JP2011100211A - Failure determining device, failure determining method, failure determining program, and program recording medium recording the program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a failure determining device that detects a failure even in a device in which a lot of periodic changes occur in the temporal change and tendency thereof cannot be easily quantified and formulated. <P>SOLUTION: The failure determining device 10 includes a collecting section 11, a temporary storage section 12, a classified data storage section 13, a classifying section 14 and a failure determining section 15. In a first phase, a category to which the manufacturing data to be determined belongs is determined. In a second phase, whether the manufacturing data has deviated with the change of time in the category to which the manufacturing data belongs is monitored to determine a failure or not. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an abnormality determination device, an abnormality determination method, an abnormality determination program, and a program recording medium on which the program is recorded.

  2. Description of the Related Art In recent years, in plant operations such as semiconductor manufacturing, in order to detect and predict abnormalities in manufacturing equipment and products at an early stage, devices that monitor manufacturing conditions, manufacturing results, and the like have been developed. The monitoring apparatus monitors the process data obtained from the manufacturing apparatus and the time transition of the manufacturing data, thereby grasping the state of the manufacturing apparatus and detecting an apparatus abnormality.

  When the manufacturing data to be monitored changes with time in the manufacturing process, there is a method for determining manufacturing data from the tendency of change with time as a method for accurately predicting and detecting an abnormality in the manufacturing apparatus (Japanese Patent Laid-Open No. 2000-146636). Gazette: see Patent Document 1).

  This first prior art compares the tendency of time transition at the time of abnormality with the tendency of time transition of the monitoring target, and determines that it is abnormal when a change pattern similar to the abnormal pattern is shown.

  However, in the first prior art, when it is difficult to formulate or model a temporal change pattern at the time of abnormality, the accuracy of abnormality detection is lowered. This becomes a problem when the frequency of device abnormality is extremely low, such as when the factory is operating stably, and sampling data for modeling abnormal conditions is not available.

  On the other hand, as a second conventional technique, there is a method of detecting an abnormality by comparing a trend of time transition of manufacturing data in a normal state and a trend of time transition of manufacturing data to be monitored (Japanese Patent Laid-Open No. 10-301619). : Patent Document 2).

  In the second prior art, the regression coefficient of the regression equation is calculated from the threshold value created based on the trend of the time transition of the manufacturing data judged to be normal and the time transition of the monitored data, and compared with the regression coefficient. is doing. The second prior art overcomes the problems of the first prior art.

  In particular, when the monitored data to be handled is a control parameter or process data in which a normal temporal change is uniquely determined, the second prior art can be said to be effective.

  However, in the second prior art, when an inspection device or a manufacturing device whose change over time is not uniquely determined regardless of whether it is normal or abnormal, it is difficult to quantify and formulate change over time. Unable to handle abnormality detection.

JP 2000-146636 A JP-A-10-301619

  Accordingly, an object of the present invention is to provide an abnormality determination device and an abnormality determination method for detecting an abnormality even for a device in which a change with time periodically changes, a large number of changes appear, and it is difficult to quantify and formulate the tendency. An object of the present invention is to provide an abnormality determination program and a program recording medium on which the program is recorded.

In order to solve the above problems, the abnormality determination device of the present invention is:
A collection unit that collects manufacturing data from the manufacturing device in time series;
A temporary storage unit for sequentially storing the manufacturing data collected at the collection unit at each time in a predetermined manufacturing period.
A classified data storage unit that stores the manufacturing data collected by the collection unit in the past and a plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. When,
The manufacturing data of the current manufacturing period stored in the temporary storage unit is converted into a set of manufacturing data stored in the classified data storage unit having a time transition similar to the time transition of the manufacturing data of the current manufacturing period. A classification part for classification;
Compare the time transition of the manufacturing data of the current manufacturing period with the time transition of the manufacturing data stored in the classified data storage unit in the set of manufacturing data in which the manufacturing data of the current manufacturing period is classified, An abnormality determination unit that detects abnormality by obtaining similarity is provided.

  According to the abnormality determination device of the present invention, since the collection unit, the temporary storage unit, the classified data storage unit, the classification unit, and the abnormality determination unit are included, the manufacturing data to be determined in the first phase In the second phase, it is determined whether or not the manufacturing data is abnormal by monitoring whether the manufacturing data deviates with time transition in the classification group to which it belongs. As described above, by classifying the manufacturing data according to the characteristics of the change in the time transition, it is possible to accurately detect the abnormality even when the change of the manufacturing data with time is not uniquely determined.

Moreover, in the abnormality determination device of one embodiment,
The classification unit and the abnormality determination unit are respectively
When calculating the similarity of the time transition between the manufacturing data,
The manufacturing data determined to be normal belonging to the same classification as the manufacturing data of the current manufacturing period is acquired from the classified data storage unit,
The manufacturing data of the current manufacturing period and the manufacturing data determined to be normal are divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity to the change trend between the time transition of the manufacturing data of the current manufacturing period and the time transition of the manufacturing data determined to be normal,
The minimum similarity is obtained from the calculated similarities.

