JP2007188471A - Method and device for analyzing quality, computer program, and computer readable storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To estimate a responsible process with sufficient precision by simple processing to present an improvement scheme when a quality defect occurs in a product to be produced through processing in a plurality of processes. <P>SOLUTION: As for a product having a defect in quality, data regarding the quality of a plurality of products processed before and after the product are arranged for the fixed number of products according to a processing order in a data sort part 103, similarity between a defect on the quality of each product to be analyzed and a defect on the quality of the product is evaluated, calculated in a defect similarity calculation part 104 and a process in which products with high similarity of defects is generated with the highest continuity is estimated as the responsible process in a responsible process estimation part 105. Furthermore, the improvement scheme is retrieved and presented from a quality improvement case accumulation part 107 based on an estimation result. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention is a quality analysis in a manufacturing factory suitable for use in estimating a cause process when a defect in quality of a product produced through processing in a plurality of processes, particularly a surface defect occurs in a thin coil as a steel product. The present invention relates to a method, a quality analysis apparatus, a computer program, and a computer-readable storage medium.

  For example, it is often very difficult to identify the causal process when a surface defect occurs in a steel product such as a thin coil, because the number of process steps that pass through the process is large. Processing can be performed by changing the manufacturing parameters of the suspicious process, and it can be judged from the presence or absence of improvement of the defect occurrence status as a result, but it is very laborious because there are many manufacturing parameters for each process. End up.

  In addition, there is a fine flaw that becomes a quality defect that cannot be measured by an automatic flaw inspection apparatus installed in a continuous production line that manufactures steel products such as thin coil. Therefore, it is necessary to use a record of human sensory inspection (hereinafter abbreviated as visual inspection) mainly for visualizing the cause process of minute wrinkles.

  On the other hand, Patent Document 1 collects manufacturing performance data and quality data of products with similar defect occurrence forms, and examines the processes through which these products have passed in common. A technique for presuming that is is disclosed.

  Patent Document 2 discloses a method of investigating actual manufacturing parameter values for each process and determining the presence or absence of abnormality by comparison with management reference data.

  By the way, in Patent Document 3, when a plurality of, especially a large number of wrinkles are generated in a thin coil, information on wrinkle distribution forms is quantified from the wrinkle coordinate data collected and accumulated by an automatic wrinkle inspection device or the like. Disclosed is a surface defect distribution pattern analysis apparatus that can utilize defect distribution information regarding a large number of coils by automatically extracting and presenting as a general index.

  That is, this distribution form analyzing apparatus is a soot distribution feature quantity calculating means for performing arithmetic processing for converting the soot distribution form of the thin coil into a quantitative value based on the coordinate data relating to the position of occurrence of the soot on the thin coil. And analysis result display means for presenting the analysis result of the feature value of the wrinkle distribution form calculated by the wrinkle distribution feature value calculation means to the user.

Japanese Patent Laid-Open No. 10-326816 Japanese Patent Laid-Open No. 9-50949 JP 2005-257660 A

  However, when the total number of processing steps is large, there are often a plurality of steps that have passed through in common, and with the technique disclosed in Patent Document 1, it is difficult to estimate the cause step, and to eliminate the cause Finding an improvement plan was not easy. In addition, wrinkles that cannot be photographed with an automatic wrinkle inspection device were not targets for estimating the causal process.

  Further, in the method disclosed in Patent Document 2, a defect may occur due to a synergistic action of a plurality of manufacturing parameters even if the actual values of the individual manufacturing parameters are not greatly deviated from the normal range. It was difficult to estimate the cause process of the occurrence.

  In addition, the records of visual inspections are characterized by the fact that there are variations among inspectors because they are human sensory inspections, and the number of data that can be entered for manual operation is small, which is more precise than automatic inspection devices. It was difficult to collect records (occurrence position, shape, image).

  In view of the above problems of the prior art, the present invention is concise when an automatic inspection device or a sensory inspection detects that a quality defect has occurred in a product produced through a plurality of processes. The first object is to estimate the causal process with high accuracy and the second object is to present an improvement plan for eliminating the cause of the occurrence of a defect.

