Data analysis apparatus, method, and program

Abstract

To improve the accuracy in estimating an anomaly cause in a manufacturing process.SOLUTION: A data analysis system in accordance with an embodiment includes a first requirement designation unit, a second requirement designation unit, a factor acquisition unit, a calculation unit, and a similarity calculation unit. The first requirement designation unit designates a first requirement representing a first product that is an object of analysis. The second requirement designation unit designates a second requirement representing a second product that is an object of comparison. The factor acquisition unit acquires first factor data representing plural first manufacture requirements on the first product on the basis of the first requirement, and acquires second factor data representing plural second manufacture requirements on the second product on the basis of the second requirement. The calculation unit calculates a first index value relevant to a degree to which each of the first manufacture requirements contributes to an anomaly cause, on the basis of the first factor data, and calculates a second index value relevant to a degree to which each of the second manufacture requirements contributes to the anomaly cause, on the basis of the second factor data. The similarity calculation unit calculates a similarity between the first index value and second index value.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Embodiments of the present invention relate to a data analysis apparatus, method, and program.

Background Art

[0002] In the production of products, improvement in productivity is required. To improve productivity, it is important to maintain and improve the yield of products. In many manufacturing industries, data in the manufacturing process is collected, monitored, and analyzed to detect abnormalities and identify the causes of the abnormalities. Thereafter, countermeasures are taken against the causes of the abnormalities to maintain and improve the yield. However, it is necessary to shorten the period from the occurrence of an abnormality to the implementation of countermeasures. This is because when the period until the implementation of countermeasures is short, the production of defective products can be reduced and a high yield can be achieved.

[0003] On the other hand, methods for automatically detecting abnormalities and methods for estimating the causes of abnormalities are known. For example, in a method for automatically detecting abnormalities, abnormal products that deviate from outlier values or standard values are automatically detected for individual data of products such as the dimensions and characteristic values of the products. In a method for estimating the cause of an abnormality, based on the individual data of the detected abnormal product, among the individual data of past abnormal products, an abnormal case having similar individual data is searched for, and the discovered past abnormal case is presented.

[0004] According to the study by the present inventor, in a method for estimating the cause of an abnormality, for example, when a plurality of past abnormal cases having similar individual data are discovered, even if the individual data is similar in the plurality of abnormal cases, the causes of abnormalities in the manufacturing process are not necessarily similar. Therefore, according to the study by the present inventor, the method for estimating the cause of an abnormality is in a situation where the accuracy of estimating the cause of an abnormality in the manufacturing process is low.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

[0006] The problem that this invention aims to solve is to provide a data analysis device, method, and program that can improve the accuracy of estimating the cause of abnormalities in the manufacturing process. [Means for solving the problem]

[0007] The data analysis device according to the embodiment comprises a first condition specification unit, a second condition specification unit, a factor acquisition unit, a calculation unit, and a similarity calculation unit. The first condition specification unit specifies a first condition indicating a first product to be analyzed. The second condition specification unit specifies a second condition indicating a second product to be compared. The factor acquisition unit acquires first factor data indicating a plurality of first manufacturing conditions for the first product based on the first condition, and acquires second factor data indicating a plurality of second manufacturing conditions for the second product based on the second condition. The calculation unit calculates a first index value related to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of the abnormality based on the first factor data, and calculates a second index value related to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of the abnormality based on the second factor data. The similarity calculation unit calculates the similarity between the first index value and the second index value. [Brief explanation of the drawing]

[0008] [Figure 1] A block diagram showing an example of a data analysis device according to the first embodiment. [Figure 2] A figure showing an example of factor data according to the first embodiment. [Figure 3] A figure showing an example of state data according to the first embodiment. [Figure 4] A flowchart illustrating the operation in the first embodiment. [Figure 5] A schematic diagram illustrating the operation in the first embodiment. [Figure 6] A figure showing an example of the first factor data D according to the first embodiment. [Figure 7]Figure showing an example of the second factor data D1 according to the first embodiment. [Figure 8] Figure showing an example of the second factor data D2 according to the first embodiment. [Figure 9] Figure showing an example of a tabulation of the first factor data D according to the first embodiment. [Figure 10] Figure showing an example of a table of bias ratios based on the tabulation according to the first embodiment. [Figure 11] Schematic diagram for explaining the operation in the first embodiment. [Figure 12] Block diagram of a data analysis device according to a modification of the first embodiment. [Figure 13] Block diagram showing an example of a data analysis device according to the second embodiment. [Figure 14] Flowchart for explaining the operation in the second embodiment. [Figure 15] Block diagram of a data analysis device according to the third embodiment. [Figure 16] Figure showing an example of a defective database according to the third embodiment. [Figure 17] Flowchart for explaining the operation in the third embodiment. [Figure 18] Figure showing an example of a display form of a display device according to the third embodiment. [Figure 19] Block diagram of a data analysis device according to a modification of the third embodiment. [Figure 20] Figure showing an example of the hardware configuration of a data analysis device according to the fourth embodiment.

Embodiments for Carrying Out the Invention

[0009] Hereinafter, each embodiment will be described with reference to the drawings. In the following description, the case where the data analysis device analyzes data on products and their manufacturing conditions will be described as an example. Note that the data analysis device may be read as any name such as a similarity calculation device in accordance with specific processing.

[0010] (First Embodiment) FIG. 1 is a block diagram showing a data analysis device according to the first embodiment. The data analysis device 200 includes a first condition specifying unit 210, a second condition specifying unit 220, a factor acquisition unit 230, a calculation unit 240, and a similarity calculation unit 250. The data analysis device 200 is connected to a manufacturing database 100 in which data related to product manufacturing is recorded. The manufacturing database 100 and a defect database (not shown) may be provided separately from the data analysis device 200, for example, or may be provided in the data analysis device 200.

[0011] As shown in FIGS. 2 and 3, the manufacturing database 100 stores manufacturing data including factor data 100D and state data 100S. Note that the factor data 100D is information on manufacturing conditions such as the devices and materials used in product manufacturing, and the settings of the devices. The state data 100S is data related to the state of the product, such as the dimensions and electrical characteristics of the product. Each of the factor data 100D and the state data 100S has a manufacturing number for identifying which product's data it is, and can be associated with each data using the manufacturing number as a join key. For example, the factor data 100D stores by associating the manufacturing number of the product with data 1 to 5 indicating the manufacturing conditions of the product. The state data 100S stores by associating the manufacturing number of the product with state data indicating the state of the product.