  According to the abnormality determination device of this embodiment, when making an abnormality determination for manufacturing data that changes with time, by comparing with normal manufacturing data belonging to the same classification recorded in the classified data storage unit, The manufacturing data that has been removed is determined to be abnormal. Thereby, even if the change of the manufacturing data in the abnormal state is not uniquely determined, it can be detected in a systematic manner. In addition, when manufacturing data are compared over the entire period, the accuracy decreases. Therefore, it is possible to improve the accuracy of abnormality determination by dividing the time intervals based on the characteristics of time transitions and individually comparing sections having the same time transition tendency.

Moreover, in the abnormality determination device of one embodiment,
The abnormality determination unit
When calculating the similarity of the time transition between the manufacturing data,
Manufacturing data determined to be abnormal belonging to each classification is obtained from the classified data storage unit,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to the change trend between the time transition of the manufacturing data of the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
The maximum similarity is obtained from the calculated similarities.

  According to the abnormality determination device of this embodiment, the abnormality determination unit compares the abnormal manufacturing data recorded in the classified data storage unit with the current manufacturing data to be determined. As a result, even when the temporal change of the manufacturing data in the normal state is not uniquely determined, it is possible to detect the abnormality with high accuracy by determining that the manufacturing data having the same temporal change as the known abnormal state is abnormal.

Moreover, in the abnormality determination device of one embodiment,
The abnormality determination unit
When calculating the similarity of the time transition between the manufacturing data,
The manufacturing data determined to be normal belonging to the same classification as the manufacturing data of the current manufacturing period is acquired from the classified data storage unit,
The manufacturing data of the current manufacturing period and the manufacturing data determined to be normal are divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity to the change trend between the time transition of the manufacturing data of the current manufacturing period and the time transition of the manufacturing data determined to be normal,
While obtaining the minimum similarity from the calculated similarities,
Manufacturing data determined to be abnormal belonging to each classification is obtained from the classified data storage unit,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
The maximum similarity is obtained from the calculated similarities.

  According to the abnormality determination apparatus of this embodiment, the abnormality determination unit performs an abnormality determination based on both the past data in the abnormal state and the past data in the normal state, thereby performing a double check, and the accuracy of the abnormality detection is increased. Can be improved.

Moreover, the abnormality determination device of the present invention is
A collection unit that collects manufacturing data from the manufacturing device in time series;
A temporary storage unit for sequentially storing the manufacturing data collected at the collection unit at each time in a predetermined manufacturing period.
A classified data storage unit that stores the manufacturing data collected by the collection unit in the past and a plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. When,
By comparing the time transition of the manufacturing data of the current manufacturing period stored in the temporary storage unit with the time transition of the manufacturing data of each classification stored in the classified data storage unit, and obtaining similarity With an abnormality determination unit that detects an abnormality,
The abnormality determination unit
When calculating the similarity of the time transition between the manufacturing data,
Manufacturing data determined to be abnormal belonging to each classification is obtained from the classified data storage unit,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
It is characterized in that the maximum similarity is obtained from the calculated similarities.

  According to the abnormality determination device of the present invention, since the collection unit, the temporary storage unit, the classified data storage unit, and the abnormality determination unit are included, the abnormality determination unit is recorded in the classified data storage unit. The abnormal manufacturing data is compared with the current manufacturing data to be determined. As a result, even when the temporal change of the manufacturing data in the normal state is not uniquely determined, it is possible to detect the abnormality with high accuracy by determining that the manufacturing data having the same temporal change as the known abnormal state is abnormal.

The abnormality determination method of the present invention is
A collection step for collecting manufacturing data from the manufacturing equipment in time series;
Temporary storage step for sequentially storing the manufacturing data collected at the collecting step at each time in a predetermined manufacturing period;
The classified data storage that is the manufacturing data collected in the collecting step and stores the manufacturing data of the plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. Steps,
The manufacturing data of the current manufacturing period stored in the temporary storage step is converted into a set of manufacturing data stored in the classified data storage step having a time transition similar to the time transition of the manufacturing data of the current manufacturing period. A classification step to classify;
Compare the time transition of the manufacturing data of the current manufacturing period with the time transition of the manufacturing data stored in the classified data storage step in the set of manufacturing data in which the manufacturing data of the current manufacturing period is classified, An abnormality determination step for detecting an abnormality by obtaining similarity is provided.

  According to the abnormality determination method of the present invention, since it includes the collection step, the temporary storage step, the classified data storage step, the classification step, and the abnormality determination step, the manufacturing data to be determined in the first phase In the second phase, it is determined whether or not the manufacturing data is abnormal by monitoring whether the manufacturing data deviates with time transition in the classification group to which it belongs. As described above, by classifying the manufacturing data according to the characteristics of the change in the time transition, it is possible to accurately detect the abnormality even when the change of the manufacturing data with time is not uniquely determined.