The quality analysis method of the present invention is a quality analysis method for estimating a causal process for a quality defect of a product produced through processing in a plurality of processes in a manufacturing factory and presenting a method for removing the cause, A data input step for inputting data on the quality of a plurality of products processed before and after the reference product for each process for the product in which the above defect has occurred (reference product); A data accumulation step for accumulating and storing data relating to quality, and a series of predetermined products in accordance with the processing order for data relating to the quality of the plurality of products processed before and after the reference product for each process A data sorting step for obtaining data to be analyzed by arranging as many as the number, quality defects of each product and the reference product for the data to be analyzed Defect similarity calculation step for evaluating and calculating defect similarity with quality defects, and causal process estimation for estimating a process in which a product with a high defect similarity is generated with the highest continuity as a causal process A quality improvement case search step for searching an improvement plan from a past quality defect improvement case database based on a step, a quality defect of the product and an estimation result of the causal process, and a search for the quality improvement case search step And an output / display step for outputting and presenting the improvement plan from the result.
Moreover, the quality analysis apparatus of the present invention is a quality analysis apparatus that presents a method for estimating a cause process for a defect in quality of a product produced through processing in a plurality of processes in a manufacturing factory and removing the cause. A data input unit for inputting data on the quality of a plurality of products processed before and after the reference product for each process for the product (reference product) in which a defect in quality occurs, and the data input unit A data accumulating unit for accumulating and storing the data relating to the quality, and data relating to the quality of the plurality of products processed before and after the reference product for each process, according to a processing order. A data sorting unit that obtains data to be analyzed by arranging as many as the number of products, and a defect in quality of each product and a defect in quality of the reference product for the data to be analyzed. Defect similarity calculator that evaluates and calculates similarity, cause process estimator that estimates that the process having the highest continuity of the product having the highest defect similarity is the cause process, and past quality Quality improvement case storage unit for storing quality defect improvement cases for defects and causal processes, and quality improvement cases for searching for improvement proposals from the quality improvement case storage unit for the product quality defects and the estimation results It has a search part, and the output and display part which outputs and presents an improvement plan from the search result of the said quality improvement example search part, It is characterized by the above-mentioned.
The computer program of the present invention is a computer program for causing a computer to execute a quality analysis process for estimating a cause process for a quality defect of a product produced through a process in a plurality of processes in a manufacturing factory. The data input process for inputting data relating to the quality of a plurality of products processed before and after the reference product for each process for the product (reference product) in which the defect occurs, and the quality input in the data input process A data accumulation process for accumulating and storing data, and for each process, data relating to the quality of the plurality of products processed before and after the reference product, a series of a predetermined number of products according to the processing order The data sorting process to obtain the data to be analyzed by arranging the data, and the data to be analyzed Defect similarity calculation processing for evaluating and calculating the defect similarity between the defect and the quality defect of the reference product, and the process in which the product having the high defect similarity is generated with the highest continuity is the cause process. Causal process estimation process to estimate, quality improvement case search process for searching improvement plan from past quality improvement case accumulation database with respect to product quality defects and estimation result, and quality improvement case search process The computer is caused to execute an output / display process for outputting and presenting an improvement plan from the search results.
A computer-readable storage medium of the present invention is characterized in that the computer program described above is stored in a storage medium.

  According to the present invention, when the occurrence history of defects in quality is indexed, when an automatic inspection device or sensory inspection detects that a quality defect has occurred in a product produced through processing in multiple steps In addition, the causal process can be estimated with high accuracy by simple processing. Furthermore, an improvement plan for removing the cause of the occurrence of the defect can be presented.

(First embodiment)
DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described with reference to the accompanying drawings. FIG. 1 is a diagram showing a schematic configuration of a quality analysis apparatus according to the first embodiment of the present invention. FIG. 2 is a flowchart showing an example of a quality analysis method executed using the quality analysis apparatus of FIG. Hereinafter, with reference to FIG. 1 and FIG. 2, explanation will be given on the estimation of a cause process for generating surface flaws in a steel product manufacturing factory, taking a thin coil (hereinafter referred to as “coil”) as a product and surface flaws as an example. To do.

  As shown in FIG. 1, a quality data storage server is installed outside the quality analysis apparatus according to the present embodiment. In the quality data storage server, data related to surface defects measured by a defect inspection device or the like is stored for a long period (for example, in the order of coil numbers (numbers uniquely assigned to the coils), for each process and each processing day) Half a year) Accumulated and stored. When a surface flaw is generated in a coil that has been manufactured in a final process after a plurality of processes, a reference coil is generated for each process in which the coil (hereinafter referred to as “reference coil”) is processed. The processing date in the relevant process is retrieved from the coil number and data on the surface defects of all the coils processed on the same day are input to the data input unit 101 to be analyzed (data input step in FIG. 2). 201). In the following data sorting step, when more coils are targeted, data on the surface flaws of the coil processed on multiple days before and after the reference coil processing date is input, and when fewer coils are targeted, Data on the surface defects of the coil processed in several hours before and after the processing time of the reference coil is input.

  Next, the input data is stored and stored in the data storage unit 102 for subsequent processing (corresponding to the data storage step 202 in FIG. 2).

  Next, in the data sorting unit 103, the data on the surface defects of the plurality of coils processed before and after the reference coil for each process is arranged by a predetermined number of products according to the processing order (data sorting step in FIG. 2). 203). This is analysis target data used for the subsequent analysis.

  A simple example shown in FIG. 3 will be described. As for the surface defects, the surface defect types (5 types of A, B, C, D, and E), the surface defect grades (three grades 1, 2, and 3), the process from step 1 It is an example of the analysis object data which made the 3 process of process 3 and selected and sorted the coil processed before and after the reference coil by 5 coils before and after each in the processing order.

  As a sorting method, in addition to selecting a certain number of continuous coils before and after the reference coil in accordance with the processing order, it is possible to select only before or after the reference coil. It goes without saying that it is possible to select by thinning out every other line.

  When multiple surface defects occur in one coil, use the defect feature quantity such as defect type and grade, for example, set the aggregation unit for the combination of defect type and grade, and count the number of defects. What is necessary is just to sort the combination of the type and grade of surface defects with many as the representative surface defects of the reference coil. However, it goes without saying that the method for selecting representative surface defects from a plurality of surface defects present in one coil is not limited to this method.