[0012] Here, factor data 100D more generally uses information related to 5M1E as manufacturing conditions. 5M1E is an acronym for Man, Machine, Material, Method, Measurement, and Environment, and is widely known as the six factors for controlling the manufacturing process. Man information includes information such as the name of the machinist. Machine information includes information such as the name of the equipment used to manufacture the product, the name of the production line, and the state of the equipment during processing, such as temperature and pressure. Material information includes information such as the ID or name of the material used to manufacture the product, and the ID or name of the parts that make up the product. Method information includes information such as the processing method of the product and the type of processing program. Measurement information includes information such as the name of the equipment used for measurement and the measurement location on the product. Environment information includes information such as the name of the factory building where measurement was performed, and the temperature and humidity at the time of measurement. For example, the following information (Da) to (Dd) may also be included as manufacturing conditions. However, the information that may be included as manufacturing conditions is not limited to the following information (Da) to (Dd).

[0013] (Da) Manufacturing lot representing the manufacturing unit, the date the product was manufactured (manufacturing date), and the time the product passed through the equipment and processes used in its manufacture.

[0014] (Db) Equipment, materials, and names of personnel involved in the manufacturing of the product.

[0015] (Dc) Settings for the manufacturing equipment, such as voltage and device mode.

[0016] (Dd) Output values ​​of manufacturing and inspection equipment, data on the product's condition such as dimensions and electrical characteristics.

[0017] The state data 100S more generally uses information related to product quality control (QC) as state data. Alternatively, the state data may include data associated with individual products that are deemed useful for analysis. For example, the state data may include the following (Sa) and (Sb) data. However, the data that may be included as state data is not limited to the following (Sa) and (Sb) data.

[0018] (Sa) Data used for product quality control (product dimensions, voltage, resistance, and other electrical characteristics).

[0019] (Sb) Flag information representing the inspection results of the product.

[0020] The manufacturing database 100 may be configured as a general relational database management system (RDBMS). The manufacturing database 100 may, for example, be a NoSQL (Not only SQL) database. Furthermore, the manufacturing data stored in the manufacturing database 100 may consist of files in a predetermined format, such as CSV (Comma Separated Value).

[0021] The first condition specification unit 210 specifies a first condition that indicates the product to be analyzed (first product). Specifically, for example, the first condition specification unit 210 specifies a first condition that indicates the group of products to be analyzed in the manufacturing database 100. For example, a list of multiple manufacturing numbers can be prepared, and the products included in that list can be specified. For example, in the case of the factor data 100D shown in Figure 2, this would correspond to specifying manufacturing numbers XXXX-00001 to XXXX-00010 as the first condition. Alternatively, the product to be analyzed may be specified using something other than the manufacturing number, such as products that satisfy predetermined conditions in the factor data 100D. For example, this would apply when specifying conditions for factor data 100D such as manufacturing lot or manufacturing date.

[0022] The second condition specification unit 220 specifies a second condition indicating the product to be compared (second product). The specification method is the same as the first condition specification unit 210; it may be specified using the serial number, or it may be specified using something other than the serial number, such as a product whose factor data 100D satisfies predetermined conditions. The second condition specifies a product different from the first condition. Note that there may be some overlap with the product specified in the first condition. Multiple conditions may be specified in the second condition, for example, when searching for similar cases from among multiple cases. In this case, the similarity score will be calculated for each second condition.

[0023] The factor acquisition unit 230 acquires first factor data indicating multiple first manufacturing conditions for the product under analysis (first product) based on the first condition, and acquires second factor data indicating multiple second manufacturing conditions for the product being compared (second product) based on the second condition. For example, the factor acquisition unit 230 acquires factor data from the factor data 100D in the manufacturing database 100 for the products specified by the first and second conditions.

[0024] The calculation unit 240 calculates a first index value based on the first factor data, which relates to the degree to which each of the multiple first manufacturing conditions contributes to the cause of abnormality in the product under analysis (first product). The calculation unit 240 also calculates a second index value based on the second factor data, which relates to the degree to which each of the multiple second manufacturing conditions contributes to the cause of abnormality in the product being compared (second product). Here, the degree to which a factor contributes to the cause of abnormality in a product is a value that represents how much the factor data indicating the manufacturing conditions of the product influenced the occurrence of the abnormality in that product.

[0025] The similarity calculation unit 250 calculates the similarity between the first index value and the second index value. As a method for calculating similarity, for example, Pearson's product-moment correlation coefficient may be used as a distance index, or other mathematical distance indexes such as L1, L2 norm, cosine similarity may be used. Alternatively, as a method for calculating similarity, an index that does not satisfy the axioms of distance but quantifies the difference between two data sets, such as Kullback-Leibler divergence, may be used. Alternatively, as a method for calculating similarity, dissimilarity (degree of dissimilarity) may be used.

[0026] Next, the operation of the data analysis device configured as described above will be explained with reference to the flowchart in Figure 4 and the schematic diagrams in Figures 5 to 11.

[0027] (Step ST10) As shown in Figures 4 and 5, the first condition specification unit 210 specifies a first condition that indicates the product to be analyzed (hereinafter also referred to as the first product). For example, the first condition specification unit 210 specifies the manufacturing numbers XXXX-00001 to XXXX-00010, which indicate the product to be analyzed, from the factor data 100D shown in Figure 2, as the first condition.

[0028] (Step ST20) The second condition specification unit 220 specifies a second condition that indicates the product to be compared (hereinafter also referred to as the second product). For example, the second condition specification unit 220 specifies the manufacturing numbers YYYY-00001 to YYYY-00010, which indicate the product to be compared, from the factor data 100D shown in Figure 2, as the first second condition. Similarly, for example, the second condition specification unit 220 specifies the manufacturing numbers ZZZZ-00001 to ZZZZ-00010, which indicate the second product, from the factor data 100D, as the second second condition. However, it is not necessary to specify a second or subsequent second condition.

[0029] (Step ST30) The factor acquisition unit 230 acquires first factor data representing multiple first manufacturing conditions for the first product based on the first condition. The factor acquisition unit 230 also acquires second factor data representing multiple second manufacturing conditions for the second product based on two second conditions. Each acquired factor data consists of table data with a number of rows equal to the number of conditions and a number of columns equal to the number of factor items. The number of conditions, in the case of the first condition, is the number of manufacturing numbers specified in the first condition. Similarly, the number of conditions, in the case of the second condition, is the number of manufacturing numbers specified in the second condition. The number of manufacturing numbers is also the number of products.