The abnormality determination method of the present invention is
A collection step for collecting manufacturing data from the manufacturing equipment in time series;
Temporary storage step for sequentially storing the manufacturing data collected at the collecting step at each time in a predetermined manufacturing period;
The classified data storage that is the manufacturing data collected in the collecting step and stores the manufacturing data of the plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. Steps,
By comparing the time transition of the manufacturing data of the current manufacturing period stored in the temporary storage step with the time transition of the manufacturing data of each classification stored in the classified data storage step, and obtaining similarity With an abnormality determination step for detecting an abnormality,
In the abnormality determination step,
When calculating the similarity of the time transition between the manufacturing data,
Manufacturing data determined to be abnormal belonging to each classification is acquired from the above classified data storage step,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
It is characterized in that the maximum similarity is obtained from the calculated similarities.

  According to the abnormality determination method of the present invention, the abnormality determination step is recorded in the classified data storage step because the collection step, the temporary storage step, the classified data storage step, and the abnormality determination step are included. The abnormal manufacturing data is compared with the current manufacturing data to be determined. As a result, even when the temporal change of the manufacturing data in the normal state is not uniquely determined, it is possible to detect the abnormality with high accuracy by determining that the manufacturing data having the same temporal change as the known abnormal state is abnormal.

  Further, the abnormality determination program of the present invention may be configured such that the computer is used as the collection unit, the temporary storage unit, the classified data storage unit, the classification unit, and the abnormality determination unit, or the collection unit, the temporary storage unit, It is characterized by functioning as the classified data storage unit and the abnormality determination unit.

  According to the abnormality determination program of the present invention, the computer is used as the collection unit, the temporary storage unit, the classified data storage unit, the classification unit, and the abnormality determination unit, or the collection unit, the temporary storage unit, Since it functions as the classified data storage unit and the abnormality determination unit, the abnormality determination device can be realized by installing the program in a computer.

  A program recording medium according to the present invention is characterized in that the abnormality determination program is recorded.

  According to the program recording medium of the present invention, since the abnormality determination program is recorded, the abnormality determination program can be easily installed in a computer.

  According to the abnormality determination device of the present invention, since the collection unit, the temporary storage unit, the classified data storage unit, the classification unit, and the abnormality determination unit are included, the manufacturing data is classified by characteristics of changes in time transition. By doing so, it is possible to detect an abnormality with high accuracy even when the change in manufacturing data over time is not uniquely determined.

  According to the abnormality determination device of the present invention, since the collection unit, the temporary storage unit, the classified data storage unit, and the abnormality determination unit are included, even when the temporal change of the manufacturing data in a normal state is not uniquely determined. It is possible to detect an abnormality with high accuracy by determining that manufacturing data having the same temporal change as a known abnormal state is abnormal.

  According to the abnormality determination method of the present invention, since the collection step, the temporary storage step, the classified data storage step, the classification step, and the abnormality determination step are included, the manufacturing data is classified by characteristics of changes in time transition. By doing so, it is possible to detect an abnormality with high accuracy even when the change in manufacturing data over time is not uniquely determined.

  According to the abnormality determination method of the present invention, since the collection step, the temporary storage step, the classified data storage step, and the abnormality determination step are included, even when a change with time in manufacturing data in a normal state is not uniquely determined. It is possible to detect an abnormality with high accuracy by determining that manufacturing data having the same temporal change as a known abnormal state is abnormal.

  According to the abnormality determination program of the present invention, the computer is used as the collection unit, the temporary storage unit, the classified data storage unit, the classification unit, and the abnormality determination unit, or the collection unit, the temporary storage unit, Since it functions as the classified data storage unit and the abnormality determination unit, the abnormality determination device can be realized by installing the program in a computer.

  According to the program recording medium of the present invention, since the abnormality determination program is recorded, the abnormality determination program can be easily installed in a computer.

It is a block diagram which shows 1st Embodiment of the abnormality determination apparatus of this invention. It is a flowchart for demonstrating the procedure of the abnormality diagnosis operation | movement of an abnormality determination apparatus. It is a flowchart for demonstrating the classification method by a classification | category part. It is a flowchart for demonstrating the abnormality determination method by an abnormality determination part. It is a flowchart for demonstrating the calculation method of the similarity degree by a classification | category part and an abnormality determination part. It is a block diagram which shows 2nd Embodiment of the abnormality determination apparatus of this invention. It is a flowchart for demonstrating the abnormality determination method by an abnormality determination part. It is a flowchart for demonstrating the abnormality determination method by the abnormality determination part which determines abnormality based on the past abnormal manufacture data. It is a graph for demonstrating accumulation | storage of the manufacturing data by a temporary memory part. It is a graph for demonstrating the classification method by a classification | category part. It is a graph for demonstrating the abnormality determination method by an abnormality determination part. It is a graph for demonstrating the abnormality determination method by an abnormality determination part. It is a graph for demonstrating the normal data storage method by the classified data storage part. It is a graph for demonstrating the similarity calculation method currently performed by the classification | category part and the abnormality determination part. It is a graph for demonstrating the similarity calculation method performed in the abnormality determination part.