  Based on the analysis target data, the defect similarity calculation unit 104 calculates the defect similarity between each coil and the reference coil (corresponding to the defect similarity calculation step 204 in FIG. 2). For example, it can be calculated as 2 points if both the species and grade match, 1 point if either one matches, and 0 point if neither match. However, it goes without saying that the method of calculating the defect similarity is not limited to this method. The calculation results are shown in the row of “Defect similarity” in FIG.

  The causal process estimation unit 105 first evaluates the continuity of the defect similarity from the series data of the defect similarity of each coil for each process calculated by the above method (in the causal process estimation step 205 of FIG. 2). Correspondence). Specifically, the coil having the maximum defect similarity of 2 is assumed to be a coil having a high defect similarity, and the coil having a high defect similarity with respect to 11 analysis target coils including the reference coil is selected. The number is used as an index of continuity of defect similarity.

This value is as follows.
Step 1: 2
Process 2: 5
Process 3: 4
In this case, the continuity of the defect similarity in the process 2 is the highest, and it is estimated that this is the causal process of the surface defect (species A) of the reference coil.

  On the other hand, in the quality improvement case accumulation unit 107, cases when quality defect improvement performed in the past are performed are accumulated as a quality improvement case database. Each case includes at least a defect feature amount and a cause process characterizing the defect, and a quality improvement method (improvement plan) associated with them. For example, when the defect feature quantity is type A, grade 3, and the cause process is process 2, proposals such as “replace the roll installed in the equipment of process 2” are accumulated as improvement plans.

  The quality improvement case search unit 108 uses the defect feature quantity (type and grade) of the coil to be analyzed and the cause process estimated by the cause process estimation unit 105 as search keys to obtain the quality from the quality improvement case storage unit 107. An improvement method (improvement plan) is searched (corresponding to the quality improvement case search step 207 in FIG. 2).

  When the simple example shown in FIG. 3 is used, according to the first embodiment, since the estimated cause process is the process 2, the search key and the type A and grade 3 of the coil are combined. Then, a quality improvement method “replace rolls installed in the equipment in step 2” is obtained from the quality improvement case accumulation unit 107 as a search result. By exchanging the corresponding roll based on this search result, it becomes possible to keep the quality of the product manufactured after the roll exchange good.

  Finally, the output / display unit 106 outputs the estimation result that the process 2 is the causal process and the improvement plan to the outside and displays them on the screen (corresponding to the output / display step 206 in FIG. 2).

(Second Embodiment)
In addition, as a method for evaluating the continuity of the defect similarity in the causal process estimation unit 105 in FIG. 1, the defect similarity calculated by the defect similarity calculation unit 104 in FIG. A total value for 11 analysis target coils may be obtained, and that value may be used as an index of continuity of defect similarity.

The values calculated in this way are as follows.
Step 1: 6
Process 2:12
Process 3:10
From this result, it is estimated that the process 2 is a causal process.

(Third embodiment)
Further, as a method for evaluating the continuity of the defect similarity in the causal process estimation unit 105 in FIG. 1, the defect similarity calculated by the defect similarity calculation unit 104 in FIG. Arbitrary two coils are selected, proportional to the product of the defect similarity of both coils, and the processing interval between the two coils (for example, 1 when the coils are processed continuously, and one coil between them) 2) (which is defined as 2) and is calculated for all two coils to obtain a total value, which may be used as an index of continuity of the defect similarity. The calculation formula is shown in Formula (1) and Formula (2).

The values calculated in this way are as follows.
Step 1: 9.2
Step 2: 62.1
Step 3: 34.6
From this result, it is estimated that the process 2 is a causal process.

  Note that the above calculation is a value calculated for the coils of all defect similarity values larger than 0 in the equations (1) and (2), but in order to simplify the calculation, The continuity index may be calculated only for the coil having the maximum similarity (2 in the example shown in FIG. 3).

(Fourth embodiment)
As another example of defect feature values other than the type and grade, an embodiment will be described in the case where there is a feature value characterizing a defect when there are a plurality of defects in one coil. As in the first to third embodiments, the data sorting unit 103 in FIG. 1 has a series of data on the surface defects of the plurality of coils processed before and after the reference coil for each process in accordance with the processing order. A predetermined number of products are arranged (corresponding to the data sorting step 203 in FIG. 2). This is analysis target data used for the subsequent analysis.

When a plurality of surface defects are generated in one coil, for example, the analysis is performed using the feature amount described below. An example will be described with reference to FIG. First, as a method for obtaining a feature amount characterizing a plurality of defects, there is a distribution form analysis method disclosed in Patent Document 3. In this method, the surface defects generated in each product (coil) are divided into one or a plurality of groups according to the generation position and form, and the feature amount related to the spatial distribution of the surface defects in each group is calculated. . As the group features, the center of gravity position (rolling direction position X [m], width direction position Y [mm]), space size (rolling direction space size σX [m], width direction space size σY [mm]), surface defect There are number [pieces] and surface flaw number density [pieces / m ² ]. The group divided here is a defect group.

