JP2009087027A - Process management method, process management program, and process management system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a process management method, a process management program, and a process management system for managing a manufacturing line in which a plurality of devices are arranged in parallel in one process, and for estimating the change of a contribution degree of each device to the finish of a product. <P>SOLUTION: This process management system 1 manages a manufacturing line 101 in which a plurality of devices 103 are arranged in parallel in one process 102. The data collection means 2 of the process management system 1 acquires data including history information and the finish of a product from the manufacturing line 101. Also, an analyzing means 3 performs the analysis of qualification type I to the data, and searches the relative relation of the contribution degree of each device 103 of the manufacturing line 101 to the finish of the product. Then, an estimation means 4 estimates the change of the contribution degree of each device 103 so that the total sum of the change of the contribution degree of the device 103 can be minimized on the basis of the relative relation. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a process management method, a process management program, and a process management system, and more particularly to a process management method, a process management program, and a process management system for managing a production line in which a plurality of devices are arranged in parallel in one process.

  With the miniaturization and high precision of electronic devices, the manufacturing process has become complicated and the margin has been reduced. Technology that ensures manufacturing quality at the product design and process design stages is being improved by using simulations, etc., but due to the situation described above, not all process mechanisms and variable factors have been fully modeled during process design. However, this situation is expected to continue in the future.

  On the other hand, as the cost of collecting and accumulating information on the production line decreases as IT technology advances, there are a variety of things that can be obtained in the mass production process in order to stabilize the process in the mass production stage and maintain and improve product reliability. The effectiveness of maximizing the use of information is being recognized. However, in order to extract useful information from a vast amount of data obtained from more than a thousand devices arranged in hundreds of processes, there are many technical problems to be solved.

  In the manufacture of electronic devices, especially in quality improvement / stabilization activities in mass production lines, analysis of equipment differences is one of the most effective methods. In an electronic device production line, particularly a mass production line, from the viewpoint of securing production capacity, a plurality of processing apparatuses are often provided in one process and used in parallel. In this case, the difference is often a problem. The machine difference is a phenomenon in which a difference that cannot be ignored is produced in spite of using the same model of equipment under the same processing conditions. Machine differences are considered to be caused by individual differences in devices and subtle differences in installation environment. As described above, when a plurality of apparatuses are used in one process, a new variation factor is added to the product finish, which is a machine difference. In many cases, the cause of a machine difference cannot be determined even if its existence can be confirmed. For this reason, it is difficult to take countermeasures, and such a machine difference is one of the biggest problems in the production line.

  However, when a plurality of devices are used in one process, it may be easy to detect an abnormality of the device. For example, if there is a production line consisting of a plurality of processes, only one device is used for each process. Then, it is assumed that a certain device has changed and a change has occurred in the final product quality. In such a case, it is difficult to determine which device has changed by simply monitoring the final performance of the product. However, if each process uses two devices in parallel, and there is a change in one of them, in this case, if there is even a history of device passage in each process of the product, each process By examining the device dependence on the final performance every time, the anomalous device can be identified.

  Of course, in practice, based on the concept of “self-process guarantee”, the basis of manufacturing line construction is to measure process performance for each process and prevent defective products from flowing into the next process. If the process performance is monitored, it is often possible to detect changes in the apparatus. However, factors other than the process performance being monitored may change due to a change in the apparatus, which may change the final performance. In such a case, anomalies cannot be detected in each process. The reason why the factor is not monitored is that measurement may not be possible, technical or cost may be difficult, and the factor may be unknown in the first place.

  As described above, when an unknown factor that varies the final product characteristics is included in the process, it is considered that a clue for estimating the factor may be obtained by examining the cause of the machine difference. However, finding the cause of the machine difference is often difficult. Therefore, it is important to grasp not only the machine difference at a certain time but also the time-series fluctuation of the machine difference. If you know when the machine difference has changed, you can collaborate with maintenance, repairs, abnormal actions, etc., performed near that point, and discover unknown causes of machine difference fluctuations and product performance. There is sex.

In many cases, the parallel use of a plurality of devices is unavoidable in terms of securing manufacturing capacity. At least it can be said that from the viewpoint of quality control, the difference information obtained from it should be used in reverse to help improve and maintain quality. In this way, using multiple devices in one process introduces a new problem of machine differences, but it is also true that there is new information that can be obtained from using multiple devices. . The information is anomaly generation process, and the other is substitution characteristics of unknown factors.
Therefore, a conventional method for grasping the machine difference will be described below.

(1) Single Regression Method Among the methods for grasping the machine difference, the simplest method pays attention to only one process and obtains the average of the final workmanship for each device in the process. By doing this, it may be possible to grasp that the final performance of a product group that has passed a specific device in the process is different from the overall final performance. Furthermore, by performing this analysis periodically, it is possible to plot and visualize the trend of the average final performance for each apparatus.

Here, in order to simulate the trend visualization state, a model of a production line is assumed.
FIG. 21 is a diagram illustrating a model of a production line.
As shown in FIG. 21, it is assumed that the production line includes three processes A, B, and C, and three apparatuses are used in parallel in each process. That is, in the process A, the devices A1, A2, and A3 are provided in parallel, in the process B, the devices B1, B2, and B3 are provided in parallel, and in the process C, the devices C1, C2, It is assumed that C3 is provided in parallel. Moreover, the use apparatus in each process shall be chosen at random every time.

