value calculation intervals C1 to C3, since the entire simplified waveform SWF has a waveform that increases or decreases, it is possible to represent the increasing / decreasing tendency with a waveform symbol Cw composed of one character.

[0029] In steps S55 to S57, the waveform encoding unit 116 divides the entire section of the simplified waveform SWF into a plurality of encoding sections Ej, and encodes the waveform in each encoding section Ej. That is, in step S55, the waveform encoding unit 116 sets a plurality of encoding sections Ej. In the example of FIG. 8, j = 1 to 3. Each encoding section Ej is a section set between two adjacent points among the start point P1 and the end point P6 of the simplified waveform SWF and the points P2 and P4 indicating extreme values. Specifically, the first encoding section E1 is the section from the start point P1 to the point P2. The second encoding section E2 is the section from the point P2 to the point P4. The third encoding section E3 is the section from the point P4 to the end point P6. Each encoding section Ej in the example of FIG. 8 is a monotonically increasing section in which the value monotonically increases, or a monotonically decreasing section in which the value monotonically decreases. In the present disclosure, the monotonically increasing section and the monotonically decreasing section are also referred to as "monotonically changing sections". However, as described regarding the method for detecting extreme values in step S53, each encoding section Ej does not have to be a monotonically changing section. That is, each encoding section Ej may include small up-and-down fluctuations having an increasing and decreasing tendency opposite to the increasing and decreasing tendency of the entire encoding section Ej.

[0030] In step S56, the waveform encoding unit 116 calculates the change Δj in the value in each encoding section Ej. The change Δj in the value is equal to the difference in the values at both end points of each encoding section Ej.

[0031] In step S57, the waveform encoding unit 116 determines the section waveform symbol of each encoding section Ej from the change Δj in the value in each encoding section Ej.

[0032] FIG. 9 is an explanatory diagram showing a method for determining a section waveform symbol. In this example, from the relationship between the change Δj in the value, the reference value Δmax / 2, and the lower limit allowable value Δmax / 4, the section waveform symbol is determined according to the following rules. <Encoding rule CR1> When Δj ≧ Δmax / 2 and the start point of the encoding section Ej is a maximum value, the period waveform symbol = "D". <Encoding rule CR2> When Δmax / 2 > Δj ≧ Δmax / 4 and the starting point of the symbolization interval Ej is a maximum value, the period waveform symbol = "d". <Symbolization rule CR3> When Δj ≧ Δmax / 2 and the starting point of the symbolization interval Ej is a minimum value, the period waveform symbol = "U". <Symbolization rule CR4> When Δmax / 2 > Δj ≧ Δmax / 4 and the starting point of the symbolization interval Ej is a minimum value, the period waveform symbol = "u".

[0033] As a reference value for determining the magnitude of the value change Δj, a positive value other than Δmax / 2 may be used. Also, as a lower limit allowable value, a positive value other than Δmax / 4 may be used.

[0034] The "D" of the interval waveform symbol indicates that the value has a tendency to decrease (going Down) and also means that the change in the value is large. The "d" of the interval waveform symbol indicates that the value has a tendency to decrease and also means that the change in the value is small. The "U" of the interval waveform symbol indicates that the value has a tendency to increase (going Up) and also means that the change in the value is large. The "u" of the interval waveform symbol indicates that the value has a tendency to increase and also means that the change in the value is small. In the example of Fig. 8, the interval waveform symbols for the three symbolization intervals E1 to E3 are "u", "D", and "U", respectively.

[0035] In symbolization rules CR2 and CR4, the condition "Δj ≧ Δmax / 4" for the magnitude relationship between the value change Δj and the lower limit allowable value Δmax / 4 may be omitted. However, if this condition is used, for the symbolization interval Ej where the value change Δj is smaller than the lower limit allowable value Δmax / 4, symbolization can be prevented. In other words, when the symbolization interval Ej shows a small variation less than the lower limit allowable value, that small variation can be prevented from being symbolized. The symbolization interval Ej that was not symbolized may simply be ignored and the symbolization of the next symbolization interval may be executed. Alternatively, the symbolization interval Ej that was not symbolized may be combined with the next symbolization interval and symbolization may be executed for the combined symbolization interval.