  FIG. 6 shows the above four feature quantities (center of gravity position, space size, number of surface defects, and surface defect number density) that characterize a plurality of defects as data related to surface defects. It is an example of the analysis object data which selected the coil processed before and behind the reference | standard coil (the said coil), respectively, and ordered 5 coils each before and after in the process order.

  Hereinafter, calculation of defect similarity will be described. The defect similarity calculation unit 104 in FIG. 1 calculates the defect similarity between the analysis target coil and the reference coil as follows (corresponding to the defect similarity calculation step 204 in FIG. 2).

  In order to obtain the defect similarity between one coil in the analysis target coil and the reference coil, first, an arbitrary one is selected from a plurality of defect groups in each coil, and the similarity is obtained for the combination. From the similarities for all combinations of defect groups, the similarity that shows the highest value among them is defined as the defect similarity between the two coils.

  The similarity between defect groups is calculated based on the judgment value. Here, the determination value is determined so as to be 1 for each feature value of the defect group, for example, 1 for complete coincidence and 0 for complete disagreement. As a method for calculating the determination value, there is a method (determination value A) in which 1 is set within ± 30% of the characteristic amount of the reference coil and 0 is set otherwise. FIG. 7 shows the result of calculating the judgment value A for the data of FIG.

  As another method of calculating the determination value, when the number of surface defects is 100 and the number of surface defects of the comparison target coil is 20, the determination value is 1− (| 100−20 | / 100) = 0.2, In the case of 120, 1− (| 100−120 | / 100) = 0.8, and in the case of 220, 1− (| 100−220 | / 100) = − 0.2. There is a method (determination value B) for rounding up to obtain a determination value.

  Next, an average of determination values or a weighted average is obtained and used as the similarity between defect groups. The defect similarity 1 in FIG. 7 shows the result of the defect similarity from the similarity based on the weighted average, and also shows the weight Wi of each feature amount used for the calculation. As a method for determining the weight Wi for the feature quantity i, for example, there is a formula (3).

  Here, Pij represents a determination value of the feature value i of the coil j and is assumed to be non-negative. N represents the number of coils, and M represents the number of types of feature values. wi is the defect group average of the determination values of the feature quantity i so that the weight becomes larger as the total of the determination values is smaller (the parameter of the defect group is a defect group belonging to the coil included in the process. 1 / x, which is a decreasing function, is applied to any of all steps. The weight Wi is normalized to a range of 0 to 1 by multiplying wi by a coefficient so that the sum is 1. Normalization is essential when the following defect similarity adjustment is performed.

  As the defect similarity between the analysis target coil and the reference coil, the maximum value obtained by calculating the similarity obtained as described above for all combinations of defect groups that can be taken by one set of coils is adopted.

  By the way, when the existence of a defect that is difficult to detect is assumed, there is a method of adjusting the defect similarity so as not to exceed the preceding and following defect similarities as a method of virtually complementing the defect oversight. The condition of the target coil for adjusting the defect similarity is 0 in which the defect similarity is completely inconsistent, and further, the coil whose defect similarity is 0 including the target coil among the front and rear coils adjacent to the target coil. Is N0, and N0 and N2 are the number of coils with non-zero defect similarity that are connected before or after the N0 coils, respectively. The target coil is the target. The above description is illustrated in FIG.

  The defect similarity after adjustment is expressed by equation (5) using N0, N1, and N2.

  Here, α indicates the value of the defect similarity that is not larger among the non-zero defect similarities for the front and rear from the adjustment target coil. Specifically, when FIG. 8B is applied, the coils having the processing orders of −2 and 1 are applicable, and the defect similarity is 1 and 1, so that α is 1. When applied to FIG. 8C, the coils having the processing orders of −2 and 1 correspond to the same, and the defect similarity is 1 and 0.8, so α is 0.8. It is only necessary to satisfy that the defect similarity after adjustment decreases as N0 increases. Furthermore, for the purpose of virtually complementing the defect oversight, it is better to set a limit so as not to exceed the defect similarity α of the coil having the non-zero defect similarity occurring immediately before and after. FIG. 7 shows a defect similarity 2 as a result of defect similarity calculation after the defect similarity 1 is adjusted.

  Hereinafter, causal process estimation will be described. The causal process estimation unit 105 first evaluates the continuity of the defect similarity from the series data of the defect similarity of each coil for each process calculated by the above method (in the causal process estimation step 205 of FIG. 2). Correspondence).

  The continuity for evaluating the continuity of the defect similarity indicates that the greater the value, the higher the continuity, and the greater the non-zero defect similarity, the greater the value, the greater the defect similarity. It is an indicator. Specifically, the continuity is calculated as follows (a) to (c).

(A) When attention is paid to one coil (starting coil) in the series of defect similarity data in a certain process, first, the defect similarity S1 of the starting coil, the defect similarity S1, and the coil adjacent to the starting coil Values S1 and S2 multiplied by the defect similarity S2 and values S1, S2 and S3 multiplied by the defect similarity S3 of the coil adjacent to the coil S2 are calculated. The value obtained by this operation is called a partial product.

