JP6385914B2 - Quality monitoring system and quality monitoring method - Google Patents

Description

  The present invention relates to a quality monitoring system that manages the quality of products manufactured by equipment and a quality monitoring method that uses the quality monitoring system. For example, a quality monitoring system that manages the quality of products that are continuously mass-produced and a quality monitoring method that uses the quality monitoring system. It can be used suitably.

  In equipment for manufacturing products, it is desirable to set suitable conditions for producing target non-defective products. However, in practice, the environmental parameters at the manufacturing site are not always set. This is partly due to the fact that a plurality of management items such as temperature and humidity in equipment in equipment and raw materials of products fluctuate every moment.

  Therefore, in managing the quality of the product, it is important to feed back the analysis result of the finished product to the production site. If the finished product is a defective product, it is expected to improve the yield by analyzing the manufacturing history data, investigating the cause, looking for measures to prevent recurrence, and improving various conditions at the production site.

  However, for example, in the case of a complex product manufactured through a plurality of processes, this feedback takes time. In particular, in the case of a product that is continuously mass-produced, there is a possibility that the mass production of defective products may be continued for a long time before it is confirmed that the finished product is defective. Note that mass production of defective products can be suppressed if production is stopped and a pass / fail judgment is made each time a product is completed, but long-term production suspension is not preferable in terms of productivity.

  Furthermore, the manufacturing condition parameters of the equipment include a wide variety of management items, and these management items may have a mutual influence relationship. In such a case, as a result of changing the setting in a direction to improve one management item, another management item may be changed in an undesired direction. Therefore, it is preferable to comprehensively manage all management items included in the manufacturing condition parameters.

  The MT method (Mahalanobis Taguchi method) is known as a method for comprehensively managing a plurality of items that affect each other. Furthermore, according to the MT method, by monitoring data of a system such as a production facility, it is possible to grasp the tendency before an abnormality occurs in the system.

  In relation to the above, Patent Document 1 (Japanese Patent Laid-Open No. 2006-23214) discloses a description relating to a method for determining the presence or absence of an abnormality in a measurement reaction process. This judgment method consists of a step of creating a reference Mahalanobis space from time series data of reaction processes that have been measured successfully, and a reaction process divided into a plurality of predetermined processes, and a new process that is divided into a plurality of processes. And creating a Mahalanobis space, and calculating a Mahalanobis distance in the Mahalanobis space at the end of a plurality of predetermined processes.

JP 2006-23214 A

  Provided are a quality monitoring system and a quality monitoring method using the same, which are further improved from the conventional MT method. Other problems and novel features will become apparent from the description of the specification and the accompanying drawings.

  The means for solving the problem will be described below using the numbers used in the (DETAILED DESCRIPTION). These numbers are added to clarify the correspondence between the description of (Claims) and (Mode for Carrying Out the Invention). However, these numbers should not be used to interpret the technical scope of the invention described in (Claims).

  According to one aspect, the quality monitoring system of the present invention includes a data recording unit (31), a Mahalanobis distance calculation unit (361), and a provisional evaluation unit (36). Here, the data recording unit (31) records log data of manufacturing condition parameter values when manufacturing a product. The Mahalanobis distance calculation unit (361) calculates a Mahalanobis distance of manufacturing condition data that is a set of manufacturing condition parameter values included in the log data when a specific product is manufactured. The temporary evaluation unit (36) performs a temporary evaluation of the quality of a specific product by comparing the Mahalanobis distance with a threshold value. According to this configuration, the quality of products that are continuously mass-produced can be tentatively evaluated using the MT method at the time of manufacture, so that feedback can be performed faster than when only quality inspection is used.

  According to another aspect, the quality monitoring system of the present invention may further include a threshold update unit (35). Here, the data recording unit (31) further records quality data indicating the result of quality inspection of the manufactured product and determining that it is a non-defective product or a defective product, and the threshold update unit (35) includes log data, quality data, and Update threshold based on Mahalanobis distance. When the minimum value of the Mahalanobis distance corresponding to the defective product is smaller than the maximum value of the Mahalanobis distance corresponding to the non-defective product, the threshold update unit (35) sets the threshold to be updated from the range between the minimum value and the maximum value. If the minimum value is greater than or equal to the maximum value, a threshold value to be updated is selected from the range between the maximum value and the minimum value. According to this configuration, the threshold value is changed, and the accuracy of the temporary evaluation can be improved.

  According to still another aspect, in the quality monitoring system of the present invention, the threshold update unit (35) updates the threshold further based on the weighted F value. The weighted F value is a harmonic average value of the relevance rate weighted with the first coefficient and the recall rate weighted with the second coefficient. The conformity rate is a ratio at which a product that has been determined to be acceptable in the provisional evaluation is determined to be a non-defective product in the quality inspection. The recall is a ratio at which a product determined to be a non-defective product in the quality inspection was determined to be acceptable in the provisional evaluation. According to this configuration, an appropriate threshold value can be calculated using the weighted F value calculated by weighting the relevance ratio and the recall ratio, and the accuracy of the temporary evaluation can be improved.

  According to still another aspect, the quality monitoring system of the present invention may further include a coefficient storage unit and a coefficient selection unit. Here, the coefficient storage unit stores a plurality of combinations of the first coefficient and the second coefficient. The coefficient selection unit selects an optimal combination from a plurality of combinations according to the Mahalanobis distance. According to this configuration, an appropriate threshold can be calculated for the calculated weighted F value according to the situation of the manufacturing facility, and the accuracy of the temporary evaluation can be improved.