  FIGS. 22A to 22C are graphs illustrating simulation results of final results classified for each apparatus used for processing, with a period on the horizontal axis and a final result on the vertical axis. a) shows the result classified for each apparatus in the process A, (b) shows the result classified for each apparatus in the process B, and (c) shows the result classified for each apparatus in the process C.

  In the production line having the configuration shown in FIG. 21, the first to third machines in each process have contributions of 1, 0, and −1 to the final product quality (for example, electrical characteristics). Differences in these contributions are machine differences. Also, the sum of these contributions in each process is the final product. The equipment used in each process was selected at random each time. Then, 1000 samples of data were created by simulation. At this time, from the 500th sample, the contribution of the apparatus A1 in the process A was changed from 1 to 3. The data was divided every 100 samples, and the average of the final performance for each device in each period was plotted with the horizontal axis as the period and the vertical axis as the performance average. For example, the item 5 on the horizontal axis is the analysis result for the data set from the 401st sample to the 500th sample. In FIG. 22, it can be seen that the change of the apparatus A1 in the process A can be visualized. Note that a method called analysis of variance is also known as a method for determining whether or not the difference in performance average of each device is statistically significant.

However, this simple regression method has the following problems.
This simple regression method is based on the premise that all variability factors other than equipment differences in the process of interest affect the products that have passed through any equipment in this process evenly. is there. As a factor that breaks this premise, there may be a case where there is a correlation in the apparatus used between processes. For example, the product processed by the apparatus A1 in the process A is often processed by the apparatus B2 in the process B. In such a case, a situation may occur in which the machine difference in the process A observed by the above-described method is actually the machine difference in the process B. In fact, the correlation of the devices used between the processes (hereinafter, also simply referred to as “correlation between devices”) exists in various production lines.

FIGS. 23A to 23C are graphs illustrating simulation results in a situation where there is a correlation between the devices used between the processes. The notation of the figure is the same as in FIG.
In the simulation shown in FIG. 23, in the case shown in FIG. 22 described above, the apparatus having the same number as that in the process A is used in the process B with a probability of 50%. The other conditions were the same as the simulation shown in FIG.
As shown in FIGS. 23 (a) to 23 (c), when there is a correlation between apparatuses worn between processes, only the apparatus A1 of the process A actually fluctuates, whereas the apparatus of the process B Observation results were obtained as if the apparatus B1 was also changing.

(2) Multiple regression method In order to avoid the above-mentioned problem, a method called quantification class I is known. For example, Patent Document 1 introduces a technique for statistically grasping the significance of machine difference by a technique based on quantification type I. This is a method of simultaneously estimating the machine difference of each process by the least square method from the passage history information and the final performance of a plurality of processes. This corresponds to multiple regression analysis when the explanatory variable is a continuous variable. In this method, first, a model for estimating the final result from the equipment used in each process is established. This is expressed by Equation 1 below.

However,
y: Final result p: Number of steps m (j): Number of devices in the j-th step a (j, i): Magnitude of contribution of the j-th step and i-th device to the final result u (j, i): Number Use of j step, i-th device (1: use, 0: not use)
For example, if device I is used in step J,
u (J, i) = 1 (i = I)
u (J, i) = 0 (i ≠ I)
b: Adjustment coefficient.

  In the mass production line, if there are measurement data of final finished products for a large number of products and passage history data of the apparatus in each process of each product, by using the method of quantification I, a (j , I) can be obtained. It can be said that the difference between a (j, i) for each device in the process j is a so-called machine difference with respect to the final result.

  However, a (j, i) for each device in the process j obtained by this method is meaningful in the difference between each other, but the absolute value of each is meaningless. This is because, as can be seen from the form of Equation 1, a (j, i) and b in each step can take infinite combinations of values for the same y. Therefore, for convenience of handling, if an appropriate numerical value is given to the estimated value of a (j, i) by the quantification class I, for example, a (j, 1) is set to 0, or a (j in step j , I) it is necessary to set an arbitrary constraint condition such that the average of i) is zero.

Next, in order to investigate the trend of machine difference, consider that the analysis by the quantification type I is periodically performed.
FIGS. 24A to 24C are graphs illustrating the simulation results by the quantification type I when there is no correlation between the apparatuses used between the processes. The notation of the figure is the same as in FIG.
FIGS. 25A to 25C are graphs illustrating the simulation results by the quantification type I when there is a correlation of the devices used between the processes. The notation of the figure is the same as in FIG.
In the simulation shown in FIG. 25, the correlation between the apparatuses is the same as the simulation shown in FIG.

  As shown in FIGS. 24 and 25, it can be seen that the machine difference analysis by the quantification type I clearly shows in which process the machine difference fluctuation occurs regardless of the presence or absence of correlation between apparatuses. However, it is not possible to specify which device has changed the degree of contribution. In the machine difference analysis by the quantification class I, a (j, i) in the above formula 1 has only a relative meaning, and therefore the estimated value of a (j, i) of each device is balanced with 0 as an average. It will change in a harmonious way. For example, in the example shown in FIG. 25, the true machine difference variation is that the average value of the final results of the products processed by the device A1 is increased by 2, and the average value of the final results of the products processed by the devices A2 and A3 is constant. Is. However, in the analysis results shown in FIGS. 24 and 25, the average of the device A1 is increased by 1.33, and the averages of the devices A2 and A3 are each decreased by 0.67. With this, it is difficult to correctly grasp the variation of the machine difference.