[0030] For example, if the first condition specifies 10 products and 5 factors, the factor data will be a 10-row, 5-column table. The second condition is similar. If multiple second conditions are specified, the table data corresponding to the number of conditions will be obtained. For example, if two second conditions are specified, the two second conditions will be referred to as Second Condition 1 (15 products) and Second Condition 2 (10 products). In this case, the factor data corresponding to Second Condition 1 will be a 15-row factor data, and the factor data corresponding to Second Condition 2 will be a 10-row factor data.

[0031] The following explanation uses the example of searching for a second product group (second condition i (i=1,…,I)) that has previously experienced an anomaly similar to the first product group (first condition) in which an anomaly occurred. "I" refers to the number of conditions in the second condition, and for the sake of explanation, we will use 2 conditions (I=2). Figure 6 shows the first factor data D obtained from the manufacturing database 100 using the first condition. Figure 7 shows the second factor data D1 obtained using second condition 1, and Figure 8 shows the second factor data D2 obtained using second condition 2. The first factor data D obtained using the first condition is a table data where each column is the manufacturing condition Cj (j=1,…,J=5) and each row is the second product. The second factor data Di obtained using the second condition is similar. Based on the above settings, the following explanation will be given.

[0032] (Step ST40) The calculation unit 240 calculates a first index value F(D) for the first factor data D, which is related to the degree to which each manufacturing condition in the first factor data D contributes to the occurrence of the first product specified by the first condition. The calculation unit 240 also calculates a second index value F(Di) for the second factor data Di, which is related to the degree to which each manufacturing condition in the first factor data D contributes to the occurrence of the second product specified by the second condition. The degree to which each manufacturing condition contributes to the cause of the first product's abnormality is a value that represents how much each manufacturing condition in the second factor data Di contributes to the cause of the second product's abnormality. Similarly, the degree to which each manufacturing condition in the second factor data Di contributes to the cause of the second product's abnormality is a value that represents how much each manufacturing condition in the second factor data Di contributes to the cause of the second product's abnormality.

[0033] Here, we quantify the bias related to specific manufacturing conditions as an index value, but the index value is not limited to this. When quantifying the bias, for example, a summary table is created for each manufacturing condition Cj (each column of the factor) of the factor data D of the first condition, and the bias rate (frequency distribution Od{d=1,2…,K}) for each element of the manufacturing condition is obtained by dividing each element of the summary table by the total number of products. Then, the maximum value of this bias rate for each element of manufacturing condition Cj is taken as the bias rate rj of manufacturing condition Cj, and the vector with rj as an element is quantified as the first index value F(D)=(r1,…,rJ). The second index value F(Di) can be obtained in a similar manner.

[0034] Regarding the manufacturing condition Cj of the first factor data D in Figure 6, Figure 9 shows the summary table T1, and Figure 10 shows table T2, which shows the bias rate obtained by dividing each element of summary table T1 by the total number of products. The bias rate for manufacturing condition C1 is 0.5, the bias rate for manufacturing condition C2 is 1.0, the bias rate for manufacturing condition C3 is 0.3, the bias rate for manufacturing condition C4 is 0.4, and the bias rate for manufacturing condition C5 is 0.4, so the first index value F(D) is obtained as (0.5, 1.0, 0.3, 0.4, 0.4). For example, the bias rate of "1.0" for manufacturing condition C2 reflects a high bias towards item C. The bias related to the manufacturing condition can also be quantified for the second condition using a similar method.

[0035] (Step ST50) The similarity calculation unit 250 calculates the similarity Si between the first index value F(D) and the second index value F(Di). In this example, as shown in Figure 11, the similarity calculation unit 250 calculates the similarity S1 between the first index value F(D) and the second index value F(D1) when the second condition i = 1. Similarly, the similarity calculation unit 250 calculates the similarity S2 between the first index value F(D) and the second index value F(D2). Note that the similarity Si may be, for example, a mathematical distance index, or an index that quantifies the difference between the two data, even if it is not a distance index.

[0036] Subsequently, by selecting the higher of the two calculated similarity scores S1 and S2, a second condition i with a similar bias in manufacturing condition Cj can be searched for. Furthermore, from the second index value F(Di) corresponding to the higher similarity score Si and the second factor data Di, the cause of the anomaly can be inferred as an item of manufacturing condition Cj with a high bias.

[0037] As described above, according to the first embodiment, the first condition specification unit 210 specifies a first condition indicating the first product to be analyzed. The second condition specification unit 220 specifies a second condition indicating the second product to be compared. The factor acquisition unit 230 acquires first factor data indicating a plurality of first manufacturing conditions for the first product based on the first condition, and acquires second factor data indicating a plurality of second manufacturing conditions for the second product based on the second condition. The calculation unit 240 calculates a first index value based on the first factor data, relating to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of abnormality in the first product. The calculation unit 240 also calculates a second index value based on the second factor data, relating to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of abnormality in the second product. The similarity calculation unit 250 calculates the similarity between the first index value and the second index value.

[0038] Thus, according to the first embodiment, by calculating an index value based on the product's manufacturing conditions and calculating the similarity of the index values, the similarity of the manufacturing conditions is taken into consideration, which improves the accuracy when estimating the cause of an abnormality in the manufacturing process.

[0039] To elaborate, the first comparative example involves a method that searches for similar individual data among past abnormal product individual data based on the individual data of the detected abnormal product, and then presents the found past abnormal cases. In the first comparative example, the individual data in question consists of values ​​output from equipment such as manufacturing equipment and inspection equipment, such as product dimensions and characteristic values. Therefore, the first comparative example does not consider the similarity of manufacturing conditions, such as which equipment was used for the product, which materials were used, or the settings of the equipment, resulting in a low accuracy in estimating the cause of abnormalities in the manufacturing process.

[0040] Furthermore, as a second comparative example, there is a method that supports root cause investigation by calculating an index value indicating the likelihood of a cause based on the bias of anomalies for each manufacturing condition in various data obtained during the manufacturing process, and estimating the manufacturing conditions that cause the anomalies. However, since the second comparative example does not consider past cases, it is not possible to search for similar past cases. For this reason, in the second comparative example, if multiple index values ​​are high and calculated to be likely causes, it may take time to identify the true cause of the anomaly.