(B) In this way, the partial product is sequentially calculated until the final coil of the series data is obtained, and is repeatedly added to the continuity. This operation covers the range from the starting coil to the ending coil, but the same operation is performed for the same operation from the starting coil in the reverse direction to the starting coil. In this case, the defect similarity S1 of the starting coil which is the second time is not added.

(C) The continuity is obtained by repeating the above operation using all coils in the series data as starting coils.

Formula (6) shows the calculation formula when the above calculation is based on the case where the defect similarity corresponding to the series j-th coil is s _j , the starting coil index is i, and the number of coils in the series data is M (= 6). .

Equation (6) is the sum of products for all combinations of adjacent defect similarities. When N coils having a defect similarity of 1 are consecutive from equation (6), the continuity is N ² . When there are two sets of consecutive sequences as in the sequence of FIG. 8A, the sum of the respective continuities N1 ² + N2 ² is the continuity of the entire sequence, indicating the positional relationship between the two sets of consecutive sequences. Since it does not depend on the number of coils N0, it is necessary to appropriately adjust the similarity reflecting N0.

  For this reason, when it is desired to evaluate the influence of the number N0 of coils that divide the two sets, the defect similarity calculation unit 104 determines the influence of N0 on the defect similarity so that the continuity increases as N0 decreases. Reflect the defect similarity. At this time, when the defect similarity is adjusted, the continuity of the adjusted series data is prevented from becoming too large by not using the coil whose defect similarity is changed by the adjustment as the starting coil. Specifically, when the defect similarity S4 of the fourth coil in the series in Equation (6) is the adjusted defect similarity, the partial product corresponding to 11 terms corresponding to the fourth column or the fourth row Is not added.

  If this operation is not performed, for example, a series A having 10 consecutive defect similarities 1 and a series B similar to FIG. 8A with N1 = N2 = 5 and N0 = 1 are compared. In this case, the continuity of the series A is 100 and the continuity of the series B is 109 (fractional truncation), and the definition of the continuity (the non-zero defect similarity is continuous, the larger the defect similarity is, the larger the value is). Contrary to

  In order to avoid this, when calculating the continuity, it is not necessary to add a partial product having a coil whose defect similarity is adjusted as a starting coil. According to this calculation method, the series A is 100, and the series B is 92 (decimal truncation), which meets the definition of continuity.

  The partial sum of products for all combinations of adjacent defect similarities is a sum obtained by removing the product for combinations starting from a coil whose defect similarity is adjusted in Equation (6). The above continuity is used as an index of continuity of defect similarity.

The continuity of equation (6) with a defect similarity of 1 is as follows.
Step 1: 2.88
Process 2: 8.29
Step 3: 2.97
In this case, the defect-like continuity of the process 2 is the highest, and it is estimated that this is a causal process of the surface flaw of the reference coil.

Similarly, the continuity by the defect similarity 2 is as follows.
Step 1: 3.77
Step 2: 8.29
Step 3: 3.26
In this case, the defect-like continuity of the process 2 is the highest, and it is estimated that this is a causal process of the surface flaw of the reference coil.

(Fifth embodiment)
FIG. 4 is a block diagram illustrating an example of a computer system that can configure the above-described quality analysis apparatus. In FIG. 4, reference numeral 400 denotes a computer (PC). A computer PC 400 includes a CPU 401, executes device control software stored in a ROM 402 or a hard disk (HD) 411, or supplied from a flexible disk drive (FD) 412, and generalizes each device connected to the system bus 404. Control. Each function unit of the present embodiment is configured by a program stored in the CPU 401, the ROM 402, or the hard disk (HD) 411 of the PC 400.

  Reference numeral 403 denotes a RAM which functions as a main memory, work area, and the like for the CPU 401. Reference numeral 405 denotes a keyboard controller (KBC) which controls to input a signal input from the keyboard (KB) 409 into the system main body. A display controller (CRTC) 406 performs display control on the display device (CRT) 410. A disk controller (DKC) 407 is a hard disk that stores a boot program (startup program: a program for starting execution (operation) of hardware and software of a personal computer), a plurality of applications, editing files, user files, a network management program, and the like. The access to the (HD) 411 and the flexible disk (FD) 412 is controlled.

  Reference numeral 408 denotes a network interface card (NIC) that exchanges data bidirectionally with a network printer, another network device, or another PC via the LAN 413.

  The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device.

(Sixth embodiment)
In FIG. 1, the quality data storage server may store and store quality data including surface defect data and quality measurement environment data, as shown in FIG. The surface wrinkle data is data relating to the surface wrinkle based on the result of sensory inspection (inspection by human). Quality measurement environment data is data that represents factors that affect the results of sensory inspections, such as operating conditions (such as line speed), skill levels, This data shows the environment prepared for performing a sensory test such as an examiner with different species discrimination ability depending on physical condition, lighting in an examination room, and the like. As an example of this case, a simple case shown in FIG. 10 will be described. This includes the inspectors (three persons a, b, c) as the data related to the quality measurement environment data, the surface flaw types as the surface flaw data (five types of A, B, C, D, E), and the occurrence position Example of analysis target data sorted by selecting the five coils before and after the reference coil in the order of processing, and selecting the coils processed before and after the reference coil as the three processes from Step 1 to Step 3, which are the target of the coordinates and cause process estimation. It is.

