JP2011145764A - Operation analysis device, operation analysis method, computer program and computer-readable storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an operation analysis device allowing efficient finding of a range of a value of an operation factor wherein variation of quality data becomes large in a process wherein the variation of the quality data changes by the operation factor. <P>SOLUTION: The operation analysis device includes: a data input part inputted with operation factor data and the quality data; an operation factor space division part setting a plurality of division candidate points wherein an operation factor space is divided; a probability density function calculation part calculating a variation index in a local area by use of a probability density function using the quality data as a probability variable about each local area of the operation factor space divided in the division candidate points; a maximum variation model selection part determining the division candidate point for generating the local area wherein the variation index calculated about each local area becomes maximum as a decision division point, and performing decision division of the operation factor space; a redivision target area decision part deciding necessity/unnecessity of redivision about the local area of the operation factor space performed with the decision division; and an analysis result display part displaying an analysis result. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an operation analysis apparatus, an operation analysis method, a computer program, and a computer-readable storage medium. In particular, in the entire process in which quality is determined as an operation result, the relationship between a plurality of operation factors and quality variations is clarified. The present invention relates to an operation analysis apparatus, an operation analysis method, a computer program, and a computer-readable storage medium that are used to efficiently find manufacturing conditions and operation conditions that require operation improvement.

  In a process in which quality is determined as an operation result, clarifying the relationship between a plurality of operation factors and quality variations is an effective means for efficiently finding manufacturing conditions and operation conditions that require improvement. Variation in quality occurs depending on manufacturing conditions and operating conditions, but it is highly region dependent, and extracting in which region of the operating factor space the variation occurs leads to efficient operational improvement.

  As an operation analysis method considering an area generated by dividing the operation factor space, methods such as Patent Documents 1 and 2 have been proposed. Patent Document 1 discloses an operation analysis device that divides an operation factor space, constructs a relational expression between operation data and quality data in each local region, and clarifies the relationship between the operation data and the quality data. Yes. In Patent Document 2, the operation factor space is divided according to the given stratification, the probability density function of the quality data in each local region is obtained, the quality index based on the probability density is calculated, and the operation factor is influenced by the quality influence. An operation analysis device that makes it easier to extract an operation factor having a large influence on a quality index by arranging them in descending order is disclosed.

JP 2009-151383 A JP 2008-146621 A

  However, the methods described in Patent Documents 1 and 2 are operation analysis methods for knowing the causal relationship between the operation data and the quality data, such as what kind of quality result is obtained with respect to the operation condition. It was not a technique to know the variation itself, such as in which area the variation in quality was large.

  If the method of dividing the operation factor is given, the magnitude of the quality variation can be evaluated using a general statistical test method such as a test of equal variance. Conventionally, based on operation knowledge and the like, manufacturing conditions and operation conditions are stratified, graphed, and a total test method is used to extract areas with large variations in quality.

  However, if the number of manufacturing conditions and operating conditions increases and the stratification of each operating factor becomes finer in order to extract various varieties, the method of dividing the region becomes enormous due to the combination thereof. For this reason, an analysis method that relies on human labor is time consuming and trial and error, so that an appropriate result is not always obtained by a person. Therefore, an apparatus that systematically analyzes the magnitude of variation depending on the operation region and extracts the region with the largest variation is desired.

  Therefore, the present invention has been made in view of the above problems, and the object of the present invention is to provide quality that requires operational improvements in a process in which variations in quality data vary depending on operational factors. It is an object of the present invention to provide a new and improved operation analysis apparatus, operation analysis method, computer program, and computer-readable storage medium capable of efficiently finding out the range of values of operation factors in which data variation increases.

  In order to solve the above problems, according to an aspect of the present invention, there is provided an operation analysis apparatus that analyzes variations in quality data in a manufacturing process. Such an operation analysis device includes a data input unit for inputting operation factor data and quality data, and an operation factor space division for setting a plurality of division candidate points for dividing an operation factor space composed of the operation factor data into a plurality of local regions. The probability of calculating a variation index that represents the degree of variation in quality data in the local region using a probability density function with quality data as a random variable for each local region in the operation factor space divided by the division candidate point The density function calculation unit and the maximum variation model selection that determines the candidate division point that generates the local region that maximizes the variation index calculated for each local region as the deterministic division point, and deterministically divides the operating factor space by the definite division point The local region of the operation factor space that has been definitely divided at the deterministic division point by the maximum variation model selection unit A subdivision target region determination unit that determines necessity and an analysis result display unit that displays an analysis result indicating a variation in quality data of a local region of the operation factor space. For the local area determined to be necessary, division candidate point setting by the operation factor space division unit, variation index calculation by the probability density function calculation unit, fixed division by the maximum variation model selection unit, and subdivision target region determination unit It is characterized by repeating the determination of whether or not re-division is necessary.

  According to the present invention, the operation factor space is divided by each operation factor, and the probability density function is applied to each divided local region to evaluate the quality variation. As a result, a region with a large variation in quality is cut out, and a range of manufacturing conditions and operation conditions with a large variation in quality can be found efficiently.