  Hereinafter, the present invention will be described in detail with reference to the illustrated embodiments.

(First embodiment)
FIG. 1 is a block diagram showing a first embodiment of the abnormality determination device according to the present invention. As shown in FIG. 1, the abnormality determination device 10 detects an abnormality in the manufacturing apparatus 20. The abnormality determination device 10 includes a collection unit 11, a temporary storage unit 12, a classified data storage unit 13, a classification unit 14, and an abnormality determination unit 15.

  The manufacturing apparatus 20 is, for example, a semiconductor manufacturing apparatus. The manufacturing data of the manufacturing apparatus 20 is, for example, process data indicating the operating state of the manufacturing apparatus 20. The collection unit 11 collects manufacturing data from the manufacturing apparatus 20 in time series.

  The temporary storage unit 12 sequentially stores the manufacturing data collected by the collecting unit 11 and the manufacturing data at each time in a predetermined manufacturing period. The manufacturing period is, for example, a maintenance cycle of the manufacturing apparatus 20. After the abnormality determination by the abnormality determination device 10, the accumulated manufacturing data is erased to prepare for the next manufacturing data collection.

  The classified data storage unit 13 stores the manufacturing data collected by the collecting unit 11 and the data of a plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. To do.

  The classification unit 14 stores the manufacturing data of the current manufacturing period stored in the temporary storage unit 12 in the classified data storage unit 13 having a time transition similar to the time transition of the manufacturing data of the current manufacturing period. Classify into production data sets.

  The abnormality determination unit 15 includes a time transition of manufacturing data in the current manufacturing period, and a time transition of manufacturing data stored in the classified data storage unit 13 in a set of manufacturing data in which the manufacturing data in the current manufacturing period is classified. Are detected, and abnormality is detected by obtaining similarity.

  Next, an outline of processing performed by the abnormality determination device will be described with reference to the flowchart of FIG. First, the collecting unit 11 starts collecting data related to manufacturing from the manufacturing apparatus 20 (step S21).

  Thereafter, the temporary storage unit 12 continues to collect the manufacturing data D201 acquired by the collection unit 11 during the period t1 (step S22). In this embodiment, as shown in the graph of FIG. 9, D201 indicated by a solid line is recorded until the period t1.

  And the said classification | category part 14 classifies the collected manufacturing data D201 based on the classification item of manufacturing data (step S23). This classification item is C201, and the classified manufacturing data is D201 '. This classification item is classified according to the tendency of manufacturing data over time.

  Here, the outline of the classification method will be described with reference to the graph of FIG. The manufacturing data D201 for the current manufacturing period and the normal manufacturing data D202 and D203 recorded in the past are respectively compared in the period up to the period t1. In this embodiment, D201 is classified into C201 of the same classification item as D202 because the trends in time transition of D201 and D202 are similar.

  Thereafter, the temporary storage unit 12 continues to accumulate the manufacturing data again during the period t2 (step S24).

  Then, the abnormality determination unit 15 compares the time trend of the manufacturing data D201 ′ with the time trend of the manufacturing data D202 belonging to the classification item C201 among the manufacturing data collected in the past. It is determined whether it is normal (step S25).

  After that, when it is determined that it is normal (step S26), and when the period t3 has not passed since the start of manufacturing data accumulation (step S28) (see FIG. 11), the process returns to step 25.

  On the other hand, when it is determined that there is an abnormality (step S26) (see FIG. 12), an abnormality is output and the process ends (step S27).

  On the other hand, if it is determined to be normal (step S26) and the period t3 has elapsed from the start of manufacturing data accumulation (step S28) (see FIG. 13), normal is output and the process ends.

  As described above, the abnormality determination device 10 having the above configuration determines which classification the manufacturing data to be determined belongs to in the first phase (step S23), and the manufacturing data belongs in the second phase. It is determined whether or not there is an abnormality by monitoring whether or not the classification group deviates with time transition (step S25). As described above, by classifying the manufacturing data according to the characteristics of the change in the time transition, it is possible to accurately detect the abnormality even when the manufacturing data changes with time.

  Next, a manufacturing data classification method corresponding to step S23 in the flowchart of FIG. 2 will be described.

  The classification unit 14 acquires a predetermined number of manufacturing data (D202, D203 in FIG. 10) for each classification item from the classified data storage unit 13, and the time of the manufacturing data (D201 in FIG. 10) for the current manufacturing period. The transition and the time transition of each of the acquired manufacturing data (D202 and D203 in FIG. 10) are compared, the similarity of the change tendency is calculated, and the current manufacturing period stored in the temporary storage unit 12 is further calculated. Manufacturing data (D201 in FIG. 10) is classified into the same set as the acquired manufacturing data (D202 in FIG. 10) for which the maximum similarity is calculated.