  According to still another aspect, the monitoring system of the present invention may further include a control unit (32), a contribution rate calculation unit (38), and a Mahalanobis distance calculation formula creation unit (37). Here, a control part (32) changes manufacturing conditions based on the result of temporary evaluation. The contribution rate calculation unit (38) calculates the contribution rate to the Mahalanobis distance for each of the plurality of manufacturing condition parameters. The control unit (32) selects some of the plurality of manufacturing condition parameters according to the calculated contribution rate. The Mahalanobis distance calculation formula creation unit (37) creates a new calculation formula for the Mahalanobis distance using the selected partial manufacturing condition parameters. According to this configuration, a highly accurate Mahalanobis distance calculation formula and a threshold value can be created.

  According to still another aspect, the monitoring system of the present invention may further include a display device (321). Here, the display device (321) highlights the graph showing the change over time of the manufacturing condition parameter value according to the contribution rate and the grace period until the limit value of the manufacturing condition parameter value. According to this configuration, it is possible to easily recognize the priority order when changing the manufacturing condition parameters.

  According to the embodiment, the quality of a product that is continuously mass-produced can be tentatively evaluated using the MT method at the time of manufacture, so that feedback can be performed faster than when only quality inspection is used. As a result, an improvement in yield is expected.

FIG. 1 is a scatter diagram illustrating an example of a quality analysis method according to the prior art. FIG. 2A is a block diagram illustrating a configuration example of the quality monitoring system according to the first embodiment. FIG. 2B is a block diagram illustrating a configuration example of a general computer. FIG. 3 is a flowchart showing an operation example of the quality monitoring method according to the first embodiment. FIG. 4 is a graph illustrating the S / N ratio exemplified in Table 1. FIG. 5 is a diagram illustrating an example of a method for displaying manufacturing condition parameters that are candidates for change in the quality monitoring system according to the first embodiment. FIG. 6 is a flowchart showing an operation example of updating the Mahalanobis distance calculation formula by the quality monitoring method according to the first embodiment. FIG. 7 is a diagram for explaining the principle of calculating a threshold in the quality monitoring system according to the second embodiment. FIG. 8 is a flowchart showing an operation example of a method for calculating an optimum threshold in the quality monitoring system according to the second embodiment.

  With reference to the accompanying drawings, a quality monitoring system and a quality monitoring method using the quality monitoring system according to the present invention will be described below.

  In the present invention, the MT method (Mahalanobis Taguchi method) is used for continuous mass production of the same product. In order to better understand the present invention, the reason why the MT method is used will be described first, and then the details of each embodiment will be described.

  Although it will be described repeatedly, it is difficult to comprehensively manage a plurality of items that affect each other, and such management is relatively easy if the MT method is used.

  FIG. 1 is a scatter diagram illustrating an example of a quality analysis method according to the prior art. In the scatter diagram of FIG. 1, the horizontal axis indicates serial numbers of a plurality of the same products continuously produced in the order of production. The vertical axis indicates one measured value of the manufacturing condition parameter value of the product.

  In the scatter diagram of FIG. 1, when the product having the serial number is a non-defective product at the intersection of the vertical line passing through the serial number on the horizontal axis and the horizontal line passing through the measured value on the vertical axis, Is drawn, and if it is a defective product, a “×” -shaped mark is drawn.

  From the scatter diagram of FIG. 1, it is desired to read the conditions under which good or defective products are likely to be manufactured in association with the serial number on the horizontal axis and the manufacturing condition parameter value on the vertical axis. In the example of FIG. It is difficult.

  Thus, even if it analyzes for every individual manufacturing condition parameter among a plurality of manufacturing condition parameters, it is difficult to obtain a meaningful result.

(First embodiment)
FIG. 2A is a block diagram illustrating a configuration example of the quality monitoring system according to the first embodiment. The components of the quality monitoring system shown in FIG. 2A will be described.

  The quality monitoring system in FIG. 2A includes a manufacturing facility 1, a facility monitor 2, a quality monitoring system server 3, and a quality inspection unit 4. Of the components shown in FIG. 2A, the main body of the quality monitoring system is the quality monitoring system server 3, and the same quality monitoring system server 3 can be combined with other manufacturing equipment, equipment monitor and quality inspection unit. It is.

  2A includes a data recording unit 31, a control unit 32, a display device 321, a non-defective product data recording unit 33, a defective product data recording unit 34, a threshold update unit 35, and a temporary evaluation unit. 36, a Mahalanobis distance calculation formula creation unit 37, and a contribution rate calculation unit 38. The temporary evaluation unit 36 includes a Mahalanobis distance calculation unit 361 and a threshold storage unit 362. The Mahalanobis distance calculation formula creation unit 37 may be a computer that executes the Mahalanobis distance calculation formula creation program 371.

  The connection relationship of the components shown in FIG. 2A will be described. The manufacturing facility 1 manufactures the product 11. Here, a case where the manufacturing equipment 1 mass-produces the same product 11 will be described. It is preferable that different serial numbers are assigned to the products 11 to be manufactured.

  The production facility 1 and the facility monitor 2 are connected to each other so that the latter monitors the former. The equipment monitor 2 is connected to a data recording unit 31 that is a transmission destination of log data obtained by monitoring the manufacturing equipment 1.

  Table 1 is a table showing an example of log data. The log data illustrated in Table 1 includes the serial number of the product 11 and a plurality of manufacturing condition parameters.

  The equipment monitor 2 is further connected to the control unit 32, and transmits manufacturing condition data corresponding to a specific product 11 among the log data to the control unit 32. The manufacturing condition data is a set of manufacturing condition parameter values when the specific product 11 is manufactured.

  Table 2 is a table showing an example of manufacturing condition data. The manufacturing condition data illustrated in Table 2 is equivalent to the last one line extracted from the log data shown in Table 1.

  The quality inspection unit 4 performs a quality inspection of the product 11 manufactured by the manufacturing facility 1. The quality inspection unit 4 is connected to the data recording unit 31 and transmits quality data to the data recording unit 31. The quality data indicates the result of quality inspection of the product 11.