A fuzzy set can be applied to the calculation of defect similarity.
In calculating the defect similarity between 疵 α of one coil and 疵 β of another coil, fuzzy indicating the mutual relationship between generation positions P, which are surface flaw data representing fuzzy sets in terms of fuzzy numbers, and between species T Calculate using sets. A qualifier, which is quality measurement environment data, is used as a factor for determining the membership function of the fuzzy number representing the occurrence position. The membership value of the fuzzy set indicating the mutual relationship between the species is 1 for the same species, 0 for the completely different species, and similar to the form, etc. even for different species. If there is a point, the intermediate value is used. Since the occurrence position is recorded by the examiner, the occurrence position is recorded with an error of a tendency that is different for each examiner from the true occurrence position. The absolute value of the error amount includes factors that are difficult to record, such as the individuality and physical condition of the examiner, in addition to the plate speed of the production line, so the range of error amount determined for each examiner as a membership function. Expressed as a reflected fuzzy number. In addition, there may be variations in judgment results due to differences in the quality measurement environment, so when determining membership values of fuzzy sets that indicate the mutual relationship between species, the quality measurement environment The difference may be reflected.

  The defect similarity is calculated as the maximum value of the membership value of the product set of the fuzzy numbers of the occurrence position Pα of 疵 α and the occurrence position Pβ of 疵 β and the membership value of the fuzzy set indicating the association of the species. .

  The membership function of the fuzzy number at the occurrence position is determined in accordance with the quality measurement environment (here, the certifier) that records the defects. For the fuzzy set showing the mutual relationship between the species, the difference in the quality measurement environment is not reflected here for the sake of simplicity.

  Equation (7) shows the calculation formula for the defect defect similarity. At this time, x represents the occurrence position of wrinkles recorded by the examiner.

A calculation example of the defect similarity is shown using the coil and coil No. 1-2 in FIG. First, the fuzzy numbers P1 and P2 representing the occurrence position of the wrinkles between the coil and the coil No. 1-2 are expressed by the equation (8) using the membership functions μ _fa and μ _fb of the examiners a and b. The functions μ _fa and μ _fb are as shown in Expression (8 ′) and Expression (8 ″), respectively, and are calculated as shown in Expression (9) from Expression (8 ′) and Expression (8 ″). .

Flaw type A of both coils, the membership value mu _A⇔B of fuzzy set indicating the relevant B, and 0.9 regardless of the flaw generation position from FIG. 11, the maximum mu _{Parufa∩Pbeta} of formula (7) Since the value is 1, it is calculated as shown in the following formula (10), and the defect similarity is 0.95.

The results calculated for the other coils in the same procedure as above are shown as defect similarity 3 in FIG.
The continuity evaluation method for defect similarity in the causal process estimation unit 105 in FIG. 9 is the same continuity evaluation method as in the second embodiment, and the continuity for each process from the defect similarity 3 in FIG. The calculated value is as follows.

Step 1: 3.85
Step 2: 7.10
Step 3: 5.90

  From the above results, since the continuity of the defect similarity in step 2 is the highest, it is presumed that step 2 is the causal step of the surface defect (type A) of the reference coil.

Furthermore, in order to present an improvement plan, the cause process was set as step 2, and the quality improvement case search was performed using the type of the reference coil and the generation position as search keys, and an improvement plan “Perform equipment repair in step 2” was obtained. Upon receiving the improvement proposal and repairing the equipment in step 2, the generation of surface flaws was improved.
In addition, as a result of detailed investigation after this analysis, it was found that surface flaws occurred due to the equipment failure in step 2.

  Another object of the present invention is to supply a storage medium storing software program codes for realizing the functions of the above-described embodiments to a system or apparatus, and store the computer (CPU or MPU) of the system or apparatus in the storage medium. Needless to say, this can also be achieved by reading and executing the programmed program code.

  In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention. As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

  Further, by executing the program code read by the computer, not only the functions of the above-described embodiments are realized, but also an OS (operating system) or the like running on the computer based on the instruction of the program code. It goes without saying that a case where the function of the above-described embodiment is realized by performing part or all of the actual processing and the processing is included.

  Further, after the program code read from the storage medium is written into a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion is performed based on the instruction of the program code. It goes without saying that the CPU or the like provided in the board or function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

  The cause process of surface flaws that occurred in the manufacturing process of thin coil and estimation of improvement were presented. The surface wrinkle data of the 43 coils processed before and the 49 coils processed after that in the three steps from Step 1 to Step 3 are input to the coil in which the surface wrinkles occurred, and the data is sorted. FIG. 5 shows the result of the defect similarity calculation. Here, as the data on the surface wrinkles, in addition to the wrinkle type and wrinkle grade, 4 points of the longitudinal position and the width direction position of the wrinkles are input, and whether or not each matches the surface wrinkles of the reference coil, from 0 point A defect similarity of 4 points was calculated. In addition, about the longitudinal direction position and the width direction position of the ridge, it was determined that the lengths and the width direction of the coil were equally divided when they were equally divided into 10 corresponding to the same section. As the continuity index, the method described in the third embodiment is used.