  Furthermore, the manufacturing data classification method by the classification unit 14 will be described more specifically using the flowchart of FIG. 3 and the graph of FIG. First, the classification unit 14 acquires the manufacturing data D201 from the temporary storage unit 12 (step S31).

  Thereafter, the classification unit 14 selects one classification item from the classification items stored in the classified data storage unit 13 (step S32). For example, as shown in the graph of FIG. 10, there are a classification item C201 to which D202 belongs and a classification item C202 to which D203 belongs, and the classification item C201 to which D202 belongs is selected.

  Then, the classification unit 14 obtains a predetermined number n1 of manufacturing data belonging to the classification item C201 from the classified data storage unit 13 (step S33). This manufacturing data list is denoted as L301. For example, as shown in the graph of FIG. 10, this list L301 is a list having the manufacturing data D202 as an element.

  Thereafter, the classification unit 14 calculates n1 similarity degrees between the time trend of the manufacturing data D201 and the time trend of the manufacturing data included in the acquired manufacturing data list L301 (step S34).

  If all the classification items stored in the classified data storage unit 13 have not been selected (step S35), the process returns to step S32.

  On the other hand, when all the classification items stored in the classified data storage unit 13 are selected (step S35), the process proceeds to step S36.

  After that, among the similarities calculated in step S34, the classification item to which the manufacturing data having the highest similarity belongs is set as the classification item of the manufacturing data D201. For example, as shown in the graph of FIG. 10, since the degree of similarity with the manufacturing data D202 is the maximum, the manufacturing data D201 is classified into the classification item C201.

  Here, when the classification unit 14 obtains the manufacturing data list L301 which is the manufacturing data to be compared, even if the number of recorded manufacturing data becomes enormous by narrowing the comparison target to the predetermined number n1. The classification determination process can be performed efficiently.

  Next, the method by which the classified data storage unit 13 stores normal manufacturing data will be described more specifically.

  The classified data storage unit 13 classifies and stores, as normal manufacturing data, manufacturing data in which no abnormality is detected in the manufacturing apparatus 20 during a predetermined period t3, as normal manufacturing data. . For example, as shown in the graph of FIG. 13, the manufacturing data D201 in the current manufacturing period has not been determined to be abnormal until the period t3. Therefore, the manufacturing data D201 is classified into the same classification items as the manufacturing data D202 and classified as normal manufacturing data. To be recorded. Here, for example, the period t3 may coincide with the maintenance cycle cycle of the manufacturing apparatus 20 mentioned in the description of the temporary storage unit 12.

  Next, an abnormality determination method corresponding to step S25 in the flowchart of FIG. 2 will be described.

  The abnormality determination unit 15 acquires a predetermined number of past normal manufacturing data belonging to the same classification as the manufacturing data of the current manufacturing period from the classified data storage unit 13, and the time transition of the manufacturing data of the current manufacturing period, Compared with the time transition of each of the acquired manufacturing data, the similarity of the change tendency is calculated, respectively, and when the calculated minimum similarity is below a predetermined threshold, it is determined as abnormal.

  Next, the manufacturing data abnormality determination method by the abnormality determination unit 15 will be described more specifically with reference to the flowchart of FIG. First, the abnormality determination unit 15 acquires the manufacturing data D401 classified by the classification unit 14 (step S41).

  Thereafter, the abnormality determination unit 15 acquires a predetermined number of manufacturing data belonging to the same classification as the manufacturing data acquired from the classification unit 14 from the normal manufacturing data stored in the classified data storage unit 13. (Step S42). This manufacturing data list is L401.

  Then, the abnormality determination unit 15 compares the tendency of the time transition of each manufacturing data included in the manufacturing data list L401 with the tendency of the time transition of the manufacturing data D401 to be determined, and obtains the similarity (step S43). ).

  After that, when all the calculated similarities are below a predetermined threshold, it is determined as abnormal (step S44). Further, when it is not determined to be abnormal (step S45) and a predetermined period has elapsed since the temporary storage unit 12 stores the data (step S46), the manufacturing data D401 to be monitored is determined to be normal. (Step S47).

  Next, a method for calculating similarity between manufacturing data performed by the classification unit 14 and the abnormality determination unit 15 will be described.

  Each of the classification unit 14 and the abnormality determination unit 15 obtains normal manufacturing data belonging to the same classification as the (classified) manufacturing data of the current manufacturing period from the classified data storage unit 13 when calculating the similarity. Then, the manufacturing data of the current manufacturing period and the acquired manufacturing data are divided into time intervals based on the characteristics of the time transition, and the time transition and acquisition of the manufacturing data of the current manufacturing period are obtained for each divided section. The time transition of the manufactured data is compared, the similarity of the change tendency is calculated, and the minimum similarity is output from the calculated similarities.