[0041] In contrast, according to the first embodiment, for example, if multiple cases with similar individual data are found, by considering the similarity of manufacturing conditions, it is possible to separate past cases in which anomalies occurred under the same manufacturing conditions from other past cases. Furthermore, by searching for anomalies with similar manufacturing conditions such as equipment and materials, the cause of the anomaly can be estimated with higher accuracy, and the time spent by engineers at the manufacturing site in identifying the cause of the anomaly can be shortened. Consequently, it is expected that the period until countermeasures can be implemented will also be shortened. Therefore, according to the first embodiment, when presenting similar past cases based on manufacturing conditions, the efficiency of cause investigation can be improved by searching and presenting past cases that are narrowed down to cases with similar causes of anomalies.

[0042] (Modified version of the first embodiment) Next, we will describe various modifications of the first embodiment. Each modification can also be applied to the following embodiments.

[0043] The first embodiment uses two second conditions i (i=1,2), but is not limited thereto. For example, the second condition specification unit 220 may specify one second condition i different from the first condition, or it may specify three or more second conditions i different from the first condition. Regardless of which of the one or more second conditions i different from the first condition is specified, the factor acquisition unit 230 acquires second factor data Di for each second condition i. The calculation unit 240 calculates a second index value F(Di) for each second factor data Di. The similarity calculation unit 250 calculates a similarity Si for each second index value F(Di). Therefore, this modification provides the same effects as the first embodiment. Furthermore, with this modification, for example, if it is desired to confirm that the cause of the abnormality in the current first product is the same as the typical cause of the abnormality in past second products, the cause of the abnormality can be confirmed by specifying one second condition i.

[0044] Furthermore, while the first embodiment uses the bias rate for each manufacturing condition as the index value, it is not limited to this. For example, the calculation unit 240 may use a method to quantify the bias related to specific manufacturing conditions as an index value within the framework of a statistical test. In this case, the calculation unit 240 calculates the first index value based on the first factor data and the statistical hypothesis test, and calculates the second index value based on the second factor data and the statistical hypothesis test. Below, a modified example using the G-test, a likelihood ratio test framework, as a statistical test for nominal scale variables such as manufacturing equipment will be described, but it is not limited to this. For example, a chi-squared test may be used for the statistical hypothesis test. Furthermore, the calculation unit 240 may use other testing methods as well.

[0045] Here, the calculation unit 240 calculates the probability value p-value obtained by testing the significance of the bias for each manufacturing condition Cj (j=1,…,J), which is each column of the first factor data D, and sets the first index value F(D)=(p1,…,pJ) as a vector whose elements are the p-values ​​for each manufacturing condition Cj. When calculating the p-value for each manufacturing condition, a summary table is created for each manufacturing condition Cj (each column) of the factor data D of the first condition, and the frequency distribution Od{d=1,2…,K} for each manufacturing condition item d is obtained by dividing each element of the summary table by the total number of products. Let K be the number of types of manufacturing condition items. At this time, the manufacturing data of the first product under analysis is considered as the population, and the null hypothesis is set that "the distribution of each manufacturing condition item for a product in a certain state (abnormal product) is the same as the distribution obtained by random sampling from the population." Next, the null hypothesis is tested and its p-value is calculated. A smaller p-value indicates a higher probability of rejecting the hypothesis, suggesting that the results are not identical to the distribution obtained from random sampling, meaning that the incidence of abnormal products is higher under specific manufacturing conditions. Therefore, a small p-value suggests that manufacturing condition Cj contributes significantly to the cause of the abnormality. The G-value, the test statistic for the G-test, is calculated using the following formula.

[0046]

number

[0047] Ed is the number of products expected under the null hypothesis, and is calculated using the following formula.

[0048]

number

[0049] P(d) is the expected probability, which is the probability that a product judged to be abnormal in item d will occur if the null hypothesis is true. If the true value of the expected probability is unknown, it is approximated by N / K, where N is the total number of products and K is the number of types of manufacturing condition items. Next, the p-value corresponding to the G-value is calculated using the following formula with the chi-squared distribution f(x,k).

[0050]

number

[0051] k represents the degrees of freedom of the chi-squared distribution, where k = K-1. The larger the degrees of freedom k, the less likely the p-value is to be small in the chi-squared distribution. When the number of types of manufacturing condition items K is large, bias is more likely to occur even with random sampling. However, due to the above properties, the significance of the bias is evaluated considering the number of items K. According to the above modification, the calculation unit 240 can calculate the p-value for each manufacturing condition Cj and calculate the first index value F(D), which is a vector with the calculated p-values ​​as elements. The second index value F(Di) can be calculated for the second condition in a similar manner. Subsequently, as described above, by determining the similarity between the first index value F(D) and the second index value F(Di), it is possible to search for a second condition i with similar manufacturing condition biases.

[0052] Furthermore, in the first embodiment, the bias rate for each manufacturing condition was used as the index value, but this is not limited to this. For example, the calculation unit 240 may use a method of quantification using a model that takes factor data D as input and outputs a first index value F(D). The model may be designed using machine learning or an arbitrarily designed function. An arbitrarily designed function is, for example, a logistic regression model, but is not limited to this. In the case of model design using machine learning, there are unsupervised models and supervised models. In the case of a supervised model, a ground truth label is given in advance for each analysis range, and the model is trained so that the output data approximates the input data given the same ground truth label. Alternatively, this can be achieved by training the model so that the output data does not approximate the input data given different ground truth labels. In the case of an unsupervised model, this can be achieved by using a clustering model such as K-Means to design the model so that similar factor data D are classified into the same class. In any case, the calculation unit 240 calculates the first index value and the second index value using a trained model that has been trained to output index values ​​based on the input factor data. Therefore, according to this modification, the method for setting the first indicator value F(D) from the factor data D can be determined in a data-driven manner.

[0053] Furthermore, in the first embodiment, the second index value F(Di) was calculated based on the second factor data Di, but the invention is not limited to this. For example, the data analysis device 200 may include a storage unit 232 in which the second condition and the second index value are stored in association with each other, as shown in Figure 12. Specifically, the storage unit 232 stores conditions such as the first condition and the second condition, and index values ​​such as the calculated first index value and the second index value in association with each other. However, since the first condition and the first index value are stored after the calculation of the first index value F(D), when a new first condition is specified, the storage unit 232 does not store the new first condition or the first index value F(D) corresponding to the new first condition. The calculation unit 240 can also search the storage unit 232 based on the specified second condition and obtain the second index value F(Di) from the storage unit 232. In other words, according to this modified version, in addition to the effects of the first embodiment, after first calculating the second index value F(Di) corresponding to the second condition, when the same second condition is specified again, the second index value F(Di) can be obtained from the storage unit 232 without calculating it again.