  Here, the operation factor space dividing unit may exclude, from the division target, a local region whose variation index is equal to or greater than a predetermined threshold when the local region is subdivided.

  Further, the variation index can be represented by, for example, the difference between the probability variable value when the cumulative probability of the probability density function is a preset value and the mode value of the probability distribution.

  Further, the operation factor space dividing unit may divide the operation factor space into two local regions at the division candidate points.

  Further, the probability density function calculation unit may exclude the division candidate point from the candidates when the number of data points belonging to at least one of the local areas divided by the division candidate points is less than the reference value. Good.

  Further, the subdivision target area determination unit may end the division process when the number of divisions of the operation factor space reaches a predetermined number.

  Further, the operation factor space dividing unit excludes, from the candidates, division candidate points that are outside the region of the local region to be divided among the division candidate points set by the operation factor space dividing unit when the local region is subdivided. May be.

  Moreover, in order to solve the said subject, according to another viewpoint of this invention, the operation analysis method which analyzes the dispersion | variation in quality data in a manufacturing process is provided. The operation analysis method includes an input step in which operation factor data and quality data are input, and a division candidate point setting step in which a plurality of division candidate points are set, each of which divides an operation factor space composed of operation factor data into a plurality of local regions. And, for each local region of the operation factor space divided by the division candidate points, using a probability density function, a variation index calculating step for calculating a variation index representing the degree of variation of quality data in the local region, and each local region A deterministic dividing point determining step for determining a dividing candidate point that generates a local region that maximizes the variation index calculated with respect to as a deterministic dividing point; A subdivision judgment that determines whether or not subdivision is necessary for the local region of the operation factor space that is deterministically divided at points. And an analysis result display step for displaying an analysis result indicating a variation in quality data of the local area of the operation factor space, and for the local area determined to be subdivided by the subdivision determination step, The division candidate point setting step, the variation index calculation step, the fixed division point determination step, the final division step, and the subdivision determination step are repeated.

  Furthermore, in order to solve the above-described problem, according to another aspect of the present invention, a computer includes a plurality of operation factor spaces each including a data input means for inputting operation factor data and quality data, and an operation factor data. Variation of quality data in the local area using probability density function for each local area of the operating factor space divided by the candidate area and operation factor space dividing means for setting multiple candidate points to be divided into local areas A probability density function calculating means for calculating a variation index that represents the degree of divergence, and a division candidate point that generates a local region that maximizes the variation index calculated for each local region is determined as a fixed division point, and operation is performed using the fixed division point. The maximum variation model selection means for deterministic division of the factor space and the definite division at the definite division point by the maximum variation model selection means Re-division target area determination means for determining whether or not re-division is necessary for the local area of the industry factor space, and analysis result display means for displaying the analysis result indicating the variation in the quality data of the local area of the operation factor space. , With respect to a local area determined to be subdivided by the subdivision target area determination means, division candidate point setting by the operation factor space division means, variation index calculation by the probability density function calculation means, and maximum variation model selection means A computer program for functioning as an operation analysis apparatus is provided, characterized by repeating the definite division according to the above and the re-division target area determination means to determine whether re-division is necessary.

  In order to solve the above-described problem, according to another aspect of the present invention, a computer includes data input means for inputting operation factor data and quality data, and a plurality of operation factor spaces each including operation factor data. Variation of quality data in the local area using probability density function for each local area of the operating factor space divided by the candidate area and operation factor space dividing means for setting multiple candidate points to be divided into local areas A probability density function calculating means for calculating a variation index that represents the degree of divergence, and a division candidate point that generates a local region that maximizes the variation index calculated for each local region is determined as a fixed division point, and operation is performed using the fixed division point. Maximum variation model selection means for deterministic division of the factor space, and operations that are definitely divided at the definite division points by the maximum variation model selection means Re-division target area determination means for determining whether or not re-division is necessary for the local area of the factor space, and analysis result display means for displaying an analysis result indicating variation in quality data of the local area of the operation factor space, For the local area determined to be subdivided by the subdivision target area determination means, the division candidate point setting by the operation factor space division means, the variation index calculation by the probability density function calculation means, and the maximum variation model selection means There is provided a computer-readable recording medium on which a program for functioning as an operation analysis device is recorded, characterized by repeating the definite division and the necessity of re-division by the re-division target area determination means.

  As described above, according to the present invention, in the process in which the variation in the quality data varies depending on the operation factor, the range of the value of the operation factor that requires the operation improvement and in which the variation in the quality data becomes large can be efficiently obtained. An operation analysis device, an operation analysis method, a computer program, and a computer-readable storage medium that can be found can be provided.

It is a functional block diagram which shows the structure of the operation analysis apparatus which concerns on embodiment of this invention. It is a flowchart which shows the analysis method by the operation analyzer which concerns on the same embodiment. It is explanatory drawing which shows the area | region division | segmentation determination method of the operation factor space. It is explanatory drawing which shows the area | region subdivision method of the operation factor space. It is explanatory drawing which shows the outline of a hot rolling line. It is explanatory drawing which shows the wave height h of material. It is explanatory drawing which shows the decision tree of the division | segmentation execution result of the operation factor space by a board flatness parameter | index. It is a table | surface which shows the division | segmentation execution result (area | region 1-3) of the operation factor space by a board flatness parameter | index. It is a table | surface which shows the division | segmentation execution result (area | region 4-6) of the operation factor space by a board flatness parameter | index. It is a table | surface which shows the division | segmentation execution result (area | region 7-10) of the operation factor space by a board flatness parameter | index.