  A non-defective product data recording unit 33 and a defective product data recording unit 34 are connected to the subsequent stage of the data recording unit 31. The data recording unit 31 transmits good product data among the log data to the good product data recording unit 33. The non-defective product data is manufacturing condition data corresponding to the product 11 determined as a non-defective product as a result of the quality inspection. Further, the data recording unit 31 transmits defective product data of the log data to the defective product data recording unit 34. The defective product data is manufacturing condition data corresponding to the product 11 determined as a defective product as a result of the quality inspection. The data recording unit 31 is further connected to the control unit 32.

  Table 3 is a table showing an example of non-defective product data. The non-defective product data exemplified in Table 3 is equivalent to a part of the log data shown in Table 1 extracted.

  Table 4 is a table showing an example of defective product data. The defective product data exemplified in Table 4 is equivalent to the log data shown in Table 1 extracted from the rows other than the non-defective product data.

  The non-defective product data recording unit 33 is connected to the threshold update unit 35 and transmits the non-defective product data to the threshold update unit 35. The defective product data recording unit 34 is also connected to the threshold update unit 35 and transmits defective product data to the threshold update unit 35.

  The non-defective product data recording unit 33 is also connected to the Mahalanobis distance calculation formula creating unit 37 and transmits the non-defective product data to the Mahalanobis distance formula creating unit 37.

  The defective product data recording unit 34 is also connected to a contribution rate calculation unit 38 and transmits defective product data to the contribution rate calculation unit 38.

  The threshold update unit 35 is connected to the threshold storage unit 362 and transmits the threshold to the threshold storage unit 362.

  The Mahalanobis distance calculation formula creation unit 37 is connected to the Mahalanobis distance calculation unit 361 and transmits the Mahalanobis distance calculation formula to the Mahalanobis distance calculation unit 361.

  The contribution rate calculation unit 38 and the Mahalanobis distance calculation unit 361 are connected to each other so that the former requests the latter to calculate the Mahalanobis distance, and the latter transmits the calculated Mahalanobis distance to the former. The contribution rate calculation unit 38 is also connected to the control unit 32 and transmits the contribution rate to the control unit 32.

  The control unit 32 and the temporary evaluation unit 36 are connected to each other so that the former requests the latter to make a temporary evaluation and the latter transmits the temporary evaluation to the former. The control unit 32 is also connected to the display device 321 and transmits display data to the display device 321. Further, the control unit 32 is also connected to the manufacturing facility 1 and transmits a control signal including the set value of the manufacturing condition parameter to the manufacturing facility 1.

  Some or all of the components shown in FIG. 2A can be realized as a general computer.

  FIG. 2B is a block diagram illustrating a configuration example of a general computer 100. The computer 100 illustrated in FIG. 2B includes a bus 110, an arithmetic device 120, a storage device 130, an input / output interface 140, and an external storage device 150.

  The arithmetic device 120, the storage device 130, the input / output interface 140, and the external storage device 150 are connected to the bus 110 and can communicate with each other via the bus 110.

  The arithmetic device 120 executes a predetermined program. This program may be read from the storage medium 151 via the external storage device 150 or may be read from the storage device 130. The storage device 130 stores input programs and data and outputs them. The input / output interface 140 performs input / output of programs and data to / from other computers. The external storage device 150 can read data and programs stored in the storage medium 151, and can write data and programs in the storage medium 151.

  With reference to FIG. 3, an operation performed by each component of the quality monitoring system shown in FIG. 2A, that is, a quality monitoring method according to the first embodiment will be described. FIG. 3 is a flowchart showing an operation example of the quality monitoring method according to the first embodiment.

  The flowchart shown in FIG. 3 includes a total of nine steps from the 0th step SA0 to the 8th step SA8. In the 0th step SA0, the quality monitoring method according to the first embodiment is started. After the 0th step SA0, the first step SA1 is executed.

  In the first step SA1, the manufacturing facility 1 manufactures the product 11. At this time, each product 11 is preferably given a unique serial number. Further, while the manufacturing facility 1 manufactures the product 11, the facility monitor 2 continues to monitor various manufacturing condition parameters of the manufacturing facility 1, and continues to transmit log data obtained as a result to the data recording unit 31. Preferably it is. After the first step SA1, the second step SA2 is executed.

In the second step SA2, the Mahalanobis distance of the manufacturing condition data of the product 11 manufactured in the first step S1 is calculated. Here, the manufacturing condition data of the product 11 is a set of manufacturing condition parameter values when the product 11 is manufactured among log data obtained by the equipment monitor 2 continuously monitoring the manufacturing equipment 1. The equipment monitor 2 preferably transmits the manufacturing condition data to the control unit 32 every time the product 11 is manufactured. The control unit 32 requests the temporary evaluation unit 36 to perform a temporary evaluation of the received manufacturing condition data. At this time, the control unit 32 transmits the manufacturing condition data to the temporary evaluation unit 36. The Mahalanobis distance calculation unit 361 included in the temporary evaluation unit 36 calculates the Mahalanobis distance of the received manufacturing condition data. At this time, Mahalanobis distance calculation section 361, you substitutes manufacturing condition data to the Mahalanobis distance calculation formula itself has previously. The operation of the control unit 32 may be automatically performed or may be performed through an operation by an operator. After the second step SA2, a third step SA3 is executed.