[0054] Next, since a specific example of calculating similarity was not described in the first embodiment, it will be described below. For example, when F(D)=(p1,…,pJ) and F(Di)=(pi,1,…,pi,J), the similarity calculation unit 250 calculates the correlation coefficient Si between the first index value F(D) of the first condition and the second index value F(Di) of the second condition i as the similarity, as shown in the following equation.

[0055]

number

[0056] Here, the "-" attached to p is a bar symbol representing the average value. From now on, p with a bar symbol will be referred to as p - It is expressed as p. - This is the average value of F(D). Similarly, pi with a bar symbol is pi - It is expressed as pi. - This is the mean value of F(Di). - pi - These are calculated using the following formulas.

[0057]

number

[0058] The correlation coefficient is an index that measures the strength of a linear relationship between two sets of data, and takes values ​​in the interval [-1, 1] depending on the strength of the relationship. When there is a correlation, the value of the correlation coefficient approaches 1, and when there is an inverse correlation, the value of the correlation coefficient approaches -1. When there is no relationship, the value of the correlation coefficient approaches 0. Therefore, by using the correlation coefficient as a measure of similarity, the relationship between two sets of data can be quantified, and a highly correlated second indicator value F(Di) can be extracted. Furthermore, based on the second factor data Di used to calculate the extracted second indicator value F(Di), a second condition i with similar biases in manufacturing conditions can be searched from the manufacturing database 100 or the storage unit 232.

[0059] Furthermore, in the first embodiment, the first condition is specified (ST10), the second condition is specified (ST20), the factor data D, Di and the index values ​​F(D) and F(Di) are obtained (ST30 and ST40), and the similarity Si is obtained (ST50), but the order is not limited to this. For example, as can be seen from Figure 5, the order may be to specify the first condition, obtain the first factor data D and the first index value F(D), then specify the second condition i, obtain the second factor data Di and the second index value F(Di), and then obtain the similarity Si. Note that the process from specifying the first condition to obtaining the first index value F(D) and the process from specifying the second condition i to obtaining the second index value F(Di) may be executed in any order. The effects of the first embodiment can be obtained even with such modifications.

[0060] (Second embodiment) Next, a second embodiment will be described. Compared to the first embodiment, the data analysis device according to the second embodiment narrows down the first product to be analyzed and the second product to be compared to products in an abnormal state. This allows the data analysis device to further improve the accuracy when estimating the cause of the abnormality.

[0061] Figure 13 is a block diagram showing the configuration of the data analysis device according to the second embodiment. Components similar to those described above are denoted by the same reference numerals, and their detailed descriptions are omitted. Here, we will mainly describe the differences. Similarly, redundant descriptions will be omitted for each of the following embodiments.

[0062] In Figure 13, the data analysis device 200 further includes a state acquisition unit 222 and an anomaly detection unit 224 compared to the configuration shown in Figure 1.

[0063] Here, the state acquisition unit 222 acquires first state data indicating the state of the first product based on the first condition specified by the first condition specification unit 210. Similarly, the state acquisition unit 222 acquires second state data indicating the state of the second product based on the second condition specified by the second condition specification unit 220. In the second embodiment, for example, data used for product quality control (such as product dimensions, voltage, and electrical characteristics like resistance) can be used as appropriate state data.

[0064] The anomaly detection unit 224 detects an abnormal state of the first product based on the first state data and modifies (re-specifies) the first condition to indicate the first product in the detected abnormal state. Similarly, the anomaly detection unit 224 detects an abnormal state of the second product based on the second state data and modifies (re-specifies) the second condition to indicate the second product in the detected abnormal state. For example, the anomaly detection unit 224 may detect an abnormal state of the first product by statistical processing based on the first state data, or it may detect an abnormal state of the second product by statistical processing based on the second state data.

[0065] Accordingly, the factor acquisition unit 230 acquires first factor data based on the modified first condition and acquires second factor data based on the modified second condition.

[0066] The other configurations are the same as in the first embodiment.

[0067] Next, the operation of the data analysis device configured as described above will be explained using the flowchart in Figure 14.

[0068] Now, by executing steps ST10 and ST20, the first and second conditions are specified, as described above.

[0069] (Step ST22) The status acquisition unit 222 acquires first status data indicating the status of the first product from the status data 100S in the manufacturing database 100, based on a specified first condition. Similarly, the status acquisition unit 222 acquires second status data indicating the status of the second product from the status data 100S in the manufacturing database 100, based on a specified second condition.

[0070] (Step ST24) The anomaly detection unit 224 detects an abnormal state of the first product based on the first state data and modifies (re-specifies) the first condition to indicate the first product in the detected abnormal state. Similarly, the anomaly detection unit 224 detects an abnormal state of the second product based on the second state data and modifies (re-specifies) the second condition to indicate the second product in the detected abnormal state.

[0071] For example, the anomaly detection unit 224 detects that a product corresponding to a status data is in an abnormal state if the status data is an outlier or deviates from a standard value. Below, as an example, a method of outlier detection using the 3-sigma statistical method is described, but the anomaly detection unit 224 is not limited to this. The anomaly detection unit 224 may also use methods to detect outliers, such as rule-based or machine learning methods.

[0072] The 3-sigma outlier detection method relies on the statistical assumption that, if the state data follows a normal distribution, 99.7% of the state data will fall within 3 standard deviations of the mean. The remaining 0.3% of state data that do not fall within 3 standard deviations of the mean are outliers and are considered abnormal.

[0073] Therefore, for example, the anomaly detection unit 224 acquires status data from the manufacturing database 100 using the manufacturing number indicated in the specified first condition as the key. Furthermore, if the average of the acquired status data is μ and the standard deviation is σ, the anomaly detection unit 224 detects that the first product with a manufacturing number whose status data is outside the range of μ±3σ is in an abnormal state.

[0074] Subsequently, the anomaly detection unit 224 modifies the specified first condition to narrow it down to the first condition of the abnormal state. Similarly, the anomaly detection unit 224 modifies the specified second condition to narrow it down to the second condition of the abnormal state.

[0075] (Step ST30) The factor acquisition unit 230 acquires first factor data based on the modified first condition and acquires second factor data based on the modified second condition.

[0076] As described above, the processing from step ST40 onwards is executed.