  Exemplary embodiments of the present invention will be described below in detail with reference to the accompanying drawings. In addition, in this specification and drawing, about the component which has the substantially same function structure, duplication description is abbreviate | omitted by attaching | subjecting the same code | symbol.

[1. Overview of operation analysis method using operation analyzer]
First, the outline | summary of the operation analysis method by the operation analysis apparatus which concerns on this embodiment is demonstrated. The operation analysis apparatus according to the present embodiment, from a large amount of operation data under various production conditions / operation conditions, has large variations in quality, and manufacturing conditions / operation conditions (range of operation factors) that need improvement. In order to extract, it is a device that performs space division by each operating factor and finds manufacturing conditions and operating conditions (range of operating factors) that need improvement. At this time, focus on the probability distribution instead of the representative value such as the expected value, median value, and mode value of the quality data in the divided area, and select the divided area where the dispersion becomes large with the variation index as the evaluation criterion. Repeat this step. As a result, the operation analysis apparatus according to the present embodiment can accurately extract a region having a large variation without overlooking a variation in quality that hardly affects the representative value of the quality data. Hereinafter, the configuration of the operation analysis apparatus and the operation analysis method based on the configuration will be described in detail.

[2. Configuration of Operation Analyzer]
First, based on FIG. 1, the structure of the operation analysis apparatus 100 which concerns on embodiment of this invention is demonstrated. FIG. 1 is a functional block diagram showing the configuration of the operation analysis apparatus 100 according to the present embodiment.

  As shown in FIG. 1, the operation analysis apparatus 100 according to the present embodiment includes a data input unit 101, an operation factor space division unit 102, a probability density function calculation unit 103, a maximum variation model selection unit 104, and a subdivision. A target area determination unit 105 and an analysis result display unit 106 are provided.

  The data input unit 101 receives operation factor data and quality data to be analyzed from a data storage unit (not shown). The data input unit 101 outputs the input operation factor data and quality data to the operation factor space division unit 102.

  The operation factor space division unit 102 sets division candidate points for each operation factor for the operation factor data input from the data input unit 101. The operation factor space dividing unit 102 sets the operation factor division candidate points as, for example, an intermediate value of values that can be taken from discrete data in discrete data, a point at which data is divided at regular intervals, or an interval of a constant number of data. It can be determined as a point to be divided by. Moreover, a division | segmentation candidate point can also be determined by operation knowledge, and can also be determined by these combination. Further, the operation factor space division unit 102 determines whether the division candidate point is within the local region to be divided, and excludes division points that are not targeted for calculation of a probability density function, which will be described later, from the division candidate points. When the candidate division points are determined, the operation factor space dividing unit 102 causes the probability density function calculating unit 103 to calculate the probability density function of each region divided by the candidate division points.

  The probability density function calculation unit 103 divides the operation factor space into a plurality of local regions at each division candidate point of each operation factor, sets a plurality of local regions, and generates probability for the quality data to which the operation factor value belongs to each local region. Fit density function. This embodiment demonstrates the case where an operation factor space is divided into two. Then, the probability density function calculation unit 103 calculates, for each local region, a variation index representing quality variation in the local region based on the probability density function. Thereby, it becomes possible to grasp the quality variation in each local region.

  Based on the variation index of each local region calculated by the probability density function calculation unit 103, the maximum variation model selection unit 104 employs a division candidate point that generates a local region having the maximum variation index as a fixed division point. Then, the maximum variation model selection unit 104 performs an operation factor space division process at the adopted fixed division point.

  The subdivision target area determination unit 105 determines whether or not it is necessary to execute subdivision processing for further dividing the two local areas obtained by the definite division processing of the maximum variation model selection unit 104 into local areas. . The necessity of executing the re-division processing is, for example, the region when the variation index in the local region is smaller than the threshold value, or when the region division depth (the number of times the region is divided into local regions) is equal to or less than a predetermined value. Is subdivided based on a subdivision condition such as subdivision. The subdivision target area determination unit 105 instructs the operation factor space division unit 102 to perform further area division when the subdivision condition is satisfied. On the other hand, when the subdivision condition is not satisfied, the subdivision target area determination unit 105 ends the division process and outputs the analysis result to the analysis result display unit 106.

  The analysis result display unit 106 displays the analysis result output from the subdivision target region determination unit 105. The analysis result display unit 106 displays the analysis results such as the range of operation factors for each local region of the operation factor space, the number of data, the variation index, and the histogram, which are finally obtained, by a display device such as a display (not shown). ).

[3. Operation analysis method]
Next, the operation analysis method by the operation analysis apparatus 100 mentioned above is demonstrated based on FIGS. FIG. 2 is a flowchart showing an analysis method performed by the operation analysis apparatus 100 according to the present embodiment. FIG. 3 is an explanatory diagram showing a region division determination method for the operation factor space. FIG. 4 is an explanatory diagram showing a region subdivision method for the operation factor space.