FIGS. 26A to 26C are graphs illustrating simulation results by the quantification type I when machine difference fluctuations occur more frequently than in FIG. The notation of the figure is the same as in FIG.
In this example, in the process A of the production line shown in FIG. 1, the contribution of the apparatus A1 is increased by 2 at the time of the 250th sample, and the contribution of the apparatus A2 is increased by 2 at the time of the 500th sample. It is assumed that the contribution of the device A3 has increased by 2 at the time of the sample.
However, in the result shown in FIG. 26 (a), the trend of each device is unclear, and it is difficult to say that the fluctuation of the machine difference can be correctly visualized.
As described above, in the conventional analysis based on the quantification type I, there is a problem that only a relative relationship between machine differences can be obtained, and a unique trend of each device cannot be known.

JP 2000-12640 A

  An object of the present invention is a process management method, a process management program, and a process management system for managing a production line in which a plurality of apparatuses are arranged in parallel in one process, and the contribution degree of each apparatus to the final product finish. To provide a process management method, a process management program, and a process management system capable of estimating a change.

  According to one aspect of the present invention, there is provided a process management method for managing a production line in which a plurality of devices are arranged in parallel in one process, wherein the data obtained from the production line is analyzed for quantification type I. And determining the relative relationship between the contributions of each of the devices to the final performance of the product in the production line and, based on the relative relationships, each of the devices so that the sum of changes in the contributions is minimized. There is provided a process management method characterized by estimating a change in the degree of contribution.

  According to another aspect of the present invention, there is provided a process management program for managing a production line in which a plurality of devices are arranged in parallel in one process, wherein the computer is configured to quantize data obtained from the production line. And a procedure for obtaining a relative relationship of contributions of the apparatus to the final performance of the product in the production line, and the sum of changes in the contributions is minimized based on the relative relationship. Thus, there is provided a process management program that executes a procedure for estimating a change in contribution of each device.

  According to still another aspect of the present invention, there is provided a process management system for managing a production line in which a plurality of devices are arranged in parallel in one process, and a data collection means for collecting data from the production line; Analyzing means for analyzing the quantification type I on the data to obtain a relative relationship of the contribution degree of the device to the final product performance in the production line, and based on the relative relation, There is provided a process management system comprising estimation means for estimating a change in contribution of each device so that a total sum of changes is minimized.

  According to the present invention, there are a process management method, a process management program, and a process management system for managing a production line in which a plurality of apparatuses are arranged in parallel in one process, and the contribution degree of each apparatus to the final workmanship of a product. A process management method, a process management program, and a process management system that can estimate a change can be realized.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
First, the concept of the embodiment of the present invention will be schematically described.
In the embodiment of the present invention, first, the data obtained from the production line is analyzed by the above-mentioned quantification type I, and the relative relationship of the degree of contribution each device has to the final product is obtained. The change (trend) in the contribution degree of each device is estimated from the physical relationship. For this purpose, the following concept is introduced. That is, it is hypothesized that machine difference fluctuations do not occur frequently and that machine differences of a plurality of devices rarely change simultaneously. Based on this hypothesis, he also stands in the position that "the observed event is considered to have happened most probably the most probable possibility". This is the same idea as the maximum likelihood method in statistics. That is, it is considered that the observed change in a (j, i) is caused by as little apparatus variation as possible. Based on these ideas, an algorithm described in detail in the following embodiment can be considered as the machine difference fluctuation waveform reproduction algorithm.

First, a first embodiment of the present invention will be described.
FIG. 1 is a block diagram illustrating a process management system according to this embodiment.
As shown in FIG. 1, the process management system 1 according to the present embodiment is a system that manages a production line 101. In the production line 101, a plurality of steps 102 are provided in series, and a plurality of devices 103 are arranged in parallel in each step 102, respectively. The plurality of apparatuses 103 arranged in one process 102 performs the same type of processing on the semi-finished product imported from the previous process in parallel independently of each other, and delivers the processed semi-finished product to the next process. . The production line 101 is, for example, an electronic device production line.

  In the process management system 1, a data collection unit 2 that collects and stores data from the production line 101 is provided. This data includes history information indicating which device 103 processed each product in which process 102, and the final performance of the product. The data collection means 2 is provided with storage means such as an HDD (Hard Disk Drive).

  In addition, the process management system 1 performs a quantification type I analysis on the data collected by the data collection unit 2 from the production line 101, and the degree of contribution each device 103 has to the final product quality in the production line 101. Analyzing means 3 for obtaining the relative relationship is provided. That is, the analysis unit 3 can calculate the relative relationship of the contributions of the plurality of apparatuses 103 belonging to a certain process 102. However, the contribution degree of each device 103 calculated at this time is a relative value. For example, the total contribution of all devices 103 belonging to a certain process 102 is a constant value.

  Further, the process management system 1 estimates the contribution change of each device 103 based on the relative relationship of the contribution degree calculated by the analysis unit 3 so that the sum of the change of contribution degree is minimized. Means 4 are provided. That is, the estimation unit 4 maintains the above-described relative relationship, and the magnitude of the change in the contribution of each device 103 is minimized so that the sum of the changes in the contribution of all the devices 103 belonging to a certain process 102 is minimized. Select the size. Thereby, the change of the contribution degree of each apparatus 103 is estimated. Moreover, the trend of the contribution degree of the apparatus 103 is estimated by continuously estimating the change of the contribution degree. The analysis unit 3 and the estimation unit 4 may be configured by dedicated hardware, respectively, or may be configured by a common general-purpose computer.