[0036] As shown in the example of FIG. 9, a one-character symbol representing the increasing or decreasing trend of data values in each symbolization interval Ej is called an "increase / decrease symbol". By constructing the interval waveform symbol with the increase / decrease symbol, the waveform shape of each symbolization interval Ej can be represented by a simple character symbol. Also, in the example of FIG. 9, in the symbolization interval Ej where the data value shows a decreasing trend, when the change in the data value is large, the capital letter "D" is used as the increase / decrease symbol, and when the change in the data value is small, the lowercase letter "d" is used as the increase / decrease symbol. Similarly, in the symbolization interval Ej where the data value shows an increasing trend, when the change in the data value is large, the capital letter "U" is used as the increase / decrease symbol, and when the change in the data value is small, the lowercase letter "u" is used as the increase / decrease symbol. In this way, if the increasing or decreasing trend of the data value and the magnitude of the change are represented by a one-character increase / decrease symbol, the waveform shape of each symbolization interval Ej can be represented by a simple character symbol. Also, if the magnitude of the change in the data value is distinguished by the capital and lowercase letters of the same alphabet, there is an advantage that the meaning of the interval waveform symbol is easy to understand. Note that the increasing or decreasing trend of the data value and the magnitude of the change are collectively referred to as the "increasing / decreasing state of the data value".

[0037] The increase / decrease symbol constituting the interval waveform symbol may be composed of a plurality of characters. In this case, the increasing / decreasing state of the data value can be classified and expressed in a larger number. Also, a fixed-length interval length symbol representing the length of the symbolization interval may be added before or after the increase / decrease symbol. If the interval waveform symbol is constructed to include the increase / decrease symbol and the interval length symbol, the characteristics of the waveform can be represented in more detail.

[0038] In step S58, the waveform symbolization unit 116 integrates the interval waveform symbols of the plurality of symbolization intervals Ej to determine the waveform symbol Cw of the entire simplified waveform SWF. In the example of FIG. 8, since the interval waveform symbols for the three symbolization intervals E1 to E3 are "u", "D", and "U", the waveform symbol Cw of the entire simplified waveform SWF is "uDU".

[0039] The waveform symbol Cw in this embodiment is composed of one or more characters. Specifically, when the entire simplified waveform SWF is a waveform in which the value increases or decreases, the waveform symbol Cw representing the simplified waveform SWF is composed of one character. Also, when the entire simplified waveform SWF is divided into n symbolized intervals E1 to En in which the values increase or decrease respectively, the waveform symbol Cw representing the simplified waveform SWF is composed of n characters. n is an integer of 2 or more, and in the case of FIG. 8, n = 3.

[0040] When the waveform symbol Cw of the simplified waveform SWF is thus determined, the process proceeds to step S60 in FIG. 4, and the analysis unit 118 calculates other characteristic values of the time-series data TD. As other characteristic values of the time-series data TD, for example, the period, average value, maximum value, minimum value, etc. of the time-series data TD can be used. However, step S60 may be omitted.

[0041] In step S70, the analysis unit 118 registers the characteristic values of the time-series data TD including the waveform symbol Cw of the simplified waveform SWF as explanatory variables in the analysis data mart.

[0042] FIG. 10 is an explanatory diagram showing an example of the analysis data mart DMa including the characteristic values of the time-series data TD. This analysis data mart DMa corresponds to the product data mart DM shown in FIG. 2 with the characteristic values of the time-series data TD added as explanatory variables. As the characteristic values of the time-series data TD, the waveform symbol Cw of the simplified waveform SWF and the average temperature during processing are registered. When the time-series data TD is data for each lot, the characteristic values of the time-series data TD are added to all products with the same lot ID. Note that only the waveform symbol Cw of the simplified waveform SWF may be registered as the characteristic value of the time-series data TD in the analysis data mart DMa. In this case, step S60 described above is omitted.

[0043] In step S80, the analysis unit 118 performs a cause analysis of the product using the analysis data mart. This cause analysis is a process of examining the causes of product defects. The cause analysis can be performed using any model such as a multivariate analysis model or a machine learning model, for example.

[0044] As described above, in this embodiment, since the simplified waveform SWF is used to create the waveform symbol Cw represented by one or more characters as an explanatory variable, the waveform symbol Cw representing the waveform of the time series data TD can be easily created as an explanatory variable suitable for the cause analysis of the product. Further, in this embodiment, the waveform symbol Cw representing the shape of the time series data TD that cannot be understood by normal characteristic values such as the period of the time series data TD can be added as an explanatory variable.

[0045] The waveform symbol Cw of this embodiment is composed of characters that are easy for the user to understand. For example, when the waveform symbol Cw is "UD", it can be understood that the simplified waveform SWF is a waveform having a simple shape convex upward. Also, when the waveform symbol Cw is "UDU", it can be understood that the simplified waveform SWF is close to the shape of a cubic function and there is a change in the value corresponding to one period. Thus, in this embodiment, since the waveform symbol Cw of the simplified waveform SWF is composed of characters that are easy for the user to understand, when the user sees the waveform symbol Cw, it is possible to imagine the approximate shape of the time series data TD.

[0046] Furthermore, in this embodiment, since other characteristic values of the time series data TD are added to the analysis data mart, the cause analysis of the product can be performed from more information, leading to the extraction of the cause of defects.