As shown in FIG. 2, the process performed by the operation analysis apparatus 100 according to the present embodiment is started when the operation factor data and the quality data are first input to the data input unit 101 (step S100). The operation factor data is composed of a plurality (p pieces) of data (u ₁ , u ₂ ,..., U _p ), which may be either continuous values or discrete values. In the following, p operation factor data are collectively expressed as a p-dimensional vector u = (u ₁ , u ₂ ,..., U _p ). It is assumed that N pieces of operation factor data u ₁ , u ₂ ,..., U _N are input to the data input unit 101 as a whole. On the other hand, the quality data y is acquired in correspondence with N pieces of operation factor data u ₁ , u ₂ ,..., U _N (y ₁ , y ₂ ,..., Y _N ). May be either a continuous value or a discrete value. The quality data y is regarded as a random variable, and the probability density function is determined depending on the region of the space due to the operation factor.

When the operation factor data and the quality data are input to the data input unit 101, the operation factor space dividing unit 102 sets each operation factor (u ₁ , u ₂ ,..., U _p ) constituting the operation factor data u. A division candidate point is created (step S102). Candidate points for division can be determined based on operational knowledge, for example. In the case of discrete data, it can be set to an intermediate value of possible values of data, set to a value that divides data at regular intervals, or constant The value can be set to a value that divides data at an interval of the number of data, or can be determined by a combination thereof. If there are m ₁ , m ₂ ,..., M _p division candidate points for each operation factor (u ₁ , u ₂ ,..., U _p ), the number of all division candidate points is as follows. It is represented by Formula 1.

・・・（数式１）

... (Formula 1)

As an example, consider a case where division candidate points are set in an operation factor space composed of _two operation factors (u ₁ , u ₂ ). When division candidate points are set using a method such as setting an intermediate value between operation knowledge and possible values, for example, as shown in FIG. 3, three division candidate points u ₁₁ , u ₁₂ , u for the operation factor u ₁ are used. ₁₃ , it is assumed that two candidate division points u ₂₁ and u ₂₂ are set for the operation factor u ₂ . Accordingly, the total number of division candidate points set in the operation factor space is five.

When division candidate points are set by the operation factor space division unit 102, the probability density function calculation unit 103 divides the operation factor space into two at each division candidate point of each operation factor (step S108). In addition, about the process of step S104 and S106, it is a process performed when the division | segmentation of the operation factor space is the 2nd time or later, and these processes are mentioned later (paragraphs 0059-0063). The division performed in step S108 is a temporary division for determining a formal division point (a definite division point). In the example of FIG. 3, the probability density function calculation unit 103 divides the operation factor space into two at the division candidate points u ₁₁ , u ₁₂ , u ₁₃ , u ₂₁ , u ₂₂ , and the division candidates 1-1 to 1- 1 5 is created.

  Next, the probability density function calculation unit 103 determines, for each of the division candidates 1-1 to 1-5, whether or not the number of data points belonging to one of the two local regions obtained by division is equal to or greater than a reference value. (Step S110). When the number of data points belonging to the local region is smaller than the reference value, the accuracy is low even if the probability density function is applied. Therefore, in step S110, a process for removing the division candidates that are not applied with the probability density function is performed. The reference value is preferably set to a number of about 10 to 100, for example, and may be set to a value of about several percent of the number of data when the total number of data is large.

  For the division candidates determined in step S110 that the number of data points belonging to the local region is smaller than the reference value, the probability density function calculation unit 103 performs the process of step S118 without applying the probability density function. On the other hand, the probability density function calculation unit 103 performs probability density function fitting for the division candidates determined that the number of data points belonging to the local region is equal to or greater than the reference value (step S112).

  In step S112, for each division candidate, the probability density function f (y; β) is applied to the quality data y to which the operation factor vector belongs to each of the two divided local regions. Here, β is a parameter of the probability density function f (hereinafter also referred to as “probability parameter”). β is composed of one or a plurality of probability parameters by a probability density function, and these are collectively expressed here.

  Examples of probability density functions include normal distribution, lognormal distribution, exponential distribution, and the like. For example, the probability density function of the normal distribution is expressed by the following formula 2.

・・・（数式２）

... (Formula 2)

  In this case, there are two probability parameters of the probability density function f: average μ and standard deviation σ. Therefore, these two probability parameters are collectively expressed as β = (μ, σ). Exp represents an exponential function.

  Further, for example, the probability density function of the lognormal distribution is expressed by the following Equation 3.

・・・（数式３）

... (Formula 3)

  In this case, there are two probability parameters of the probability density function f: logarithmic mean μ ′ and logarithmic standard deviation σ ′. Therefore, these two probability parameters are collectively expressed as β = (μ ′, σ ′). Ln represents a natural logarithm.

  Further, for example, the probability density function of the exponential distribution is expressed by the following mathematical formula 4.

・・・（数式４）

... (Formula 4)

  In this case, the probability parameter of the probability density function f is λ. Although there is one probability parameter, it is expressed as β = λ as in other probability density functions.