  In order to compare time-series data that changes with time and obtain a similarity, the accuracy of the entire time-series data is reduced. Therefore, in the embodiment of the present invention, the accuracy can be improved by dividing the time intervals and comparing sections having the same time transition tendency.

  In the classification unit 14, the similarity is calculated in step S34 in the flowchart of FIG. Further, the abnormality determination unit 15 performs similarity calculation in step S43 of the flowchart of FIG.

  Next, the similarity calculation method performed by the classification unit 14 and the abnormality determination unit 15 will be described more specifically with reference to the flowchart of FIG. 5 and the graph of FIG. 14.

  First, the manufacturing data D401 acquired from the classification unit 14 or the abnormality determination unit 15 and the manufacturing data D402 stored in the classified data storage unit 13 selected as the comparison target are acquired (step S51).

  Thereafter, the production data D401 and D402 are respectively divided at time intervals (step S52). As a method of dividing manufacturing data at time intervals, for example, there is a method of decomposing the data into a plurality of time-series data having different frequency bands by wavelet transform processing or window Fourier transform.

  For example, as another method of decomposing manufacturing data at time intervals, when a change with time in a normal pattern of manufacturing data occurs at a predetermined time, the manufacturing data may be divided at a predetermined interval. For example, in the graph of FIG. 14, in the classification to which the manufacturing data D402 belongs, it is known that the time change occurs at the manufacturing time t1 + t2, and the sections before the time t1 + t2 are compared with the sections after the time t1 + t2.

  Then, one section is selected from the divided sections of D401 and set as D401 ′ (step S53), and the section D402 ′ corresponding to the selected section D401 ′ is selected from the divided sections of D402 (step S53). S54), the tendency of the time transition of the section D401 ′ and the tendency of the time transition of the section D402 ′ are compared, and the similarity is calculated (step S55).

  Thereafter, when the similarity is obtained for all the sections (step S56), the minimum similarity is obtained from the obtained similarities (step S57).

  On the other hand, when similarity is not calculated | required with respect to all the sections (step S56), it returns to step S53. Here, in this embodiment, as the degree of similarity, the difference between the regression coefficients obtained from the production data that is time series data is used. Further, as the degree of similarity, for example, a difference in dispersion of manufacturing data (more precisely, a reciprocal of the difference in dispersion) may be used.

  According to this embodiment, even for manufacturing data that changes over time over many stages, it is similar to each other with high accuracy by dividing the time for each feature of change over time and comparing the trends in each section. The degree can be calculated.

  The abnormality determination method of the present invention includes a collection step, a temporary storage step, a classified data storage step, a classification step, and an abnormality determination step. The collection step collects manufacturing data from the manufacturing apparatus 20 in time series. The temporary storage step sequentially stores the manufacturing data collected at the collecting step and the manufacturing data at each time in a predetermined manufacturing period. The classified data storage step classifies the manufacturing data collected in the collecting step and a plurality of manufacturing data in the past in the manufacturing period into sets of manufacturing data having similar time transition trends in the manufacturing period. And remember. The classification step stores the manufacturing data of the current manufacturing period stored in the temporary storage step in the classified data storage step having a time transition similar to the time transition of the manufacturing data of the current manufacturing period. Classify into production data sets. The abnormality determination step includes a time transition of manufacturing data in the current manufacturing period and a time transition of manufacturing data stored in the classified data storing step in a set of manufacturing data in which the manufacturing data in the current manufacturing period is classified. Is detected, and the abnormality is detected by obtaining the similarity.

  Therefore, in the first phase, it is determined to which classification the manufacturing data to be judged belongs, and in the second phase, the manufacturing data deviates with time in the classification group to which it belongs. It is judged whether it is abnormal by monitoring. As described above, by classifying the manufacturing data according to the characteristics of the change in the time transition, it is possible to accurately detect the abnormality even when the change of the manufacturing data with time is not uniquely determined.

(Second Embodiment)
FIG. 6 shows a second embodiment of the abnormality determination device of the present invention. As shown in FIG. 6, in the abnormality determination device 60 of the second embodiment, among the configurations of the abnormality determination device of the first embodiment shown in FIG. 14 and the abnormality determination unit 15, a classified data storage unit 63 and an abnormality determination unit 65 are included. Note that the same reference numerals as those in the first embodiment have the same configurations as those in the first embodiment, and thus description thereof is omitted.

  The abnormality determination device 60 includes a collection unit 11, a temporary storage unit 12, a classified data storage unit 63, and an abnormality determination unit 65.

  The collection unit 11 collects manufacturing data from the manufacturing apparatus 20 in time series. The temporary storage unit 12 sequentially stores the manufacturing data collected by the collecting unit 11 and the manufacturing data at each time in a predetermined manufacturing period.

  The classified data storage unit 63 classifies the manufacturing data collected by the collecting unit 11 and stores the data of a plurality of past manufacturing periods into sets of manufacturing data having similar time transition trends in the manufacturing period. To do.