  In the third step SA3, the temporary evaluation unit 36 performs a temporary evaluation of the product 11 by comparing the Mahalanobis distance calculated in the second step S2 with a threshold that the threshold storage unit 362 has in advance. Here, if the Mahalanobis distance is smaller than the threshold value, the provisional evaluation of the product 11 is determined to be acceptable. On the other hand, if the Mahalanobis distance is greater than or equal to the threshold value, the provisional evaluation of the product 11 is determined to be rejected. In addition, what is evaluated at this time is a manufacturing condition parameter of the manufacturing facility 1 when the product 11 is manufactured, and the quality inspection of the product 11 itself has not been performed yet. The result of the temporary evaluation is transmitted from the temporary evaluation unit 36 to the control unit 32. Following the third step SA3, a fourth step SA4 is executed.

  In 4th step SA4, the control part 32 determines whether the result of temporary evaluation performed by 3rd step SA3 is a pass. If the result of the provisional evaluation is acceptable, the production of the product 11 may be continued as it is. In this case, after the fourth step SA4, the process returns to the first step SA1 (branch “YES”). On the contrary, if the result of the temporary evaluation is rejected, it is considered that the production of the product 11 needs to be temporarily stopped and the production condition parameter of the production equipment 1 needs to be changed. In this case, the fifth step SA5 is executed after the fourth step SA4 (branch “NO”).

  In the fifth step SA5, the production facility 1 is stopped to prevent mass production of defective products. At this time, it is preferable that the control unit 32 transmits a control signal indicating a production stop to the production facility 1. Alternatively, the production stop of the production facility 1 may be performed through an operation by an operator. When the production facility 1 is stopped through an operation by the worker, the control unit 32 may display a predetermined screen on the display device 321 to alert the worker. After the fifth step SA5, a sixth step SA6 is executed.

  In 6th step SA6, the control part 32 determines the priority which changes a manufacturing condition parameter. Next, it is necessary to change the values of various manufacturing condition parameters of the manufacturing facility 1 so that the provisional evaluation of the product 11 to be manufactured is passed, but it is not always necessary to change all the manufacturing condition parameters at once. Absent. The control unit 32 analyzes the log data recorded in the data recording unit 31, and considers the contribution rate to the Mahalanobis distance and the delay to the limit value set for each manufacturing condition parameter in combination. It is preferable to determine the order of manufacturing condition parameters to be preferentially changed. Detailed description regarding the priority order will be described later. After the sixth step SA6, a seventh step SA7 is executed.

  In 7th step SA7, the control part 32 changes the manufacturing condition parameter of the manufacturing equipment 1 according to the priority determined in 6th step SA6. At this time, it is preferable that the control unit 32 transmits a control signal indicating the change of the manufacturing condition parameter to the manufacturing facility 1. Alternatively, the manufacturing condition parameter may be changed through an operation by an operator. When changing the manufacturing condition parameters through the operation by the worker, the control unit 32 may display a predetermined screen on the display device 321 to indicate the priority order to the worker. Detailed description regarding this screen display will be described later. After the seventh step SA7, an eighth step SA8 is executed.

  In the eighth step SA8, the production facility 1 resumes the production of the product 11. At this time, it is preferable that the control part 32 transmits the control signal which shows manufacture restart to the manufacturing equipment 1. FIG. Or resumption of manufacture may be performed through operation by an operator. After the eighth step SA8, the process returns to the first step SA1.

  By executing each step of the flowchart shown in FIG. 3, quick feedback to the manufacturing facility 1 is possible based on the temporary evaluation of the product 11.

  Here, the priority order determined in the sixth step S6 will be described in more detail. As described above, the priority order for changing the manufacturing condition parameter is determined based on the contribution rate and the delay to the limit value. Here, the contribution rate is the magnitude of the factor effect that the manufacturing condition parameter has on the Mahalanobis distance. Further, the postponement until the limit value is the degree to which the upper limit value or the lower limit value allowed for the manufacturing condition parameter is approached, and is the degree to which the manufacturing condition parameter is far from the target value.

  Table 5 is a table showing an example of the factor effect that the manufacturing condition parameter has on the Mahalanobis distance. In the example shown in Table 5, there are a total of nine manufacturing condition parameters, from 1 to 9 thereof. For each of these nine manufacturing condition parameters, the factor effect in the case of “ON” used as the coefficient of the Mahalanobis distance calculation formula and the factor effect in the case of “OFF” not used are divided into two types. Think.

  In the example of Table 5, the S / N ratio when “ON” and the S / N ratio when “OFF” are calculated for each of a total of nine manufacturing condition parameters. Here, the S / N ratio is an extension of the concept of signal / noise ratio in communication engineering to quality engineering, and represents the degree of product quality variation in units of dB (decibel).

  FIG. 4 is a graph illustrating the S / N ratio exemplified in Table 5. On the horizontal axis of the graph shown in FIG. 4, each of the nine manufacturing condition parameters shown in Table 5 is arranged adjacent to “ON” and “OFF”. The vertical axis represents the S / N ratio. For each manufacturing condition parameter, the slope of the line connecting the “on” and “off” points corresponds to the amount of displacement shown in Table 5. The average value shown in Table 5 corresponds to the horizontal broken line shown in FIG.

The contribution rate calculation method shown in Table 5 will be described. The contribution ratio of a certain manufacturing condition parameter to the Mahalanobis distance is defined by the following equation.
Contribution rate = displacement / (sum of absolute values of displacement of all manufacturing condition parameters)
Defective product data is used for the calculation of the contribution rate. In the quality monitoring system according to the present embodiment shown in FIG. 2A, the contribution rate calculation unit 38 reads out the defective product data from the defective product data recording unit 34 and calculates the contribution rate.

As an example, the manufacturing condition parameter 2 shown in Table 5 is calculated. The displacement amounts of the manufacturing condition parameters 1 to 9 are 4.00, 2.91, 1.20, 0.70, 0.50, 0.50, -0.10, 0.10 in this order, respectively. 0.005, so
Contribution rate = 2.91 / (| 4.00 | + | 2.91 | + | 1.20 | + | 0.70 | + | 0.50 | + | 0.50 | + | −0.10 | + | 0.10 | + | 0.005 |) = 29.1%
Is obtained.