The evaluation results are as follows.
Step 1: 1763
Process 2: 1903
Step 3: 1341
From the above results, it was estimated that step 2 was the causal step.

  In order to present the improvement plan, the cause process is step 2, and the quality improvement case search is performed using the type of the reference coil, the grade, the grade, the longitudinal position and the width direction position of the cage as a search key, and the improvement plan “step 2”. "Do the equipment repairs." Upon receiving the improvement proposal and repairing the equipment in step 2, the generation of surface flaws was improved.

  In addition, as a result of detailed investigation after this analysis, it was found that surface flaws occurred due to the equipment failure in step 2.

It is a figure which shows schematic structure of the quality analysis apparatus of the 1st Embodiment of this invention. It is a flowchart for demonstrating the quality analysis method of the 1st Embodiment of this invention. It is a figure for demonstrating the data sorting method of the 1st Embodiment of this invention. It is a block diagram which shows an example of the computer system which can comprise the quality analysis apparatus of embodiment of this invention. It is a figure which shows the result of having performed the data sort in the 1st Embodiment of this invention, and defect similarity calculation. It is a figure for demonstrating the method of the data sort in the 4th Embodiment of this invention. It is a figure for demonstrating the defect similarity in the 4th Embodiment of this invention. It is a figure for demonstrating the determination formula of the defect similarity adjustment availability which concerns on the 4th Embodiment of this invention. It is a figure which shows schematic structure of the quality analyzer of the 6th Embodiment of this invention. It is a figure for demonstrating the data sorting method of the 6th Embodiment of this invention. It is a figure which shows the example of the membership value of the fuzzy set which shows the relationship of two varieties concerning the 6th Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Data input part 102 Data storage part 103 Data sort part 104 Defect similarity calculation part 105 Cause process estimation part 106 Output / display part 107 Quality improvement example storage part 108 Quality improvement example search part

Claims (34)