  The abnormality determination unit 65 compares the time transition of the manufacturing data of the current manufacturing period stored in the temporary storage unit 12 with the time transition of the manufacturing data of each classification stored in the classified data storage unit 63. Thus, an abnormality is detected by obtaining similarity.

  When calculating the similarity of the time transition between the manufacturing data, the abnormality determining unit 65 acquires the manufacturing data determined to be abnormal belonging to each classification from the classified data storage unit 63, and manufacturing data for the current manufacturing period. Is divided into time intervals from the characteristics of the time transition, and for each of the divided sections, the change in the time transition of the manufacturing data during the current manufacturing period and the time transition of the manufacturing data determined to be abnormal The similarity is calculated, and the maximum similarity is obtained from the calculated similarities.

  In the abnormality determination device 60 of the second embodiment, abnormality detection is performed based on past abnormal manufacturing data. Contrary to the first embodiment, the abnormality determination apparatus 60 is more abnormal than normal data at the time of factory startup or the like. This is effective when data is easier to model. In addition, by using the abnormality determination unit 15 of the first embodiment and the abnormality determination unit 65 of the second embodiment together, a check with past data in a normal state and a check with past data in an abnormal state are performed. By performing the double check, the accuracy of abnormality detection can be improved.

  Next, a method in which the classified data storage unit 63 stores abnormal manufacturing data will be described in more detail.

  The classified data storage unit 63 extracts characteristic time-series sections determined to be abnormal from the manufacturing data of the current manufacturing period detected as abnormal by the abnormality determination unit 65, and manufactures the extracted section The data is classified and stored as similar production data sets as abnormal production data.

  Here, for example, as a process of extracting a section showing characteristic fluctuations from the abnormal manufacturing data, the target manufacturing data is decomposed into a plurality of time series data having different frequency bands by wavelet transform or window Fourier transform, There is a method of extracting a section including the time when the abnormality occurred.

  Further, for example, as a method of extracting a section showing a characteristic variation from the abnormal manufacturing data, a section where the regression coefficient of the regression equation suddenly changes before and after the time when the abnormality is determined is extracted. Also good.

  Further, for example, as processing for extracting a section showing a characteristic variation from the abnormal manufacturing data, time series data may be visualized to extract a section where the feature can be seen.

  In this embodiment, as shown in the graph of FIG. 12, when it is determined that the manufacturing data D201 is abnormal, the section in which the abnormality has occurred is decomposed by the frequency band, and the manufacturing data D801 is shown in the graph of FIG. Take out.

  According to this embodiment, the classified data storage unit 63 stores the classification result at the end of manufacturing data storage by the temporary storage unit 12 together with the abnormal state, thereby classifying and accumulating changes over time at the time of abnormality. be able to. By comparing with the time-dependent change at the time of abnormality, the manufacturing data to be monitored can be used for abnormality detection when a characteristic abnormality change that occurred in the past has occurred.

  Next, the abnormality determination unit 65 will be described in more detail. The abnormality determination unit 65 acquires abnormal manufacturing data belonging to each classification from the classified data storage unit 63, and compares the time transition of the manufacturing data in the current manufacturing period with the time transition of the abnormal manufacturing data of each acquired classification. Then, the similarity of the change tendency is obtained, and when the similarity that exceeds the predetermined threshold is calculated, it is determined as abnormal.

  In the first embodiment, the abnormality determination unit 15 obtains the similarity only for the manufacturing data belonging to the same classification, but the abnormality determination unit 65 determines that all the classifications are determined to be abnormal. On the other hand, the similarity is obtained. This is because there may be a plurality of abnormal temporal change patterns.

  Next, the process of the abnormality determination unit 65 that determines abnormality based on the past abnormal manufacturing data will be described more specifically with reference to the flowchart of FIG. First, the manufacturing data D701 memorize | stored in the said temporary memory part 12 is acquired (step S71).

  Thereafter, one category item to which the manufacturing data determined to be abnormal belongs is selected, and this is set as C701 (step S72), and one piece of manufacturing data belonging to the selected category item C701 is selected as D702 (step S73).

  And similarity is calculated | required from the manufacturing data D701 and the manufacturing data D702 (step S74). Here, the method for obtaining the similarity is different from the method for calculating the similarity, and will be described later.

  Thereafter, it is determined whether or not the similarity exceeds a predetermined threshold (step S75). If the similarity exceeds a predetermined threshold, it is determined as abnormal and the process ends (step S76).

  On the other hand, if the similarity does not exceed the predetermined threshold, it is determined whether or not a predetermined number of manufacturing data is acquired from the manufacturing data belonging to the classification item C701 and the similarity is obtained (step S77). If a predetermined number of manufacturing data is obtained from the manufacturing data belonging to the classification item C701 and the similarity is obtained, the process proceeds to step S78, while a predetermined number of manufacturing data is manufactured from the manufacturing data belonging to the classification item C701. When data is acquired and the similarity is not obtained, the process returns to step S73.