  Furthermore, the process management system 1 outputs the data collected by the data collection means 2, the analysis result by the analysis means 3, and the estimation result by the estimation means 4, and also outputs to the data collection means 2, the analysis means 3 and the estimation means 4. On the other hand, input / output means 5 for inputting commands and information is provided. The input / output means 5 is provided with, for example, a keyboard and a display.

Next, the operation of the process management system according to this embodiment configured as described above, that is, the process management method according to this embodiment will be described.
FIG. 2 is a flowchart illustrating the process management method according to this embodiment.

  First, as shown in step S <b> 1 of FIG. 2, the data collection unit 2 of the process management system 1 acquires final workmanship and history data of products manufactured in the manufacturing line 101 from the manufacturing line 101. This data is acquired periodically. Further, analysis and estimation described later on this data are also performed periodically.

Next, as shown in step S <b> 2, the analyzing unit 3 analyzes the data acquired by the data collecting unit 2 using the quantification type I. Specifically, the above-described data is substituted into the following formula 2, and an estimated value of x (j, i) is calculated. In addition, the following formula 2 is a formula obtained by replacing a (j, i) with x (j, i) in the above formula 1. That is,
y: Final result p: Number of steps m (j): Number of devices in the j-th step x (j, i): Magnitude of contribution to final result of the j-th step and i-th device u (j, i): Number Use of j step, i-th device (1: use, 0: not use)
For example, if the I equipment is used in the J process,
u (J, i) = 1 (i = I)
u (J, i) = 0 (i ≠ I)
b: Adjustment coefficient.
Thereby, x (j, i) is acquired as a relative value. That is, the relative relationship of the contribution degree of each device is obtained.

  Next, as shown in step S <b> 3, the estimation unit 4 estimates the change in contribution of each device based on this relative relationship. That is, among the above-mentioned x (j, i), a set of arbitrary one process is used as input data, and true movement of machine difference is estimated to convert the input data into output data. Let n be the number of samples of input data and m be the number of devices in the process. Further, as input data, the contribution of the j-th device with respect to the i-th sample is assumed to be x (i, j). Similarly, let w (i, j) be the output data. Tables 1 and 2 show the format of input data and output data.

  In step S3, the estimation means 4 increases / decreases this while maintaining the relative vertical relationship of the analysis results of each time, and finds a place where the total sum of changes in each contribution from the previous time is minimized. He concludes that this is what actually happened. This is based on the hypothesis that, as described above, the variation in machine difference is rare. Hereinafter, this process will be described in detail.

FIG. 3 is a flowchart illustrating the algorithm of the process shown in step S3 of FIG.
First, as shown in step S31 of FIG. 3, the sample number i (i is an integer from 1 to n) is set to i = 1. Since the sample numbers are given in time series, the sample number i also represents the period during which the product was manufactured.
Next, as shown in step S32, w (i, k) = x (i, k) is set. However, k is an integer of 1 to m.
Next, as shown in step S33, the numerical value of i is increased by 1. That is, i = i + 1.

Next, as shown in step S34, when the sum I (s) of the change in contribution is defined as the following formula 3 and formula 4 below, the sum I (s) of the change in contribution is minimized. The extra value s is obtained by a search method. The value of the added value s obtained in this way is defined as s _min .

FIG. 4 is a diagram illustrating the concept of the total change I (s) of the above-described contribution degrees.
As shown in FIG. 4, in step S34, the value of s is changed from −∞ to + ∞ to find s (s _min ) that minimizes I (s).

FIG. 5 is a diagram illustrating a case where the contribution degree of any of the three devices (device 1, device 2, and device 3) is fixed in FIG.
As shown in FIG. 5, as a result, I (s) takes the minimum value when the contribution of any device is fixed. For this reason, when the number of apparatuses arranged in a certain process is three, the three cases shown in FIG. 5 may be examined. For example, if s = −d (i, 1), s = −d (i, 2), or s = −d (i, 3), the contribution of the device 1, device 2 or device 3 is fixed, respectively. It will be done.

Next, as shown in step S35 of FIG. 3, w (i, k) is calculated by the following equation 5 using the value of s _min obtained in step S34.

  Next, the process proceeds to step S36, and if i = n, step S3 shown in FIG. On the other hand, if i <n, the process returns to step S33, and steps S33 to S36 are repeated. As a result, w (2, k), w (3, k),... W (n, k) and w (i, k) are sequentially obtained. However, w (1, k) is x (1, k).

  And the process shown to above-mentioned step S1-S3 is repeated regularly. Thereby, the trend of the contribution degree of each apparatus 103 is estimated. The estimation means 4 outputs the calculated value of w (i, k) to the input / output means 5, and the input / output means 5 outputs this result. For example, the input / output unit 5 displays the estimation result on a display.

  According to the present embodiment, with respect to the relative relationship of the machine differences obtained by the quantification class I analysis, the hypothesis that the machine differences of a plurality of devices rarely change at the same time is introduced, and the contribution degree of each device By making the total sum of the changes in the minimum, it is possible to estimate the change in the contribution of each device to the final product performance. As a result, for example, when the degree of contribution of an apparatus in a certain process changes suddenly at a certain point in time, it can be estimated that some abnormality has occurred in this apparatus, and measures can be taken. Thereby, process management of the production line 101 can be performed.