[0077] As described above, according to the second embodiment, the state acquisition unit 222 acquires first state data indicating the state of the first product based on the first condition specified by the first condition specification unit 210. Similarly, the state acquisition unit 222 acquires second state data indicating the state of the second product based on the second condition specified by the second condition specification unit 220. The anomaly detection unit 224 detects an abnormal state of the first product based on the first state data and modifies the first condition to indicate the first product in the detected abnormal state. Similarly, the anomaly detection unit 224 detects an abnormal state of the second product based on the second state data and modifies the second condition to indicate the second product in the detected abnormal state. The factor acquisition unit 230 acquires first factor data based on the modified first condition and acquires second factor data based on the modified second condition. Therefore, in addition to the effects of the first embodiment, by narrowing down the first product to be analyzed and the second product to be compared to products in an abnormal state, the accuracy of estimating the cause of the anomaly can be further improved. Furthermore, after detecting outliers and other abnormal conditions in the product status data, it is possible to search for similar cases for products exhibiting abnormal conditions.

[0078] Furthermore, according to the second embodiment, the anomaly detection unit 224 may detect an abnormal state of the first product by statistical processing based on the first state data, or it may detect an abnormal state of the second product by statistical processing based on the second state data. In this case, in addition to the effects described above, when narrowing down the products to be analyzed or compared to products in an abnormal state, there is no need to prepare a rule base or a trained model in advance, and the abnormal state can be detected by statistical processing.

[0079] (Modified version of the second embodiment) Next, we will describe various modifications of the second embodiment. Each modification can also be applied to the following embodiments.

[0080] The second embodiment detects abnormal product states by statistical processing of state data, but is not limited thereto. For example, the abnormality detection unit 224 may detect abnormal states of the first product based on first state data using a pre-trained machine learning model, or it may detect abnormal states of the second product based on second state data using a machine learning model. In this case, in addition to the effects of the second embodiment, abnormal states can be detected by the machine learning model even in situations where the number of state data is small and unsuitable for statistical processing.

[0081] Furthermore, in the second embodiment, data used for product quality control (product dimensions, voltage, resistance, and other electrical characteristics) was used as state data, but the embodiment is not limited to this. That is, flag information, which is the result of product inspection, may be used as state data. In this case, a first or second condition may be specified based on the flag information. For example, if flag information "1" represents an abnormal state and flag information "0" represents a normal state, the specified first condition may be modified to a first condition indicating the manufacturing number of flag information "1". Similarly, the specified second condition may be modified to a second condition indicating the manufacturing number of flag information "1". According to this modification, the effects of the second embodiment can be obtained without using statistical processing, machine learning models, rule-based systems, etc.

[0082] (Third embodiment) Next, a third embodiment will be described. Compared to the first embodiment, the data analysis device according to the third embodiment outputs similarity and a second condition as data analysis results. As a result, the data analysis device presents the similarity and the second condition to the user via the output destination device.

[0083] Figure 15 is a block diagram showing the configuration of a data analysis device according to the third embodiment. This data analysis device 200 further includes an output unit 260 and a defective database 270 compared to the configuration shown in Figure 1. The output unit 260 is connected to a display device 300.

[0084] Here, the output unit 260 obtains the calculated similarity score Si and outputs the similarity score Si and the second condition i to the display device 300. The output unit 260 may also receive information about the first condition from the first condition specification unit 210 and information about the second condition from the second condition specification unit 220. The output unit 260 may also output the first condition, the second condition i and the similarity score Si to the display device 300. The output unit 260 may also obtain information related to the second condition i from the defective database 270 and output the obtained information, the similarity score Si and the second condition i to the display device 300.

[0085] The defective product database 270 is a storage device that stores information about defective products. The information about defective products includes, but is not limited to, the following information (Ia) to (Ic).

[0086] (Ia) The serial number, manufacturing date, manufacturing lot, and other manufacturing conditions of the defective product.

[0087] (Ib) Information regarding defective products (defect occurrence status, cause of defect, response to defective products).

[0088] (Ic) Links to defect reports (Word, PDF).

[0089] The defective database 270 stores, for example, the management number, manufacturing number, manufacturing date, cause of defect, and a link to the report for each defective product in its respective columns, as shown in Figure 16. Each row in the defective database 270 records information about the defective product for each product group. The manufacturing number and manufacturing date correspond to a second condition indicating the second product for comparison. Accordingly, the output unit 260 uses the second condition i as a query and can retrieve information about defective products that match the query.

[0090] The display device 300 is a display that shows the similarity score Si and the second condition i output from the output unit 260. In other words, the display device 300 presents the user with the second condition i, which corresponds to manufacturing conditions similar to the manufacturing conditions corresponding to the first condition.

[0091] The other configurations are the same as in the first embodiment.

[0092] Next, the operation of the data analysis device configured as described above will be explained using the flowchart in Figure 17 and the schematic diagram in Figure 18.

[0093] Now, as described above, when steps ST10 to ST50 are executed and the first and second conditions i are specified, the similarity score Si is calculated after each process.

[0094] (Step ST60) The output unit 260 receives information regarding the first condition from the first condition specification unit 210 and information regarding the second condition i from the second condition specification unit 220. Based on the second condition i, the output unit 260 searches the defective database 270 and obtains information regarding defective products related to the second condition i. Thereafter, the output unit 260 outputs the information regarding the first condition, the second condition i, the similarity score Si, and the defective product to the display device 300. However, the first condition and the information regarding the defective product do not necessarily have to be output.

[0095] (Step ST70) The display device 300 displays the analysis results based on the output of the output unit 260. For example, as shown in Figure 18, the display device 300 displays the analysis unit, the similarity score Si, and related information in association. In the example shown in Figure 18, the date of the analysis unit is the second condition indicating the second product to be compared, and is the manufacturing date when the second product is indicated by the second condition specifying the manufacturing date (factor data). The manufacturing date of the second product corresponds to the manufacturing date of the defective product in the defective database 270, in addition to the manufacturing date included in the factor data. Link information near the date of the analysis unit is linked to the factor data 100D, and when selected, the screen transitions to the factor data 100D. Related information is information about the defective product linked to the second condition i, and corresponds to the cause of defect occurrence in the defective database 270. Link information near the related information is linked to the defective database 270, and when selected, the screen transitions to the defective database 270. The display device 300 also displays the date of the search query and the pagination pgn. In the example shown in Figure 18, the date in the search query is the first condition that identifies the first product to be analyzed, and it is the manufacturing date when the first product is identified by the first condition that specifies the manufacturing date (factor data). The pagination (pgn) has multiple page buttons to divide and display the area associated with the analysis unit, similarity, and related information on each page.