  It is preferable that the priority order of the manufacturing condition parameter to be changed when the provisional evaluation of the product 11 is rejected is basically the same as the order in which the contribution ratio calculated as described above is large. However, if there are manufacturing condition parameters with less time to the limit value or manufacturing condition parameters that have already exceeded the limit value, it is preferable to preferentially change these manufacturing condition parameters regardless of the contribution rate. It is preferable that the relationship between the contribution rate, the grace until the limit value, and the change priority order is automatically determined using a preset program or the like. Or after presenting the contribution rate and the grace until the limit value, the final judgment regarding the priority of the change may be left to the worker.

  Here, the screen display in the seventh step SA7 will be described in more detail. FIG. 5 is a diagram illustrating an example of a method for displaying manufacturing condition parameters that are candidates for change in the quality monitoring system according to the first embodiment.

  In the example of FIG. 5, a total of three graphs of the first graph 210 to the third graph 230 are displayed on the display device 321. FIG. 5 is merely an example, and the total number of graphs displayed on the display device 321 at a time may be an integer other than 3. Further, the display device 321 may be able to display a further graph by a scroll function.

  The first graph 210 to the third graph 230 show temporal changes of different manufacturing condition parameters. Hereinafter, details of the first graph 210 will be described.

  The horizontal axis of the first graph 210 indicates the passage of time. The vertical axis indicates the value of a certain manufacturing condition parameter. The first graph 210 includes a graph line 211 indicating a change in manufacturing condition parameter over time, an upper limit value line 212 indicating an allowable upper limit value by a broken line, and a target value line 213 indicating a target value by a one-dot chain line. And a lower limit value line 214 that indicates the lower limit value with a broken line. The first graph 210 further includes a contribution rate value.

  Note that the details of the second graph 220 and the third graph 230 are the same as those of the first graph 210, and thus further detailed description thereof is omitted.

  The first graph 210 to the third graph 230 are arranged from the top to the bottom of the display device 321 in descending order of contribution. By arranging in this way, the worker who visually recognizes the display device 321 can easily recognize the priority order based on the contribution rate when changing the manufacturing condition parameter.

  When attention is paid to the rightmost point of the third graph line 231, the upper limit value line 232 is exceeded. In such a case, it is preferable to change the manufacturing condition parameter of the third graph 230 with the highest priority regardless of the contribution rate. To make it easier for workers to recognize this situation, points that have exceeded the limit value or points that have approached the limit value beyond the specified standard are displayed in different colors or flashed. Or the entire graph may be displayed in a different background color from the other graphs.

  Up to this point, feedback based on provisional evaluation by the MT method has been described. From here, the Mahalanobis distance calculation formula based on the actual quality inspection of the product 11 and feedback related to the threshold value update will be described.

(Second Embodiment)
Since the actual quality inspection of the product 11 often takes more time than the temporary evaluation, it is preferable to perform it regardless of the feedback based on the temporary evaluation described as the first embodiment.

  Quality data indicating the result of quality inspection of the product 11 is also preferably recorded in the data recording unit 31. In the present embodiment, the quality data and log data accumulated in this way are combined to create a more accurate Mahalanobis distance calculation formula and threshold value according to the product 11.

  FIG. 6 is a flowchart illustrating an operation example in which the Mahalanobis distance calculation formula is updated by the quality monitoring method according to the second embodiment. FIG. 6 includes a total of seven steps of the 0th step SB0 to the 6th step SB6. The flowchart of FIG. 6 starts from the 0th step SB0. After the 0th step SB0, the 1st step SB1 is executed.

  In the first step SB1, the log data recorded in the data recording unit 31 is sorted into good product data and defective product data. Here, the non-defective product data is manufacturing condition data corresponding to the product 11 that is actually determined to be non-defective by quality inspection in the log data. Similarly, the defective product data is manufacturing condition data corresponding to the product 11 that is actually determined to be defective by quality inspection in the log data. The sorted good product data is recorded in the good product data recording unit 33. Similarly, the sorted defective product data is recorded in the defective product data recording unit 34. The operation of distributing the manufacturing condition data for each product 11 included in the log data as non-defective product data or defective product data may be performed by the control unit 32 or may be performed by another arithmetic device (not shown). After the first step SB1, the second step SB2 is executed.

  In the second step SB2, the Mahalanobis distance is calculated for each of the defective product data recorded in the defective product data recording unit 34. Here, for example, the contribution rate calculation unit 38 reads out the defective product data from the defective product data recording unit 34 and transmits the read out defective product data to request the Mahalanobis distance calculation unit 361 to calculate the Mahalanobis distance, and the Mahalanobis distance. The calculation unit 361 may calculate the Mahalanobis distance of the defective product data and transmit the calculated Mahalanobis distance to the contribution rate calculation unit 38. Following the second step SB2, a third step SB3 is executed.

  In the third step SB3, the contribution rate calculation unit 38 calculates the contribution rate of each manufacturing condition parameter to the Mahalanobis distance based on the Mahalanobis distance calculated in the second step SB2. Following the third step SB3, a fourth step SB4 is executed.

  In the fourth step SB4, the manufacturing condition parameters used in the new Mahalanobis distance calculation formula are selected based on the contribution ratio calculated in the third step SB3. Of the manufacturing condition parameters, what is selected here is basically one having a high contribution rate. After the fourth step SB4, a fifth step SB5 is executed.