A quality analysis method for estimating a cause process for a defect in quality of a product produced through processing in a plurality of processes in a manufacturing factory and presenting a method for removing the cause,
A data input step for inputting data relating to the quality of a plurality of products processed before and after the reference product for each process for a product in which a defect in quality occurs (reference product);
A data accumulation step for accumulating and storing the data relating to the quality input in the data input step;
A data sorting step for obtaining data to be analyzed by arranging data relating to the quality of the plurality of products processed before and after the reference product for each process by a predetermined number of products according to the processing order; ,
A defect similarity calculation step for evaluating and calculating a defect similarity between the quality defect of each product and the quality defect of the reference product for the analysis target data;
A causal process estimation step for estimating that the process in which the product having a high degree of defect similarity is generated with the highest continuity is a causal process;
A quality improvement case search step for searching for an improvement plan from a past quality defect improvement case database based on the quality defect of the product and the estimation result of the cause process;
A quality analysis method comprising: an output / display step for outputting and presenting the improvement plan from the search result of the quality improvement case search step.   The quality analysis method according to claim 1, wherein the product is a steel product.   The quality analysis method according to claim 2, wherein the defect is a surface defect.   The defect similarity calculation step evaluates and calculates the defect similarity between the defect in quality of each product and the defect in quality of the reference product based on one or a plurality of defect feature quantities characterizing the defect. The quality analysis method according to any one of claims 1 to 3, wherein:   5. The quality analysis method according to claim 4, wherein, in the defect similarity calculation step, the defect feature amount includes a cocoon type and a cocoon grade.   The quality analysis method according to claim 4, wherein the defect feature amount is one or a combination of an occurrence frequency, an occurrence position, and a form in the defect similarity calculation step.   The quality analysis method according to claim 4, wherein the defect feature amount is one or a combination of a space size, a position, and a form in the defect similarity calculation step. The defect similarity calculation step divides a plurality of defects of each product into a plurality of defect groups based on the defect feature amount by a distribution form analysis method,
8. The defect similarity in quality between each of the products and the reference product is calculated by a predetermined process based on the feature amount of the defect group. Quality analysis method.   The quality analysis method according to claim 1, wherein in the data input step, data related to a quality measurement environment is input in addition to the data related to the quality.   The quality analysis method according to claim 9, wherein the defect similarity calculation step calculates using a defect feature amount represented by a fuzzy set.   The quality analysis method according to claim 9 or 10, wherein the defect similarity calculation step defines a membership function of a defect feature amount represented by a fuzzy set using data relating to the quality measurement environment.   2. The causal process estimation step estimates the causal process using the number of products having high defect similarity in the analysis target data for each process as an index of continuity of defect similarity of the process. The quality analysis method of any one of -11.   2. The causal process estimation step estimates a causal process using a sum or a partial sum of defect similarities in the analysis target data for each process as an index of continuity of defect similarities of the process. The quality analysis method of any one of -11.   In the causal process estimation step, the defect similarities in the analysis target data for each process are summed up or partial sums of products for all combinations of adjacent defect similarities in the order arranged in the data sorting step. The quality analysis method according to claim 1, wherein the quality analysis method is used as an index of continuity of the defect similarity.   The cause process estimation step has an index indicating a higher value as the number of products having a high defect similarity in the analysis target data for each process is larger and a processing interval between the products having a higher defect similarity is narrower. The quality analysis method according to claim 1, wherein the quality analysis method is used as an index of continuity of defect similarity in the process.   The causal process estimation step has an index indicating a higher value as the total value of defect similarities of products existing in the analysis target data for each process is larger and as the processing interval between the products is narrower. The quality analysis method according to claim 1, wherein the quality analysis method is used as an index of continuity of defect similarity. A quality analysis device that estimates a causal process for a quality defect of a product produced through processing in a plurality of processes in a manufacturing factory and presents a method for removing the cause,
A data input unit for inputting data on the quality of a plurality of products processed before and after the reference product for each process for a product in which a defect in quality has occurred (reference product);
A data storage unit for storing and storing data relating to the quality input to the data input unit;
A data sorting unit that obtains data to be analyzed by arranging data relating to the quality of the plurality of products processed before and after the reference product for each process by a predetermined number of products according to the processing order; ,
For the analysis target data, a defect similarity calculation unit that evaluates and calculates the defect similarity between the quality defect of each product and the quality defect of the reference product,
A cause process estimation unit that estimates that a process in which the product having a high degree of defect similarity is generated with the highest continuity is a cause process;
A quality improvement case accumulation unit that accumulates quality defect improvement cases for past quality defects and causal processes,
A quality improvement case search unit that searches for an improvement plan from the quality improvement case storage unit for the product quality defect and the estimation result;
An output / display unit that outputs and presents an improvement plan from a search result of the quality improvement example search unit. A computer program for causing a computer to execute a quality analysis process for estimating a cause process for a defect in quality of a product produced through processing in a plurality of processes in a manufacturing factory,
A data input process for inputting data on the quality of a plurality of products processed before and after the reference product for each process for a product in which a defect in quality has occurred (reference product);
A data accumulation process for accumulating and storing the data related to the quality input in the data input process;
A data sorting process for obtaining data to be analyzed by arranging data relating to the quality of the plurality of products processed before or after the reference product for each process by a predetermined number of products according to the processing order; ,
A defect similarity calculation process for evaluating and calculating a defect similarity between a quality defect of each product and a quality defect of the reference product for the analysis target data;
Cause process estimation process for estimating that the product having a high degree of defect similarity has the highest continuity and is a cause process;
Quality improvement case search processing for searching for improvement plans from a past quality improvement case accumulation database for the product quality defects and the estimation results;
A computer program for causing a computer to execute an output / display process for outputting and presenting an improvement plan from a search result of the quality improvement case search process.   The computer program product according to claim 18, wherein the product is a steel product.   The computer program according to claim 19, wherein the defect is a surface defect.   The defect similarity calculation process evaluates and calculates the defect similarity between the defect in quality of each product and the defect in quality of the reference product based on one or more defect feature quantities characterizing the defect. The computer program according to any one of claims 18 to 20, wherein   The computer program product according to claim 21, wherein in the defect similarity calculation process, the defect feature amount is a cocoon type and a cocoon grade.   The computer program according to claim 21, wherein in the defect similarity calculation process, the defect feature amount is one or a combination of an occurrence frequency, an occurrence position, and a form.   The computer program according to claim 21, wherein in the defect similarity calculation process, the defect feature amount is one or a combination of a space size, a position, and a form. The defect similarity calculation process divides a plurality of defects of each product into a plurality of defect groups based on the defect feature amount by a distribution form analysis method,
25. The defect similarity in quality between each product and the reference product is calculated by a predetermined process based on the feature amount of the defect group, according to any one of claims 21 to 24. Computer program.   19. The computer program according to claim 18, wherein in the data input process, data related to a quality measurement environment is input in addition to the data related to the quality.   27. The computer program according to claim 26, wherein the defect similarity calculation processing is calculated using a defect feature amount represented by a fuzzy set.   28. The computer program according to claim 26 or 27, wherein the defect similarity calculation process defines a membership function of a defect feature amount represented by a fuzzy set using data relating to the quality measurement environment.   19. The causal process estimation process estimates the causal process using the number of products having high defect similarity in the analysis target data for each process as an index of continuity of defect similarity in the process. The computer program according to any one of to 28.   19. The causal process estimation process estimates a causal process using a sum or a partial sum of defect similarities in the analysis target data for each process as an index of continuity of defect similarities of the process. The computer program according to any one of to 28.   In the causal process estimation process, the defect similarities in the analysis target data for each process are summed up or partial sums of products for all combinations of adjacent defect similarities in the order arranged in the data sorting step. The computer program according to any one of claims 18 to 28, wherein the computer program is used as an index of continuity of defect similarity.   The cause process estimation process has an index indicating a higher value as the number of products having a high defect similarity in the analysis target data for each process is larger and a processing interval between the products having a higher defect similarity is narrower. The computer program according to any one of claims 18 to 28, wherein the computer program is used as an index of continuity of defect similarity in the process.   The cause process estimation process has an index indicating a higher value as the total value of defect similarities of products existing in the analysis target data for each process is larger and as the processing interval between the products is narrower. The computer program according to any one of claims 18 to 28, wherein the computer program is an index of continuity of defect similarity.   A computer-readable storage medium storing the computer program according to any one of claims 18 to 33.

Images