[0096] The display mode of such a display device 300 may be changed according to the similarity Si of the products specified in the second condition i. For example, the display device 300 may display each data element, such as the analysis unit, similarity, and related information, in order of similarity Si. Alternatively, the display device 300 may sort and display each data element in descending or ascending order of its similarity Si. The display device 300 may also highlight data elements with similar similarity Si. For example, the highlighting can be done by enlarging the font size, using bold text, coloring, etc., as appropriate. Alternatively, the display device 300 may change the display color according to the magnitude of the similarity Si. For example, a gradient may be used in which the color becomes closer to red as the similarity Si is greater, and closer to blue as the similarity Si is smaller. In Figure 18, the shading of the similarity represents this gradient. Also, for example, the display device 300 does not have to display each data element with similarity Si below a threshold. In this case, the display and hiding of each data element with similarity Si below a threshold may be changed, for example, by user operation. For example, each data element with similarity below a predetermined threshold may be displayed after transitioning to the next page by operating the pagination function pgn. Also, when the display device 300 receives defect information from the defect database 270 from the output unit 260, the display device 300 may display each data element and the defect information side by side. In any case, the display device 300 displays the analysis results from the data analysis device 200. The user identifies the cause of the anomaly by visually inspecting the displayed analysis results.

[0097] As described above, according to the third embodiment, the output unit 260 obtains the calculated similarity score Si and outputs the similarity score Si and the second condition i. This allows the user to be presented with the similarity score Si and the second condition i, in addition to the effects described above.

[0098] Furthermore, according to the third embodiment, the output unit 260 may acquire information related to the second condition i, and may output the acquired information, the similarity Si, and the second condition i. In this case, in addition to the effects described above, information related to the second condition i can be presented to the user.

[0099] (Modified version of the third embodiment) Next, we will describe various modifications of the third embodiment. Each modification can also be applied to the following embodiments.

[0100] The third embodiment has been described as a modification of the first embodiment, but is not limited thereto. For example, the data analysis device 200 may be a modification of the second embodiment, as shown in Figure 19. This data analysis device 200 further includes an output unit 260 and a defective database 270 compared to the configuration shown in Figure 15. This output unit 260 is connected to a display device 300. Here, the configuration of the output unit 260, the defective database 270, and the display device 300 is the same as in the third embodiment. Other configurations are the same as in the second embodiment. Therefore, the effects and advantages of the second and third embodiments can be obtained by this modification.

[0101] (Fourth embodiment) Figure 20 is a block diagram illustrating the hardware configuration of a data analysis device according to the fourth embodiment. The fourth embodiment is a specific example of the first to third embodiments, in which the data analysis device 200 is implemented using a computer.

[0102] The data analysis device 200 includes, as hardware, a CPU (Central Processing Unit) 201, RAM (Random Access Memory) 202, program memory 203, auxiliary storage device 204, and input / output interface 205. The CPU 201 communicates with the RAM 202, program memory 203, auxiliary storage device 204, and input / output interface 205 via a bus. In other words, the data analysis device 200 of this embodiment is realized by a computer with such hardware configuration.

[0103] The CPU 201 is an example of a general-purpose processor. The RAM 202 is used by the CPU 201 as working memory. The RAM 202 includes volatile memory such as SDRAM (Synchronous Dynamic Random Access Memory). The program memory 203 stores a data analysis program for realizing each part according to each embodiment. This data analysis program may be, for example, a program for the computer to realize the functions of the first condition specification unit 210, the second condition specification unit 220, the state acquisition unit 222, the anomaly detection unit 224, the factor acquisition unit 230, the calculation unit 240, the similarity calculation unit 250, and the output unit 260. In addition, the program memory 203 may be, for example, ROM (Read-Only Memory), a part of the auxiliary storage device 204, or a combination thereof. The auxiliary storage device 204 stores data non-temporarily. The auxiliary storage device 204 includes non-volatile memory such as an HDD (hard disk drive) or SSD (solid state drive).

[0104] The input / output interface 205 is an interface for connecting to other devices. For example, the input / output interface 205 is used to connect to a keyboard, mouse, database, and display.

[0105] The data analysis program stored in program memory 203 includes computer-executable instructions. When the data analysis program (computer-executable instructions) is executed by the processing circuit, the CPU 201, it causes the CPU 201 to perform predetermined processing. For example, when the data analysis program is executed by the CPU 201, it causes the CPU 201 to perform a series of processing as described with respect to each part in Figures 1, 5, 12, 13, 15, or 19. For example, when the computer-executable instructions included in the data analysis program are executed by the CPU 201, they cause the CPU 201 to perform a data analysis method. The data analysis method may include steps corresponding to each of the functions of the first condition specification unit 210, the second condition specification unit 220, the state acquisition unit 222, the anomaly detection unit 224, the factor acquisition unit 230, the calculation unit 240, the similarity calculation unit 250, and the output unit 260 described above. The data analysis method may also include the steps shown in Figures 4, 14, or 17 as appropriate.

[0106] The data analysis program may be provided to the data analysis device 200, which is a computer, in a state where it is stored on a computer-readable storage medium. In this case, for example, the data analysis device 200 further includes a drive (not shown) for reading data from the storage medium and retrieving the data analysis program from the storage medium. As the storage medium, for example, magnetic disks, optical disks (CD-ROM, CD-R, DVD-ROM, DVD-R, etc.), magneto-optical disks (MO, etc.), semiconductor memory, etc. can be used as appropriate. The storage medium may also be called a non-transitory computer-readable storage medium. Alternatively, the data analysis program may be stored on a server on a communication network, and the data analysis device 200 may download the data analysis program from the server using the input / output interface 205.

[0107] The processing circuit that executes the data analysis program is not limited to a general-purpose hardware processor such as the CPU 201, but may also use a dedicated hardware processor such as an ASIC (Application Specific Integrated Circuit). The term "processing circuit" (processing unit) includes at least one general-purpose hardware processor, at least one dedicated hardware processor, or a combination of at least one general-purpose hardware processor and at least one dedicated hardware processor. In the example shown in Figure 20, the CPU 201, RAM 202, and program memory 203 correspond to the processing circuit.

[0108] According to at least one embodiment described above, the accuracy of estimating the cause of an abnormality in the manufacturing process can be improved.

[0109] While several embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These embodiments can be carried out in a variety of other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims and their equivalents. [Explanation of symbols]

[0110] 100...Manufacturing database, 100D...Factor data, 100S...Status data, 200...Data analysis device, 203...Program memory, 204...Auxiliary storage device, 205...Input / output interface, 210...First condition specification unit, 220...Second condition specification unit, 222...Status acquisition unit, 224...Anomaly detection unit, 230...Factor acquisition unit, 232...Storage unit, 240...Calculation unit, 250...Similarity calculation unit, 260...Output unit, 270...Defect database, 300...Display device.