  In the fifth step SB5, the Mahalanobis distance calculation formula creating unit 37 creates a new Mahalanobis distance formula using the manufacturing condition parameters selected in the fourth step SB4. The Mahalanobis distance calculation formula creation unit 37 may create a Mahalanobis distance calculation formula by executing a Mahalanobis distance calculation formula creation program. The created Mahalanobis distance calculation formula is preferably transmitted to and stored in the Mahalanobis distance calculation unit 361. After the fifth step SB5, a sixth step SB6 is executed.

  In the sixth step SB6, the flowchart ends.

  In the quality monitoring system and the quality monitoring method according to the present embodiment, the threshold value is also changed in order to improve the accuracy of the temporary evaluation as compared with the first embodiment. In the general MT method, the threshold value to be compared with the Mahalanobis distance is set to an integer “4”. As described above, in the case of the quality monitoring system and the quality monitoring method according to the first embodiment, if the Mahalanobis distance of the manufacturing condition data at the time of manufacturing the product 11 is smaller than the integer “4” that is the threshold value, the pass “integer“ If it is 4 ”or more, it is provisionally evaluated as rejected.

  However, in practice, the product 11 that is determined to be acceptable in the provisional evaluation by the MT method may be determined as a defective product in the actual quality inspection. On the contrary, the product 11 that is determined to be unacceptable in the provisional evaluation by the MT method may be determined as a non-defective product in the actual quality inspection. Such a determination error in the provisional evaluation by the MT method can be defined by two scales called determination accuracy precision and recall.

  In the tentative evaluation by the MT method, the accuracy rate of determination accuracy is defined as the ratio of products 11 that are included in the products 11 that have been determined to be acceptable in the tentative evaluation and that are determined to be non-defective products in the actual quality inspection. Similarly, the reproduction rate of the determination accuracy is defined as a ratio of the products 11 that are included in the products 11 that are determined to be non-defective products in the actual quality inspection and that are determined to be acceptable in the provisional evaluation.

  Table 6 is a table showing an example of the relationship between the determination result of the temporary evaluation by the MT method and the determination result by the actual quality inspection.

In the example shown in Table 6, out of a total of 121 products 11 determined to be acceptable in the provisional evaluation by the MT method, 116 products 11 are determined to be non-defective products in the actual quality inspection. The precision in such cases is
116/121 = about 95.9%
Is calculated.

Similarly, out of a total of 118 products 11 that are determined to be non-defective products in the actual quality inspection, 116 products 11 are determined to be acceptable in the provisional evaluation by the MT method. The recall rate in such cases is
116/118 = about 98.3%
Is calculated.

  Ideally, the precision and recall are both 100%. However, as long as there is a possibility that the result of the provisional evaluation by the MT method does not necessarily match the result of the actual quality inspection, it is required to use a threshold value such that the precision rate and the recall rate are as close to 100% as possible.

In the present embodiment, in order to take into consideration both the precision “X” and the recall “Y”, the harmonic average of both is used. This harmonic average value is hereinafter referred to as “F value”. The F value is defined by the following equation.
F value = (2 × X × Y) / (X + Y)
As an example, if the numerical values illustrated in Table 6 are substituted into the above formula,
F value = (2 × 95.9% × 98.3%) / (95.9% + 98.3%) = about 0.97
Is obtained.

  From the above definition, it is considered that the maximum value of the F value is 100%, and that the larger the F value, the closer the relevance ratio and the recall ratio are to 100%. Therefore, in the present embodiment, a threshold value that maximizes the F value is searched, and the found threshold value is adopted as a comparison target with the Mahalanobis distance.

  FIG. 7 is a diagram for explaining the principle of calculating a threshold in the quality monitoring system according to the second embodiment. The horizontal axis shown in FIG. 7 indicates the Mahalanobis distance of the product 11. A plurality of “◯” marks shown on the horizontal axis of FIG. 7 indicate the distribution of Mahalanobis distances of manufactured products 11 that are determined to be non-defective products by quality inspection. Similarly, a plurality of “x” marks shown below the horizontal axis of FIG. 7 indicate the distribution of Mahalanobis distances of manufactured products 11 that are determined to be defective by quality inspection. .

  “D_max” shown in FIG. 7 indicates the maximum value of the Mahalanobis distance of a good product. Similarly, “d_min” illustrated in FIG. 7 indicates the minimum value of the Mahalanobis distance of a defective product. In the example of FIG. 7, since d_min is smaller than d_max, the product 11 included in the range where the Mahalanobis distance is not less than d_min and not more than d_max in the preliminary evaluation includes both non-defective products and defective products. This means that the result of provisional evaluation does not match the result of quality inspection. In the present embodiment, a threshold that is considered optimal in a range of d_min to d_max is obtained based on the criterion that the F value is maximized.

  FIG. 8 is a flowchart showing an operation example of a method for calculating an optimum threshold value in the quality monitoring system according to the present embodiment. The flowchart shown in FIG. 8 includes a total of nine steps of the 0th step SC0 to the 8th step SC8.

  The flowchart of FIG. 8 starts from the 0th step SC0. After the 0th step SC0, the first step SC1 is executed.

  In the first step SC1, a total of three constants are set. The first constant “Mesh” is the reciprocal of the mesh division number. The number of mesh divisions is the number of divisions for equally dividing a range from d_min to d_max in a mesh shape. As an example, if the number of mesh divisions is 10, the constant Mesh is 0.1. Since the second constant “d_min” and the third constant “d_max” are as defined above, further detailed description is omitted. Following the first step SC1, the second step SC2 is executed.

  In the second step SC2, a total of three variables are initialized. A threshold value candidate is stored in the first variable “criteria”. In the example of FIG. 8, the initial value of the variable criterion is “0”. The second variable “F” stores a candidate for the maximum value of the F value. In the example of FIG. 8, the initial value of the variable F is “0”. The third variable “T” stores an optimal threshold candidate. In the example of FIG. 8, the initial value of the variable T is a constant d_min. The second step SC2 may be executed before the first step SC1. After the second step SC2, the third step SC3 is executed.