In this embodiment, a logarithmic normal distribution probability density function is used. The probability parameter β of the probability density function is the quality data {y _i | u _i εM corresponding to the operation factor vector u _i = (u _i1 , u _i2 ,..., U _ip ) belonging to the target region M _{j. j} } can be determined by a statistical probability technique. The statistical probability method includes, for example, a maximum likelihood method. Since the probability parameter β is determined for each region M _j , it can be expressed as β _j .

When the probability density function is applied to the local region, the probability density function calculation unit 103 calculates a variation index for each local region, and the maximum variation model selection unit 104 evaluates the variation index for each local region. (Step S114). The variation index is an index indicating the degree of variation of the quality data included in the region, and in the present embodiment, “the value of the cumulative probability 90% −the mode of the probability distribution” (which is obtained from the probability density function) (Also referred to as “FM value”) as a variation index. For example, the difference between the value y _A of the cumulative probability 90% of the probability density function shown in FIG. 3 and the mode y _{B of the} probability distribution is the FM value (variation index) V.

  In the present embodiment, “value of cumulative probability 90% −moderate value of probability distribution” is defined as FM value, but as the dispersion index, the cumulative probability is set between 70% and 95%. May be. If the cumulative probability is greater than 95%, there is a high possibility that an abnormal value is included, and it is impossible to define an appropriate range for evaluating variation in quality data after removing the abnormal value. On the other hand, if the cumulative probability is less than 70%, it is impossible to evaluate the variation of the original quality data. Thus, the difference between the cumulative probability value set in advance between 70% and 95% and the mode value of the probability distribution can be set as the variation index.

  In addition to the FM value, the variation index may be, for example, “a value of cumulative probability 90% −a value of cumulative probability 50%”, “a value of cumulative probability 80% −a value of cumulative probability 20%”, It is possible to set a probability distribution range that is set so as not to overlook a region with large variations, such as a value derived from “value of cumulative probability 90% −value of cumulative probability 10%” or the like. A standard deviation can also be used as a variation index.

  The probability density function calculation unit 103 calculates the variation index of each local region by such a method. Then, the maximum variation model selection unit 104 determines, as a definite division point, a division candidate point that generates a local region having the maximum variation index among the variation indexes calculated for each local region.

  In the present embodiment, the processes in steps S106 to S116 are performed for each division candidate. For this reason, in step S114, when the variation index for one division candidate is calculated, the division candidate point having the larger variation index is compared with the largest variation index of the division candidates calculated before that. The data at is saved as division candidate point data. At this time, division candidate points, probability parameters, FM values, and the like are stored as divided data. Then, when the determination for all the division candidate points is finished, the division candidate point having the maximum variation index is adopted as the fixed division point.

  If the variation index is smaller than the maximum value of the variation index calculated before then, the divided data at the division candidate point is not saved, and the process proceeds to step S118. On the other hand, when the variation index is equal to or greater than the maximum value of the variation index calculated before, the divided data is stored in a memory (not shown) (step S116). Then, the maximum variation model selection unit 104 determines whether or not the processing of steps S106 to S116 has been executed for all division candidate points of all operation factors (step S118). If the process of steps S106 to S116 has not been completed for all division candidate points for all the operating factors, the process returns to step S106 and the process is repeated. On the other hand, when the process of steps S106 to S116 has been completed for all the division candidate points of all the operation factors, the area division process for dividing and dividing the local area at the fixed division point is executed (step S120).

  The processes in steps S104 to S120 are performed for all areas to be divided (step S122). When the operation factor space is divided for the first time, since there is only one region to be divided, the determination process in step S122 is always YES. On the other hand, when the operation factor space is divided for the second time and thereafter, in steps S <b> 122, steps S <b> 104 to S <b> 120 are performed for each of the two local regions generated by dividing the operation factor space at the determined division point determined in the immediately preceding step. Whether or not the above process has been executed is determined.

For example, in FIG. 4, when the value u ₁₃ of the operation factor u ₁ is determined as a definite division point in the previous round, the operation factor space is definitely divided into two regions A ₁ and A ₂ . Then, when the division depth is increased by ₁ , and when the determined division point of the next region is determined, the processes of steps S104 to S120 are executed for each of the local regions A ₁ and A ₂ . In the example of FIG. 4, for the local area A _2, since distribution index is excluded from the division target because it is above the threshold, in practice and only processing in step S104 is executed.

  If it is determined in step S122 that the processes in steps S104 to S120 have not been executed for all local regions, the process returns to step S104 and the process is repeated. On the other hand, if it is determined that the processes in steps S104 to S120 have been executed for all local regions, the division depth is increased by 1 (step S124), and whether or not the region division depth has reached the maximum division depth. Is determined (step S126). Step S126 is a process of determining whether or not to end the division process shown in FIG. It is appropriate that the maximum division depth is gradually increased according to the obtained result with an initial value of about 10.