[0047] · Other forms: The present disclosure is not limited to the above-described embodiments, and can be implemented in various forms without departing from the spirit thereof. For example, the present disclosure can also be implemented by the following aspects (aspect). The technical features in the above embodiments corresponding to the technical features in each of the following aspects can be appropriately replaced or combined in order to solve part or all of the problems of the present disclosure, or to achieve part or all of the effects of the present disclosure. Further, if the technical feature is not described as essential in this specification, it can be appropriately deleted.

[0048] (1) According to a first aspect of the present disclosure, a method for creating an explanatory variable used for factor analysis of a product is provided. This method includes: (a) a step of acquiring time-series data in a manufacturing process of the product; (b) a step of extracting a simplified waveform obtained by simplifying the time-series data; and (c) a step of creating a waveform symbol representing the simplified waveform with one or more characters as the explanatory variable. According to this method, a waveform symbol representing the waveform of time-series data can be easily created as an explanatory variable suitable for factor analysis of a product.

[0049] (2) In the above method, the step (b) may include a step of extracting a trend waveform of the time-series data as the simplified waveform. According to this method, the simplified waveform can be easily extracted.

[0050] (3) In the above method, the step (c) may include: (c1) a step of dividing the entire section of the simplified waveform into a plurality of symbolized sections according to a preset rule; (c2) a step of determining a section waveform symbol representing the waveform in each of the plurality of symbolized sections; and (c3) a step of determining the waveform symbol by integrating the section waveform symbols for the plurality of symbolized sections. According to this method, a waveform symbol representing the overall waveform shape of the simplified waveform can be determined in a simple process.

[0051] (4) In the above method, the interval waveform symbol may include a one-character increase / decrease symbol that represents the increasing / decreasing trend of the data value in each symbolization interval. According to this method, the waveform shape of each symbolization interval can be represented by a simple character symbol.

[0052] (5) In the above method, the interval waveform symbol may include a one-character increase / decrease symbol that represents the increasing / decreasing trend and the magnitude of change of the data value in each symbolization interval. According to this method, the waveform shape of each symbolization interval can be represented by a simple character symbol.

[0053] (6) In the above method, the increase / decrease symbol may distinguish and indicate the magnitude of the change of the data value in each symbolization interval using uppercase and lowercase letters of the same alphabet. According to this method, depending on whether the alphabet is uppercase or lowercase, the magnitude of the data value in the symbolization interval can be easily recognized.

[0054] (7) According to the first aspect of the present disclosure, a computer program for creating an explanatory variable to be used in factor analysis of a product is provided. This computer program causes a computer to execute: (a) a process of acquiring time-series data in the manufacturing process of the product; (b) a process of extracting a simplified waveform obtained by simplifying the time-series data; and (c) a process of creating, as the explanatory variable, a waveform symbol that represents the simplified waveform with one or more characters.

[0055] The present disclosure can also be realized in various other forms. For example, it can be realized in the form of an analysis device, a computer program for realizing the functions of the analysis device, a non-transitory storage medium on which the computer program is recorded, and the like.

Explanation of Signs

[0056] 100…Analysis device, 110…Processor, 112…Time series data acquisition unit, 114…Waveform extraction unit, 116…Waveform symbolization unit, 118…Analysis unit, 120…Memory, 130…Interface circuit, 140…Input device, 150…Display device

Claims

1. A method for creating explanatory variables used in factor analysis of a product, comprising: (a) obtaining time-series data in the manufacturing process of the product; (b) extracting a simplified waveform obtained by simplifying the time-series data; (c) creating, as the explanatory variable, a waveform symbol representing the simplified waveform with one or more characters. A method comprising the above.

2. The method according to claim 1, wherein step (b) includes extracting a trend waveform of the time-series data as the simplified waveform.

3. The method according to claim 1, wherein step (c) includes: (c1) dividing the entire section of the simplified waveform into a plurality of symbolized sections according to a preset rule; (c2) determining a section waveform symbol representing the waveform in each of the plurality of symbolized sections; (c3) determining the waveform symbol by integrating the section waveform symbols for the plurality of symbolized sections. A method comprising the above.

4. The method according to claim 3, wherein the section waveform symbol includes an increase / decrease symbol of one character representing the increase / decrease trend of data values in each symbolized section.

5. The method according to claim 3, wherein the section waveform symbol includes an increase / decrease symbol of one character representing the increase / decrease trend and the magnitude of change of data values in each symbolized section.

6. The method according to claim 5, wherein the increase / decrease symbol indicates the magnitude of change of the data values in each symbolized section by distinguishing between capital and small letters of the same alphabet.

7. A computer program for creating explanatory variables used in factor analysis of a product, the program causing a computer to execute: (a) a process of obtaining time-series data in the manufacturing process of the product; (b) a process of extracting a simplified waveform obtained by simplifying the time-series data; (c) a process of creating, as the explanatory variable, a waveform symbol representing the simplified waveform with one or more characters. A computer program.

Images