In the third step SC3, the F value when the current criteria is used is calculated, and the result is stored in the variable “F_temp”. Here, first, a threshold value corresponding to the variable criterion is calculated by the following equation.
Threshold = d_min + criteria
Next, using this threshold as a reference, the Mahalanobis distance of the product 11 manufactured in the past is temporarily evaluated. The relationship between the result of the temporary evaluation obtained in this way and the result of the past quality inspection stored in the data recording unit 31, the non-defective product data recording unit 33, or the defective product data recording unit 34, for example, In the same format as 6, the matching rate and recall rate at the current threshold are calculated. Next, the F value is obtained by calculating the harmonic average of the precision and recall. The F value thus obtained is stored in the variable F_temp as described above. Following the third step SC3, a fourth step SC4 is executed.

  In the fourth step SC4, the values of the variable F_temp and the variable F are compared to determine whether the former is equal to or less than the latter. If F_temp ≦ F is false (NO), the fifth step SC5 is executed next, and then the sixth step SC6 is executed. On the other hand, if F_temp ≦ F is true (YES), the fifth step SC5 is not executed next. The sixth step SC6 is executed.

  In the fifth step SC5, the variable F_temp is stored in the variable F, and the current threshold value is stored in the variable T. Following the fifth step SC5, a sixth step SC6 is executed.

  In the sixth step SC6, a constant Mesh is added to the variable criteria. After the sixth step SC6, a seventh step SC7 is executed.

  In the seventh step SC7, it is determined whether or not the variable criteria exceeds the constant mesh. If it exceeds (YES), the eighth step SC8 is executed next, and if it does not exceed (NO), the next is executed. The third step SC3 to the seventh step SC7 are repeated once more.

  In the eighth step SC8, the flowchart ends. At this time, the value stored in the variable F is the maximum F value, and the value stored in the variable T is the optimum threshold value.

  In the flowchart of FIG. 8, the larger the number of divided meshes, the higher the accuracy of the calculated threshold value. On the other hand, since the calculation amount increases in proportion to the number of divided meshes, it is preferable to set the number of divided meshes to an appropriate value.

  If d_max is smaller than d_min, the result of the temporary evaluation matches the result of the quality inspection, so any value from the range of d_min to d_max may be selected as the threshold value. As an example, the average value of d_min and d_max may be selected as the threshold value, or the threshold value that maximizes the F value may be selected by the same algorithm as in the flowchart of FIG.

(Weighting F value)
The following change can be further added to the appropriate threshold value calculation method according to the present embodiment. That is, when calculating the F value, the precision and recall are weighted. The F value defined in this way is hereinafter referred to as “weighted F value”.

The weighting F value is a harmonic average value of the precision “X” weighted by the first coefficient “A” and the recall “Y” weighted by the second coefficient “B”. It is calculated by the formula.
Weighted F value = (A + B) / ((A / X) + (B / Y))
If the sum of A and B is 1, the weighted F value may be defined by the following equation.
Weighted F value = 1 / ((A / X) + ((1-A) / Y))

  Here, if both the first coefficient “A” and the second coefficient “B” are 0.5, the weighted F value is a harmonic average value of the precision “X” and the recall “Y”, that is, normal. Is equal to the F value.

  Also, if the first coefficient “A” is equal to 1 and the second coefficient “B” is equal to 0, the weighted F value is equal to the precision “X”. Conversely, if the first coefficient “A” is equal to 0 and the second coefficient “B” is equal to 1, then the weighted F value is equal to the recall “Y”. Therefore, when there is a high risk that a defective product is mistakenly regarded as a non-defective product, that is, when it is desired to reliably consider only a non-defective product as a non-defective product (or an acceptable product), the first coefficient “A” is adjusted to a small value, Adjust “B” largely. On the other hand, if there are high costs such as confirmation or misuse, which is caused by mistakenly judging a non-defective product as a defective product, that is, only a defective product is definitely regarded as a defective product (or rejected product). If it is desired to consider, the first coefficient “A” is adjusted to be large, and the second coefficient “B” is adjusted to be small. Adjusting the first coefficient “A” and the second coefficient “B” has such implications.

  The inventor confirms that the yield of the product 11 may be improved as a result of using the weighted F value as described above, rather than the F value as described above, depending on the situation of the manufacturing facility 1. did. Here, the situation of the manufacturing facility 1 can be defined as a combination of manufacturing condition parameters of the manufacturing facility 1. As a specific method of utilizing the weighted F value, for example, a plurality of optimum combinations of the first coefficient A and the second coefficient B are prepared for each situation of the manufacturing facility 1 and correspond to the situation of the manufacturing facility 1. It is conceivable to read out the optimal combination to be changed, change the calculation method of the weighted F value, and calculate the optimal threshold value corresponding to the calculated weighted F value. At this time, it is preferable that a plurality of combinations of the first coefficient A and the second coefficient B are stored in advance in the threshold update unit 35 or the like so as to be readable.

  The invention made by the inventor has been specifically described based on the embodiment. However, the present invention is not limited to the embodiment, and various modifications can be made without departing from the scope of the invention. Needless to say. In addition, the features described in the embodiments can be freely combined within a technically consistent range.