  If it is determined in step S126 that the area division depth has not reached the maximum division depth and it is determined that the division process is to be further performed, the processes after step S104 are performed. That is, when the operation factor space is divided for the second time or later (in the case of re-division of the local region), all the local regions whose region variation index is smaller than the threshold are set as the division targets (step S104). In the present embodiment, FM values are used as the variation index. When the number of data of manufacturing conditions and operating conditions that increase the quality variation is relatively small or the probability that the quality variation is large is relatively small, the quality variation is large in a wide area in the operating factor space. Data may be buried in the whole and may not appear in the variation index. For this reason, there is a possibility that a local area having a large variation in quality is included in a local area where the variation index is smaller than the threshold value. Therefore, for a local region whose variation index is smaller than the threshold value, a region having a large variation in quality is cut out by re-dividing the region.

  As the threshold value of the variation index, for example, about 1.2 to 2.0 times the variation index of the entire data can be set, and the resulting area is still large. If the conditions / operation conditions are not sufficiently narrowed down, the threshold value is increased and adjustment is performed so as to obtain a region in which the manufacturing conditions / operation conditions are further narrowed down. If it is determined in step S104 that the variation index is smaller than the threshold, the local region is set as a division target, and the process proceeds to step S106. On the other hand, if it is determined that the variation index is greater than or equal to the threshold value, the process of step S122 is executed.

For example, as shown in FIG. 4, the operating factor space is determined divided by the value u ₁₃ for operating factors u _1. Calculating a probability density function for each of the generated two local regions, when calculated variation index (F-M value), the variation index of the local area A ₁ is less than the threshold value, the variation index of the local region A ₂ is the threshold When was over, the local area a ₁ becomes subdivided subject, the division of local regions a ₂ is completed.

  In the case of subdivision of the local area, the operation factor space division unit 102 determines whether or not the division candidate points created in step S102 are set in the area of the local area that is the division target in step S104. Is determined (step S106). This is because the region is not divided even if the region is divided by the division candidate points set outside the local region to be divided, so that such division candidate points are excluded.

For example, in FIG. 4, when the local area A ₁ is subdivided, the division candidate point u ₁₃ for the operation factor u ₁ is excluded. Therefore, in step S106, candidate division points that do not exist in the local region to be divided are excluded, and the process of step S118 is executed. On the other hand, the division candidate points set in the region are directly used as the division candidate points. Accordingly, the number of initial division candidate points is M, but the number of division candidate points at the time of subdivision is smaller than that at the first time. In the example of FIG. 4, the local region A ₁ by the four candidate dividing points are subdivided, candidate dividing 2-1 to 2-4 is created.

  As described above, in the case of the re-division of the local region, after the processing of steps S104 and S106, the processing after step S108 is executed. On the other hand, when the division depth reaches the maximum division depth, the subdivision target area determination unit 105 outputs the analysis result to the analysis result display unit 106 (step S128), and ends the division process. The analysis result display unit 106 causes the display device to display, for example, the range of operation factors, the number of data, the variation index, and the histogram as the analysis results for each local region of the finally obtained operation factor space. At this time, it is possible to display only an area having a large variation in quality data or an abnormal distribution, or to display such an area with emphasis.

  The configuration of the operation analysis apparatus 100 according to the present embodiment and the operation analysis method using the same have been described above. According to the present embodiment, the operation factor space is divided by each operation factor, and the probability density function is applied to each divided local region to evaluate the quality variation. As a result, a region with a large variation in quality is cut out, and a range of manufacturing conditions and operation conditions with a large variation in quality can be found efficiently. Concentrated efforts to improve the quality of the extracted areas with large variations in quality lead to efficient operation improvement.

  In addition, when the number of data of manufacturing conditions and operating conditions that increase the quality variation is relatively small, or the probability that the quality variation is large is relatively small, the quality variation within a wide area in the operating factor space Large data is buried in the whole, and representative values such as average value and variation of quality data generally used in the past may not change so much. Even in such a case, according to the present embodiment, it is possible to efficiently find the range of values of the operation factor that requires the operation improvement and in which the variation in the quality data becomes large.

  Hereinafter, based on FIGS. 5-8C, the example of the analysis of the flatness data of the plate in the rolling mill delivery side for the hot rolling process of the steel process is shown as an example. For example, as shown in FIG. 5, the rolling line 200 rolls the slab yard 210 into which the continuously cast slab is carried in, the heating furnace 220 that heats the slab transported from the slab yard 210, and the heated slab. A rolling mill 230, a cooling device 240 that cools the rolled slab, and a cooling floor 250 that naturally cools the slab are provided. The cooled slab is transported to the finishing line 260, cut into a product size, inspected, and shipped.

  Here, if the flatness of the material conveyed to the finishing line 260 is out of specification, correction processing by the correction machine 262 is required, which adversely affects productivity and cost. Therefore, it is required to reduce the flatness of the material as much as possible. The flatness varies depending on the manufacturing conditions and operating conditions of the product, but there are cases of good quality and bad quality even under the same conditions. Furthermore, since various varieties and product sizes are manufactured, in order to improve quality efficiently, manufacturing conditions and operating conditions with large variations in quality are found from a large amount of operating data, and focus is placed on targets with high improvement costs. It is effective to do.