Claims

1. A first condition specification unit that specifies the first condition indicating the first product to be analyzed, A second condition specification unit that specifies a second condition indicating the second product to be compared, A factor acquisition unit that acquires first factor data indicating a plurality of first manufacturing conditions for the first product based on the first condition, and acquires second factor data indicating a plurality of second manufacturing conditions for the second product based on the second condition, A calculation unit calculates a first index value related to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of abnormality in the first product, based on the first factor data and statistical hypothesis testing, and calculates a second index value related to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of abnormality in the second product, based on the second factor data and statistical hypothesis testing. A similarity calculation unit that calculates the similarity between the first index value and the second index value, A data analysis device equipped with the following features.

2. A first condition designation unit that designates a first condition indicating a first product to be analyzed, A second condition specification unit that specifies a second condition indicating the second product to be compared, A state acquisition unit that acquires first state data indicating the state of the first product based on the first condition, and acquires second state data indicating the state of the second product based on the second condition, An abnormality detection unit that detects an abnormal state of the first product based on the first state data and modifies the first condition to indicate the first product in the detected abnormal state, and detects an abnormal state of the second product based on the second state data and modifies the second condition to indicate the second product in the detected abnormal state, A factor acquisition unit acquires first factor data indicating a plurality of first manufacturing conditions for the first product based on the modified first condition, and acquires second factor data indicating a plurality of second manufacturing conditions for the second product based on the modified second condition. A calculation unit calculates a first index value related to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of abnormality in the first product, based on the first factor data, and calculates a second index value related to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of abnormality in the second product, based on the second factor data. A similarity calculation unit that calculates the similarity between the first index value and the second index value, A data analysis device equipped with the following features.

3. The second condition designation unit designates one or more second conditions that are different from the first condition, The factor acquisition unit acquires the second factor data for each of the second conditions, The calculation unit calculates the second indicator value for each of the second factor data, The similarity calculation unit calculates the similarity for each of the second index values. The data analysis apparatus according to claim 1 or 2.

4. The calculation unit calculates the first index value and the second index value using a trained model that has been trained to output index values ​​based on the input factor data. The data analysis apparatus according to claim 1 or 2.

5. The data analysis device according to claim 2, wherein the abnormality detection unit detects an abnormal state of the first product by statistical processing based on the first state data, and detects an abnormal state of the second product by statistical processing based on the second state data.

6. The data analysis device according to claim 2, wherein the anomaly detection unit detects an abnormal state of the first product based on the first state data using a pre-trained machine learning model, and detects an abnormal state of the second product based on the second state data using the machine learning model.

7. A storage unit in which the second condition and the second index value are stored in association with each other, The calculation unit obtains the second index value from the storage unit based on the second condition. The data analysis apparatus according to claim 1 or 2.

8. An output unit that obtains the calculated similarity and outputs the similarity and the second condition. A data analysis apparatus according to claim 1 or 2, further comprising the above.

9. The data analysis apparatus according to claim 8, wherein the output unit acquires information related to the second condition and outputs the acquired information, the similarity, and the second condition.

10. The data analysis apparatus according to claim 1, wherein the statistical hypothesis test is a G-test.

11. The data analysis apparatus according to claim 1, wherein the statistical hypothesis test is a chi-squared test.

12. Specify the first condition that indicates the first product to be analyzed, Specify a second condition that indicates the second product to be compared, Based on the first condition, first factor data representing multiple first manufacturing conditions for the first product is obtained, Based on the second condition, obtain second factor data indicating multiple second manufacturing conditions for the second product, Based on the first factor data and statistical hypothesis testing, a first index value is calculated that relates to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of the abnormality of the first product. Based on the second factor data and the statistical hypothesis test, a second index value is calculated that relates to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of the abnormality of the second product. The similarity between the first indicator value and the second indicator value is calculated, A data analysis method equipped with [specific features / features].

13. Specify the first condition that indicates the first product to be analyzed, Specify a second condition that indicates the second product to be compared, Based on the first condition, first state data indicating the state of the first product is obtained, Based on the second condition, second state data indicating the state of the second product is obtained, Based on the first state data, an abnormal state of the first product is detected, and the first condition is modified to indicate the first product in the detected abnormal state. Based on the second state data, an abnormal state of the second product is detected, and the second condition is modified to indicate the second product in the detected abnormal state. Based on the modified first condition, first factor data representing multiple first manufacturing conditions for the first product is obtained, Based on the modified second condition, obtain second factor data indicating multiple second manufacturing conditions for the second product, Based on the first factor data, a first index value is calculated that relates to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of the abnormality of the first product. Based on the second factor data, a second index value is calculated that relates to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of the abnormality of the second product. The similarity between the first indicator value and the second indicator value is calculated, A data analysis method equipped with [specific features / features].

14. Computers, Means for specifying a first condition indicating the first product to be analyzed, Means for specifying a second condition indicating a second product for comparison, Means for obtaining first factor data indicating a plurality of first manufacturing conditions for the first product based on the first condition, and for obtaining second factor data indicating a plurality of second manufacturing conditions for the second product based on the second condition, A means for calculating a first index value related to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of abnormality in the first product, based on the first factor data and statistical hypothesis testing, and for calculating a second index value related to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of abnormality in the second product, based on the second factor data and statistical hypothesis testing. Means for calculating the similarity between the first index value and the second index value, A data analysis program designed to function as such.

15. Computers, Means for specifying a first condition indicating the first product to be analyzed, Means for specifying a second condition indicating a second product for comparison, Means for acquiring first state data indicating the state of the first product based on the first condition, and for acquiring second state data indicating the state of the second product based on the second condition, Means for detecting an abnormal state of the first product based on the first state data, modifying the first condition to indicate the first product in the detected abnormal state, detecting an abnormal state of the second product based on the second state data, and modifying the second condition to indicate the second product in the detected abnormal state, Means for obtaining first factor data indicating a plurality of first manufacturing conditions for the first product based on the modified first condition, and for obtaining second factor data indicating a plurality of second manufacturing conditions for the second product based on the modified second condition, A means for calculating a first index value related to the degree to which each of the plurality of first manufacturing conditions contributes to the cause of abnormality in the first product, based on the first factor data, and for calculating a second index value related to the degree to which each of the plurality of second manufacturing conditions contributes to the cause of abnormality in the second product, based on the second factor data. Means for calculating the similarity between the first index value and the second index value, A data analysis program designed to function as such.

Images