DESCRIPTION OF SYMBOLS 1 Manufacturing equipment 11 Product 2 Equipment monitor 3 Quality monitoring server 31 Data recording part 32 Control part 321 Display apparatus 33 Good product data recording part 34 Defective product data recording part 35 Threshold update part 36 Temporary evaluation part 361 Mahalanobis distance calculation part 362 Threshold memory part 37 Mahalanobis distance calculation formula creation unit 371 Mahalanobis distance calculation formula creation program 38 Contribution rate calculation unit 4 Quality inspection unit 51 d_min
52 d_max
53 mesh division 100 computer 110 bus 120 arithmetic device 130 storage device 140 input / output interface 150 external storage device 151 storage medium 210, 220, 230 graph 211, 221, 231 graph line 212, 222, 232 upper limit value 213, 223, 233 target Value 214, 224, 234 Lower limit

Claims (10)

A data recording unit for recording log data of manufacturing condition parameter values when manufacturing the product;
A Mahalanobis distance calculator that calculates a Mahalanobis distance of manufacturing condition data that is a set of values of the manufacturing condition parameters when the specific product is manufactured included in the log data;
Comparing the Mahalanobis distance with a threshold value to evaluate the state of the manufacturing facility, and a provisional evaluation unit for provisional evaluation of the quality of the specific product;
When the result of the temporary evaluation is unacceptable, comprising a control unit that stops and controls the manufacturing equipment so as to change the value of one or more types of the manufacturing condition parameters,
The manufacturing condition parameter to be changed is determined based on a contribution rate to the Mahalanobis distance and a delay to a limit value set for each manufacturing condition parameter ,
The data recording unit further records quality data indicating a result of determining the quality of the manufactured product as a good product or a defective product,
A threshold update unit that updates the threshold based on the log data, the quality data, and the Mahalanobis distance
Further comprising
The threshold update unit
When the minimum value of the Mahalanobis distance corresponding to the defective product is smaller than the maximum value of the Mahalanobis distance corresponding to the non-defective product, the threshold to be updated is selected from a range between the minimum value and the maximum value. And
A quality monitoring system that selects the threshold to be updated from a range between the maximum value and the minimum value when the minimum value is equal to or greater than the maximum value . The quality monitoring system according to claim 1 ,
The threshold update unit updates the threshold based further on a weighted F value,
The weighting F value is a harmonic average value of the relevance rate weighted with the first coefficient and the recall rate weighted with the second coefficient,
The relevance rate is a ratio at which the product determined to be acceptable in the provisional evaluation is determined to be non-defective in the quality inspection,
The reproducibility is a ratio at which a product determined to be non-defective in the quality inspection is determined to be acceptable in the provisional evaluation. Quality monitoring system. The quality monitoring system according to claim 2 ,
A coefficient storage unit for storing a plurality of combinations of the first coefficient and the second coefficient;
A quality monitoring system further comprising: a coefficient selection unit that selects an optimum combination from the plurality of combinations according to the Mahalanobis distance. In the quality monitoring system as described in any one of Claims 1-3 ,
For each of a plurality of manufacturing condition parameters, a contribution rate calculation unit that calculates a contribution rate to the Mahalanobis distance;
A quality monitoring system, further comprising: a Mahalanobis distance calculation formula creation unit that creates a new calculation formula of the Mahalanobis distance using the selected part of the manufacturing condition parameters. In the quality monitoring system according to claim 4 ,
A quality monitoring system, further comprising: a display device that highlights a graph showing a change over time of the value of the manufacturing condition parameter according to the contribution rate and a delay to a limit value of the value of the manufacturing condition parameter. Recording log data of the values of manufacturing condition parameters when manufacturing the product;
Calculating a Mahalanobis distance of manufacturing condition data, which is a set of values of the manufacturing condition parameters when the specific product is manufactured included in the log data;
Comparing the Mahalanobis distance with a threshold value to evaluate the state of the manufacturing facility, and making a provisional evaluation of the quality of the specific product;
When the result of the temporary evaluation is unacceptable, comprising controlling the production equipment to stop so as to change the value of one or more kinds of the production condition parameters,
Said controlling is
Determining the manufacturing condition parameter to be changed based on a contribution rate to the Mahalanobis distance and a grace period set for each manufacturing condition parameter ;
Further recording quality data indicating the result of quality inspection of the manufactured product and determining that the product is non-defective or defective;
Updating the threshold based on the log data, the quality data and the Mahalanobis distance;
Further comprising
Updating the threshold includes
When the minimum value of the Mahalanobis distance corresponding to the defective product is smaller than the maximum value of the Mahalanobis distance corresponding to the non-defective product, the threshold to be updated is selected from a range between the minimum value and the maximum value. To do
If the minimum value is greater than or equal to the maximum value, selecting the threshold to be updated from a range greater than or equal to the maximum value and less than or equal to the minimum value;
A quality monitoring method comprising : The quality monitoring method according to claim 6 ,
Updating the threshold includes
Updating the threshold further based on a weighted F value,
The weighting F value is a harmonic average value of the relevance rate weighted with the first coefficient and the recall rate weighted with the second coefficient,
The relevance rate is a ratio at which the product determined to be acceptable in the provisional evaluation is determined to be non-defective in the quality inspection,
The reproducibility is a ratio at which the product determined to be non-defective in the quality inspection is determined to be acceptable in the provisional evaluation. Quality monitoring method. The quality monitoring method according to claim 7 ,
Preparing a plurality of combinations of the first coefficient and the second coefficient;
Selecting an optimal combination from the plurality of combinations according to the Mahalanobis distance; and a quality monitoring method. In the quality monitoring method as described in any one of Claims 6-8 ,
For each of a plurality of manufacturing condition parameters, calculating a contribution rate to the Mahalanobis distance;
Creating a new calculation formula of the Mahalanobis distance using the selected part of the manufacturing condition parameters. Quality monitoring method. The quality monitoring method according to claim 9 ,
A quality monitoring method, further comprising: highlighting a graph showing a time change of the value of the manufacturing condition parameter according to the contribution rate and a delay to a limit value of the value of the manufacturing condition parameter.

Images