  Conventionally, quality data is analyzed manually by stratifying the above conditions, but it takes a lot of time and depends on trial and error, so it has been difficult to find conditions with large variations in quality. The flatness data of the plate on the exit side of the rolling mill 230 of the rolling line 200 is analyzed using the operation analysis apparatus 100 according to the present embodiment, and a condition for increasing the quality variation is found.

In this example, about 28,000 quality data were analyzed. The quality data is the wave height h (see FIG. 6) of the material (slab) 10 measured by the finishing line 260, and the following 29 factors were used as operation factors.
<Quality data>
Wave height h (rolling direction, 0.1 mm unit)
<Operation factor>
Product thickness, product width, product length, plate cutting code, cooling pattern, component C, component Mn, component Nb, component B, transformation temperature, slab thickness, slab width, slab length, heating conditions, charging temperature, heating furnace presence Furnace time, heating time, heating temperature, actual extraction temperature, total number of finishing passes, water cooling start temperature, water cooling stop temperature, water cooling temperature drop, water density 1-6

  The analysis results are shown in FIGS. FIG. 7 is an explanatory diagram showing a decision tree of the result of dividing the operation factor space by the steel plate flatness index. 8A to 8C are tables showing the execution results of the division of the operation factor space by the steel plate flatness index. As calculation conditions, the maximum division depth of the decision tree to be divided is set to 10, and the minimum number of data in the region is set to 100. The threshold value of the FM value in the divided execution area is 100, that is, the wave height is 10 mm. All areas that can be divided are divided, and abnormal values are removed.

  FIG. 7 is a decision tree showing division points obtained by dividing the operation factor space at each region division depth. For example, in the first area division, the operation factor “actual water density 1” is divided at a division point of 0.255. Then, a local region having an FM value equal to or greater than the threshold value 100 is cut out as the region 1, and a local region having an FM value smaller than the threshold value 100 is set as a division target in the second region division. Similarly, area division was performed, and the operation analysis was completed when the maximum division depth, which is the condition for terminating the division process, was 10.

  As shown in FIG. 7, the operating factors used for the division are 8 factors out of 29 factors, and factors that have a large influence on quality variation are extracted. An operation factor space having an appropriate operation range with a small flatness variation including about 20,000 pieces of data is extracted as a region 10 finally remaining as a result of the division. On the other hand, the operation factor space regions having large variations in quality are separated as regions 1 to 9. In FIG. 8A to FIG. 8C, (*) is added to the operation factor used to divide each region.

  Although the number of area data in the areas 1 to 9 is smaller than the total number of data, there is a material in which the wave height h is increased. Detailed analysis of these materials can lead to quality improvement. Further, as can be seen from the histograms of FIGS. 8A to 8C, since all of the areas 1 to 9 are concentrated in an area where the wave height h is small, compared with the area 10 which is a normal operation range. Thus, the difference in the representative value of the quality data such as the average value is small. Such a region is difficult to extract simply by paying attention to the representative value of the quality data, and is considered to have been extracted by setting a variation index based on the probability distribution.

  The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to such examples. It is obvious that a person having ordinary knowledge in the technical field to which the present invention pertains can come up with various changes or modifications within the scope of the technical idea described in the claims. Of course, it is understood that these also belong to the technical scope of the present invention.

  For example, in the above embodiment, the processing of steps S106 to S116 is performed for each division candidate, and each time the variation index of the local region is calculated, the variation index is calculated by comparing with the previously calculated variation index. Although the division candidate point which becomes the maximum is specified, the present invention is not limited to such an example. For example, the probability density function is applied to each of the local regions generated by dividing at all the division candidate points, and after calculating the variation index of all the local regions, the division candidate point having the maximum variation index is specified. Also good.

  Moreover, in the said embodiment, although the space division process performed by step S104-S120 was only 2 division, you may add the division method of 3 or more division to a division candidate. In this case, the probability density function may be fitted to each divided area in step S112.

  Each step of the operation analysis method according to the embodiment of the present invention may be executed by dedicated or general-purpose hardware, but may be executed by software. When the above steps are performed by software, the above steps can be executed by causing a general-purpose or dedicated computer to execute the program. Here, as a structure of the said computer, the structure of the operation analyzer 100 mentioned above is mentioned, for example. In addition, the computer executes each of the processes by executing the program recorded on a storage unit or a removable external recording medium included in the device, or the program acquired via a network, for example. be able to. Furthermore, although the above-described operation analysis apparatus 100 can be configured by a single computer, the operation analysis apparatus 100 is configured by a plurality of computers connected via a network. The configuration function may be divided into roles.