As described above, the analysis unit 3 and the estimation unit 4 may be configured by a common general-purpose computer. In this case, the process management described above is executed by a process management program.
This process management program is a process management program for managing a production line 101 in which a plurality of devices 103 are arranged in parallel in one process 102, and is a program for causing a computer to execute the following procedure.
(1) A procedure in which the data obtained from the production line 101 is subjected to quantification type I analysis to determine the relative relationship of the degree of contribution that the apparatus 101 has to the final product quality in the production line 101 (step S2).
(2) A procedure for estimating the change in contribution of each device 103 based on this relative relationship so that the sum of the change in contribution is minimized (step S3).

Moreover, estimation of the change of the contribution degree of each apparatus mentioned above (step S3) makes this computer perform the following procedures.
(2-1) Procedure for finding an added value s _min that is a common added value s to be added to the contribution degree of each device 103 and that takes a minimum value of the total change I (s) of the contribution degree (step S34) ).
(2-2) A procedure for estimating a change in the degree of contribution of each device 103 using the additional value s _min (step S35).

Next, a second embodiment of the present invention will be described.
A diagram illustrating the process management system according to the present embodiment is the same as FIG.
FIG. 6 is a flowchart illustrating a part of the process management method according to this embodiment.
A flowchart illustrating the entire process management method according to the present embodiment is the same as FIG. FIG. 6 illustrates the algorithm shown in step S3 of FIG.
7A to 7C are diagrams illustrating the process of step S37 shown in FIG.
FIGS. 8A and 8B are diagrams exemplifying steps S38 and S39 shown in FIG.

  The process management method according to the present embodiment is different from the first embodiment described above in the content of step S3. That is, as shown in FIG. 6, in this embodiment, steps S37 and S39 are provided instead of steps S34 and S35, and step S38 is provided between steps S37 and S39.

  In the present embodiment, similarly to the first embodiment described above, steps S31 to S33 shown in FIG. 6 are performed after the steps S1 and S2 shown in FIG. That is, data is read from the production line (step S1) and analyzed by quantification type I (step S2). Thereafter, the sample number (period) i (i is an integer from 1 to n) is set to 1 (step S31), w (i, k) = w (i, k) (step S32), and i = i + 1 ( Step S33).

Then, as shown in step S37, when the sum I (s) of the change in contribution is defined as in the following Equation 6, an added value s that minimizes the sum I (s) of the change in contribution is obtained. Obtained by a search method. However, k is an integer of 1 to m. The value of the added value s at this time is s _min .

At this time, the value of d (i, k) is obtained for each case as follows. That is,
(1) If neither x (i, k) nor x (i-1, k) is missing data, the following formula 7 is used.
(2) When x (i, k) is not missing data and x (i-1, k) is missing data, it is obtained by the following formula 8.
Note that “missing data” refers to data that does not have a numerical value, for example, because a device corresponding to the data is not used.

For example, as shown in FIG. 7A, it is assumed that the number of devices is 4, and the previous data of the device 2 and the current data of the device 3 are missing data. In this case, as shown in FIG. 7B, for this time, a search for s _min is performed using the data of the devices 1, 2, and 4 that are not missing data by the same method as in the first embodiment. Do. However, for the device 2 whose previous data was missing data, the previous value is the same as this time. On the other hand, the device 3 in which the current data is missing data is not searched for s _min . Then, as shown in FIG. 7C, w is calculated for the devices 1 and 4 using the value of s _min obtained in FIG. 7B.

Next, as shown in step S38, the value of a is obtained by the following formula 9.
However, a1 is the average value of x (i, k) for k where x (i, k) and x (i-1, k) are not missing data.
Further, a2 is an average value of {w (i−1, k) + d (i, k) + s _min } for k in which x (i, k) and x (i−1, k) are not missing data. It is.

  That is, as shown in FIG. 8A, for the devices 1 and 4 in which data exists for both the previous time and the current time, an average value of x (i) is obtained and the value is set to a1. In addition, for the devices 1 and 4, an average value of w (i) is obtained, and the value is set to a2. And the value of a is calculated | required by the said Numerical formula 9.

Next, as shown in step S39, the value of w (i, k) is obtained. At this time, it is obtained separately for each case as follows.
(1) When neither x (i, k) nor x (i-1, k) is missing data, the value of w (i, k) is obtained by the following formula 10.
(2) When x (i, k) is not missing data and x (i-1, k) is missing data, the value of w (i, k) is obtained by the following formula 11.

  For example, as shown in FIG. 8B, for the devices 1 and 4 in which data exists for both the previous time and the current time, w (i, k) is calculated by the above formula 10, and the previous data is missing data. For a certain device 2, w (i, k) is calculated by the above equation 11. Note that w (i, k) is not calculated for the device 3 in which the current data is missing data.

  Next, the process proceeds to step S36. If i = n, step S3 (see FIG. 2) is terminated, and if i <n, the process returns to step S33. The process management method other than the above in this embodiment is the same as that in the first embodiment.

  In an actual production line, there may be a device that is not used at all for a certain period due to reasons such as device maintenance, inspection or repair, or production adjustment. In this case, the data of the device is missing data during that period. At this time, if a periodic machine difference analysis based on the quantification type I is performed, the analysis result is displayed every time so that the average value of the contributions to the devices excluding the devices that have not been used for that period is zero. The For this reason, the analysis result fluctuates due to the influence of use / nonuse of the apparatus, and the result becomes more difficult to see. For example, even if there is actually no machine difference fluctuation, a wavy graph is output.

  Therefore, in the present embodiment, processing in consideration of missing data is performed in steps S37 to S39. Specifically, for the device that was previously missing data but not this time, the above processing is performed without using the device data. The current value of this device is shifted by the change in the average value of the values before and after the conversion of other devices. Thereby, even if there is missing data, the analysis result does not fluctuate, and a highly reliable result can be obtained.

Hereinafter, examples of the above-described embodiments will be described.
First, the first embodiment will be described.
This example is an example of the basic operation in the first embodiment described above.
FIGS. 9A to 9C are graphs showing simulation results of the first embodiment, with the horizontal axis representing the period and the vertical axis representing the final workmanship, and FIG. The result classified for every apparatus is shown, (b) shows the result classified for every apparatus of the process B, (c) shows the result classified for every apparatus of the process C.

In the simulation of the first embodiment, as in the case shown in FIG. 26, the influence of the apparatus A1 increases by 2 at the time of the 250th sample in the process A of the production line shown in FIG. It is assumed that the influence degree of the apparatus A2 increases by 2 at the time point, and the influence degree of the apparatus A3 increases by 2 at the time point of the 750th sample.
As shown in FIGS. 9A to 9C, in this embodiment, the trend of each device is clear, and the variation of the machine difference can be visualized accurately.

Next, a second embodiment will be described.
This example is an example in which there is noise in data in the first embodiment.
FIG. 10 is a diagram showing a production line assumed in this embodiment.
As shown in FIG. 10, in this embodiment, the production line is composed of five steps A to E. In each process, three devices of Unit 1 to Unit 3 are arranged, and each device has contributions of (+1), 0, and (−1), respectively. The machine difference analysis was performed only for the processes A, B, and C as in the first embodiment. That is, the influence of the processes D and E acts as noise on the final product. The variation in contribution of each device was the same as in the first example (see FIG. 9).

FIGS. 11A to 11C are graphs showing analysis results by quantification type I in this embodiment, with the horizontal axis representing the period and the vertical axis representing the final workmanship, and FIG. The result classified for every apparatus of A is shown, (b) shows the result classified for every apparatus of process B, (c) shows the result classified for every apparatus of process C.
As shown in FIG. 11, depending on the quantification class I analysis, it was impossible to accurately evaluate the variation in the contribution of each device.

12A to 12C are graphs showing simulation results of the second embodiment, with the horizontal axis representing the period and the vertical axis representing the final workmanship, and FIG. The result classified for every apparatus is shown, (b) shows the result classified for every apparatus of the process B, (c) shows the result classified for every apparatus of the process C.
As shown in FIG. 12, according to the method according to the first embodiment, it was possible to evaluate the variation of the contribution degree of each device fairly accurately even when the final result includes noise.

Next, a third embodiment will be described.
This example is an example of the above-described second embodiment. In other words, this is an embodiment in the case where there is a deficiency in data for a certain period due to temporary interruption of use of the apparatus.
In this embodiment, it is assumed that there are five devices in the process A, and the device A1 and the device A2 are not used during the period from the 201st sample to the 400th sample, and the device A4 and the device are used after the 601st sample. It was assumed that A5 was not used. However, it was assumed that there was no machine difference variation, that is, no change in the contribution of each device.

FIG. 13 is a graph showing the result of analysis by quantification type I in this example, with the horizontal axis representing the period and the vertical axis representing the final performance.
As shown in FIG. 13, in the analysis result by the quantification type I, although the machine difference is actually constant, the final result value fluctuates due to switching between use and non-use of the apparatus. Therefore, in the analysis based on the quantification type I, it is impossible to accurately reproduce the change in the contribution degree of each device, that is, the difference in machine difference.

FIG. 14 is a graph showing the simulation results of the third example, with the horizontal axis representing the period and the vertical axis representing the final performance.
As shown in FIG. 14, according to the method of the second embodiment, even if there is missing data, it is possible to accurately reproduce the variation of the machine difference.

Next, a fourth embodiment will be described.
This embodiment is an embodiment in which the same machine difference variation as that of the second embodiment (see FIG. 12) is added to the devices A1, A2, and A3 in the third embodiment.
FIG. 15 is a graph showing the result of analysis by quantification type I in this example, with the horizontal axis representing the period and the vertical axis representing the final workmanship.
FIG. 16 is a graph showing the simulation results of the fourth example, with the horizontal axis representing the period and the vertical axis representing the final performance.

  As shown in FIG. 15 and FIG. 16, the machine difference variation could not be accurately reproduced by the quantification type I analysis. However, according to the method of the second embodiment, there is missing data and Even when the difference fluctuated, the machine difference fluctuation could be accurately reproduced.