DESCRIPTION OF SYMBOLS 100 Operation analyzer 101 Data input part 102 Operation factor space division part 103 Probability density function calculation part 104 Maximum dispersion | variation model selection part 105 Subdivision target area determination part 106 Analysis result display part

Claims (10)

An operation analysis device that analyzes variations in quality data in a manufacturing process,
A data input section for input of operating factor data and quality data;
An operation factor space dividing unit for setting a plurality of division candidate points each dividing the operation factor space composed of the operation factor data into a plurality of local regions;
For each local region of the operation factor space divided by the division candidate points, a variation index representing the degree of variation of the quality data in the local region is calculated using a probability density function using the quality data as a random variable. A probability density function calculator to
A maximum variation model selection unit that determines the division candidate point that generates a local region that maximizes the variation index calculated for each local region as a definite division point, and deterministically divides the operation factor space by the definite division point When,
A re-division target area determination unit that determines whether or not re-division is necessary for a local area of the operation factor space that has been definitely divided at the definite division point by the maximum variation model selection unit;
An analysis result display unit for displaying an analysis result indicating variation in quality data of a local region of the operation factor space;
With
For the local area determined to require re-division by the re-division target area determination unit, division candidate point setting by the operation factor space division unit, variation index calculation by the probability density function calculation unit, and the maximum variation An operation analysis apparatus that repeats the definite division by the model selection unit and the determination of the necessity of subdivision by the subdivision target region determination unit.   The operation analysis device according to claim 1, wherein the operation factor space dividing unit excludes the local region whose variation index is equal to or greater than a predetermined threshold from the division target when the local region is subdivided. .   3. The variation index is represented by a difference between a random variable value when the cumulative probability of the probability density function is a preset value and a mode value of a probability distribution. The operation analysis device described in 1.   The operation analysis device according to claim 1, wherein the operation factor space dividing unit divides the operation factor space into two local regions at the division candidate points. .   The probability density function calculation unit excludes the candidate division point from candidates when the number of data points belonging to at least one of the local regions divided by the candidate division points is less than a reference value. The operation analysis apparatus according to any one of claims 1 to 4, wherein the operation analysis apparatus is characterized.   The operation analysis according to claim 1, wherein the subdivision target region determination unit ends the division process when the number of divisions of the operation factor space reaches a predetermined number. apparatus.   The operation factor space dividing unit, when the subdivision of the local region, the division candidate points that are outside the region of the local region to be divided among the division candidate points set by the operation factor space division unit. The operation analysis apparatus according to claim 1, wherein the operation analysis apparatus is excluded from candidates. An operation analysis method for analyzing variations in quality data in a manufacturing process,
An input step in which operating factor data and quality data are entered;
A division candidate point setting step for setting a plurality of division candidate points each dividing the operation factor space composed of the operation factor data into a plurality of local regions;
A variation index calculation step for calculating a variation index representing the degree of variation in the quality data in the local region using a probability density function for each local region of the operation factor space divided at the division candidate points;
A deterministic division point determining step for determining, as a deterministic division point, the division candidate point that generates a local region in which the variation index calculated for each local region is maximized;
A deterministic dividing step of deterministically dividing the operating factor space by the definite dividing point;
A re-division determination step for determining whether or not re-division is necessary for a local region of the operation factor space that is definitely divided at the definite division point
An analysis result display step for displaying an analysis result indicating a variation in quality data of a local region of the operation factor space;
Including
For the local area determined to require re-division in the re-division determination step, the division candidate point setting step, the variation index calculation step, the definite division point determination step, the definite division step, and the re-division determination An operation analysis method, characterized in that the steps are repeated. Computer
A data input means for inputting operation factor data and quality data;
Operation factor space dividing means for setting a plurality of division candidate points each dividing the operation factor space consisting of the operation factor data into a plurality of local regions;
Probability density function calculation means for calculating a variation index representing the degree of variation of the quality data in the local region, using a probability density function for each local region of the operation factor space divided at the division candidate points;
Maximum variation model selection means for determining the division candidate point that generates the local region having the maximum variation index calculated for each local region as a definite division point, and deterministically dividing the operation factor space by the definite division point When,
Re-division target area determination means for determining whether or not re-division is necessary for a local area of the operation factor space that has been definitely divided at the definite division point by the maximum variation model selection means,
Analysis result display means for displaying an analysis result indicating a variation in quality data of a local area of the operation factor space;
With
For the local area determined to be subdivided by the subdivision target area determination means, division candidate point setting by the operation factor space division means, variation index calculation by the probability density function calculation means, and the maximum variation A computer program for functioning as an operation analysis device, characterized by repeating a definite division by a model selection unit and a determination of necessity of subdivision by the subdivision target region determination unit. On the computer,
A data input means for inputting operation factor data and quality data;
Operation factor space dividing means for setting a plurality of division candidate points each dividing the operation factor space consisting of the operation factor data into a plurality of local regions;
Probability density function calculation means for calculating a variation index representing the degree of variation of the quality data in the local region, using a probability density function for each local region of the operation factor space divided at the division candidate points;
Maximum variation model selection means for determining the division candidate point that generates the local region having the maximum variation index calculated for each local region as a definite division point, and deterministically dividing the operation factor space by the definite division point When,
Re-division target area determination means for determining whether or not re-division is necessary for a local area of the operation factor space that has been definitely divided at the definite division point by the maximum variation model selection means,
Analysis result display means for displaying an analysis result indicating a variation in quality data of a local area of the operation factor space;
With
For the local area determined to be subdivided by the subdivision target area determination means, division candidate point setting by the operation factor space division means, variation index calculation by the probability density function calculation means, and the maximum variation A computer-readable recording medium recorded with a program for functioning as an operation analysis apparatus, characterized by repeating a definite division by a model selection unit and a determination of necessity of subdivision by the subdivision target region determination unit.

Images