Next, a fifth embodiment will be described.
The present embodiment is an embodiment in the case where noise caused by an unobserved process is mixed in the fourth embodiment described above, as in the second embodiment described above.
FIG. 17 is a graph showing the analysis result by the quantification type I in this example, with the horizontal axis representing the period and the vertical axis representing the final performance.
FIG. 18 is a graph showing the simulation results of the fifth example, with the horizontal axis representing the period and the vertical axis representing the final performance.
As shown in FIGS. 17 and 18, also in this example, the method according to the second embodiment had better reproducibility of the machine difference variation waveform than the analysis of quantification class I.

Next, a sixth embodiment will be described.
In this example, the hypothesis introduced in the above-described embodiment, that is, the hypothesis that “variation of machine differences does not occur frequently and that machine differences of a plurality of devices rarely change simultaneously” does not hold. Example of the case.
In the present embodiment, in the situation assumed in the fifth embodiment (see FIGS. 17 and 18), the influences of the apparatuses A1 and A2 both increase by 2 at the time of the 250th sample in step A, and 500 It was assumed that the degree of influence of the device A3 increased by 2 at the time of the sample. The analysis / estimation described in the first embodiment was performed.

FIG. 19 is a graph showing the result of analysis by quantification type I in this example, with the horizontal axis representing the period and the vertical axis representing the final performance.
FIG. 20 is a graph showing the simulation results of the sixth example, with the period on the horizontal axis and the final performance on the vertical axis.

  As described above, in the present embodiment, two of the three devices (devices A1 and A2) change in the same manner in the 250th sample, that is, in the third period. In such a case, in the analysis result of the quantification type I shown in FIG. 19, the two units (devices A1 and A2) change slightly and the remaining one (device A3) changes greatly in the opposite direction. It has become. In the method according to the first embodiment described above, this tendency is further emphasized, and only the remaining one (device A3) has a profile that is greatly changed in the opposite direction. In either case, the original waveform has not been reproduced, but the relative relationship between the influence levels of the three units is preserved, and it can be seen that the devices A1 and A2 behaved differently from the device A3. Originally, the information obtained from the analysis result by the quantification class I is only the information on the relative relationship. Therefore, even when the situation assumed in the embodiment of the present invention does not hold, the waveform reproducibility is worse than before application. I can say no.

  As described above, in the above-described embodiments and examples, a method for reproducing a fluctuation waveform of a machine difference in a product production line, for example, an electronic device production line, is provided, and its effectiveness is shown using numerical examples. It was. The method according to the present embodiment is based on repeating the machine difference analysis method using the quantification type I which is a conventional method on a regular basis. It is characterized by high reproducibility. In this post-processing, it is assumed that the machine difference fluctuation occurs independently and suddenly in each device, and attempts to reproduce the machine difference waveform based on the concept of the maximum likelihood method. As a result, it was confirmed for several numerical examples that the machine difference fluctuation waveform that could not be reproduced by the conventional method could be reproduced. It was also shown that the assumptions made here are not always true, but the degradation of waveform reproducibility when the assumptions are not true is limited to the level before application of this method.

  The present invention has been described above with reference to the embodiment. However, the present invention is not limited to this embodiment. For example, those in which those skilled in the art appropriately added, deleted, and changed the design of the above-described embodiments are also included in the scope of the present invention as long as they have the gist of the present invention.

1 is a block diagram illustrating a process management system according to a first embodiment of the present invention. It is a flowchart figure which illustrates the process management method which concerns on 1st Embodiment. FIG. 3 is a flowchart illustrating an example of an algorithm of processing shown in step S3 of FIG. It is a figure which illustrates the concept of total I (s) of the change of a contribution. In FIG. 4, it is a figure which illustrates the case where the contribution of either of three apparatuses is fixed. It is a flowchart figure which illustrates a part of process management method which concerns on the 2nd Embodiment of this invention. (A) thru | or (c) is a figure which illustrates the process of step S37 shown in FIG. (A) And (b) is a figure which illustrates the process of step S38 and S39 shown in FIG. (A) thru | or (c) is a graph which shows the simulation result of a 1st Example, taking a period on a horizontal axis and taking a final result on a vertical axis | shaft, (a) is for every apparatus of the process A. The result of classification is shown, (b) shows the result of classification for each apparatus in the process B, and (c) shows the result of classification for each apparatus in the process C. It is a figure which shows the manufacturing line assumed in the 2nd Example. (A) thru | or (c) is a graph which shows the analysis result by the quantification I class in a 2nd Example, taking a period on a horizontal axis and taking a final result on a vertical axis | shaft, (a) is a process. The result classified for every apparatus of A is shown, (b) shows the result classified for every apparatus of process B, (c) shows the result classified for every apparatus of process C. (A) thru | or (c) is a graph which shows a simulation result of a 2nd Example, taking a period on a horizontal axis and taking a final result on a vertical axis | shaft, (a) is for every apparatus of the process A. The result of classification is shown, (b) shows the result of classification for each apparatus in the process B, and (c) shows the result of classification for each apparatus in the process C. It is a graph which shows the analysis result by the quantification I class in a 3rd Example, taking a period on a horizontal axis and taking a final result on a vertical axis | shaft. It is a graph which shows the simulation result of a 3rd Example, taking a period on a horizontal axis and taking a final workmanship on a vertical axis | shaft. It is a graph which shows the analysis result by the quantification type I in a 4th Example by taking a period on a horizontal axis and taking a final result on a vertical axis | shaft. It is a graph which shows the simulation result of a 4th Example, taking a period on a horizontal axis and taking the final result on a vertical axis | shaft. It is a graph which shows the analysis result by the quantification I class in a 5th Example, taking a period on a horizontal axis and taking the final result on a vertical axis | shaft. It is a graph which shows the simulation result of a 5th Example, taking a period on a horizontal axis and taking the final result on a vertical axis | shaft. It is a graph which shows the analysis result by the quantification I class in a 6th Example, taking a period on a horizontal axis and taking the final result on a vertical axis. It is a graph which shows the simulation result of a 6th Example, taking a period on a horizontal axis and taking the final result on a vertical axis | shaft. It is a figure which illustrates the model of a manufacturing line. (A) thru | or (c) is a graph which illustrates the simulation result of the final result classified into every apparatus used for the process, taking a period on a horizontal axis and taking a final result on a vertical axis | shaft, (a) Shows the result of classification for each apparatus of the process A, (b) shows the result of classification for each apparatus of the process B, and (c) shows the result of classification for each apparatus of the process C. (A) thru | or (c) is a graph which illustrates the simulation result of the situation where a use apparatus has a correlation between processes, (a) shows the result classified for every apparatus of the process A, (b) The result classified for every apparatus of the process B is shown, (c) shows the result classified for every apparatus of the process C. (A) thru | or (c) are the graphs which illustrate the simulation result by quantification class I, when there is no correlation of the apparatus used between processes, (a) is the result classified for every apparatus of the process A. (B) shows the result classified for every apparatus of the process B, (c) shows the result classified for every apparatus of the process C. (A) thru | or (c) is a graph which illustrates the simulation result by quantification class I about the case where there exists a correlation of the apparatus used between processes, (a) is the result classified for every apparatus of the process A. (B) shows the result classified for every apparatus of the process B, (c) shows the result classified for every apparatus of the process C. (A) thru | or (c) is a graph which illustrates the simulation result by the quantification I class in case machine difference fluctuation | variation has occurred more frequently than FIG. 24, (a) is for every apparatus of the process A. The result of classification is shown, (b) shows the result of classification for each apparatus in the process B, and (c) shows the result of classification for each apparatus in the process C.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Process management system, 2 Data collection means, 3 Analysis means, 4 Estimation means, 5 Input / output means, 101 Production line, 102 Process, 103 Apparatus

Claims (7)

A process management method for managing a production line in which a plurality of devices are arranged in parallel in one process,
Analyzing the quantification type I on the data obtained from the production line, and determining the relative relationship of the contribution degree each device has to the final product performance in the production line,
Based on the relative relationship, the change in contribution of each device is estimated so that the sum of the change in contribution is minimized.
A process management method characterized by that. The estimation of the change in contribution of each device is as follows:
A common addition value to be added to the contribution of each device, and find an addition value such that the sum of changes in the contribution takes a minimum value,
The process management method according to claim 1, wherein a change in contribution of each device is estimated using the added value. In the step of finding the added value, the analysis result of the quantification class I relating to the k-th device of the i-th sample is set to x (i, k), the added value is set to s, and the total change in contribution is calculated. When I (s) is set, an added value s _min that minimizes the sum I (s) given by the following equation is found by a search method.

In the step of estimating the change in contribution of each device, when the contribution is w (i, k), the added value s _min is substituted into the following equation to calculate the contribution w (i, k). Calculate

The process management method according to claim 2, wherein: In the step of finding the added value,
When the analysis result of the quantification class I regarding the k-th device of the i-th sample is x (i, k), the added value is s, and the total change in contribution is I (s),
If neither x (i, k) nor x (i-1, k) is missing data, an additional value s _min that minimizes the sum I (s) given by the following equation is found by a search method. ,

When x (i, k) is not missing data and x (i-1, k) is missing data, an additional value s _min that minimizes the sum I (s) given by the following equation. By searching,

In the process of estimating the change in contribution of each device,
The average value of x (i, k) for k where x (i, k) and x (i-1, k) are not missing data is a1, and x (i, k) and x (i-1, k) When the average value of {w (i−1, k) + d (i, k) + s _min } for k which is not missing data is a2, and a = a2−a1,
If neither x (i, k) and x (i-1, k) are missing data, the contribution w (i, k) is calculated by the following equation:

When x (i, k) is not missing data and x (i-1, k) is missing data, the contribution w (i, k) is calculated by the following formula.

The process management method according to claim 2, wherein: A process management program for managing a production line in which a plurality of devices are arranged in parallel in one process,
On the computer,
A procedure for performing a quantification class I analysis on the data obtained from the production line to determine a relative relationship of contributions of the device to the final product quality in the production line;
A step of estimating a change in contribution of each device based on the relative relationship so that a sum of changes in the contribution is minimized;
A process management program characterized by causing The estimated change in contribution of each device is sent to the computer,
A common addition value to be added to the contribution of each device, and a procedure for finding an addition value such that the sum of changes in the contribution takes a minimum value;
The process management program according to claim 5, further comprising: a step of estimating a change in contribution of each device using the added value. A process management system for managing a production line in which a plurality of devices are arranged in parallel in one process,
Data collection means for collecting data from the production line;
Analyzing means for performing quantification class I analysis on the data to determine a relative relationship of contributions of the apparatus to the final product performance in the production line;
Estimating means for estimating a change in contribution of each device based on the relative relationship so that a sum of changes in the contribution is minimized;
A process management system characterized by comprising:

Images