  Thereafter, when the similarity calculation has been completed for all the classification items (step S78), the abnormality determination process is ended. On the other hand, if similarity calculation has not been completed for all classification items (step S78), the process returns to step S72.

  Next, the similarity calculation process performed by the abnormality determination unit 65 will be described. When calculating the similarity, the abnormality determination unit 65 divides the current manufacturing period by the time interval of the manufacturing data in the current manufacturing period, and divides it by the time interval from the characteristics of the manufacturing data time transition in the current manufacturing period. Then, the temporal transition of abnormal manufacturing data is compared, the similarity of change tendency is calculated, and the maximum similarity is output from the calculated similarities.

  Next, when the abnormality is determined based on the past abnormal manufacturing data using the flowchart of FIG. 8, the graph of FIG. 12, and the graph of FIG. 15, The similarity calculation method based on the comparison with the trend of the time transition of the abnormal manufacturing data stored in the classified data storage unit 63 will be described more specifically.

  First, the abnormal manufacturing data D801 stored in the classified data storage unit 63 and the manufacturing data D201 for the current manufacturing period are acquired (step S81), and the acquired manufacturing data D802 is divided at time intervals from the characteristics of the time transition. (Step S82).

  Here, the production data D801 is not divided. This is because when the classified data storage unit 63 stores the manufacturing data determined to be abnormal, the section where the abnormality has occurred is extracted from the manufacturing data and recorded.

  Thereafter, one of the divided sections of the manufacturing data D201 is selected as D201 ′ (step S83), and the trend of the time transition of the manufacturing data D201 ′ is compared with the trend of the time transition of the abnormal manufacturing data section D801. The degree of similarity is obtained (step S84).

  Here, for example, as a method of obtaining the similarity from the tendency of time transition, there is a method of obtaining a difference of regression coefficients from a regression equation or a difference of variance as described in the first embodiment.

  Thereafter, when the similarity is calculated in all the divided sections (step S85), the maximum similarity is output among the calculated similarities (step S86), and the process is terminated. On the other hand, when the similarity is not calculated in all the divided sections (step S85), the process returns to step 83.

  In this embodiment, referring to the graph of FIG. 12 and the graph of FIG. 15, it is similar by taking the difference between the variance of each section before and after time t1 + t2 of the production data D201 and the variance of the production data D801. (Accurately, the reciprocal of the difference in variance is used as the similarity). As a result of the similarity calculation, the similarity with the section after time t1 + t2 is maximized, and this similarity is output.

  The abnormality determination method of the present invention includes a collection step, a temporary storage step, a classified data storage step, and an abnormality determination step. The collection step collects manufacturing data from the manufacturing apparatus 20 in time series. The temporary storage step sequentially stores the manufacturing data collected at the collecting step and the manufacturing data at each time in a predetermined manufacturing period. The classified data storage step classifies the manufacturing data collected in the collecting step and a plurality of manufacturing data in the past in the manufacturing period into sets of manufacturing data having similar time transition trends in the manufacturing period. And remember. The abnormality determination step compares the time transition of the manufacturing data of the current manufacturing period stored in the temporary storage step with the time transition of the manufacturing data of each classification stored in the classified data storage step. By detecting the similarity, the abnormality is detected. In the abnormality determination step, when calculating the similarity of the time transition between the manufacturing data, the manufacturing data determined to be abnormal belonging to each classification is acquired from the classified data storage step, and manufacturing in the current manufacturing period is performed. The data is divided into time intervals from the characteristics of the time transition, and for each of the divided sections, the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal The similarity to the change tendency is calculated, and the maximum similarity is obtained from the calculated similarities.

  Therefore, the abnormality determination step compares the abnormal manufacturing data recorded in the classified data storage step with the current manufacturing data to be determined. As a result, even when the temporal change of the manufacturing data in the normal state is not uniquely determined, it is possible to detect the abnormality with high accuracy by determining that the manufacturing data having the same temporal change as the known abnormal state is abnormal.

  In addition, this invention is not limited to the above-mentioned embodiment. For example, in addition to the configuration of the abnormality determination unit 15 of the first embodiment configured to perform abnormality determination based on past data in a normal state, like the abnormality determination unit 65 of the second embodiment, It is good also as a structure which performs abnormality determination based on the past data of an abnormal state. Alternatively, the configuration of the abnormality determination unit 15 of the first embodiment may be configured to perform abnormality determination based on normal state past data and abnormal state past data.

  As an abnormality determination program of the present invention, a computer is used as the collection unit 11, the temporary storage unit 12, the classified data storage unit 13, the classification unit 14, and the abnormality determination unit 15 of the first embodiment. Or you may make it function as the said collection part 11, the said temporary storage part 12, the said classified data storage part 63, and the said abnormality determination part 65 of the said 2nd Embodiment. Therefore, by installing this program in a computer, the abnormality determination device can be realized by the computer.

  Further, the abnormality determination program may be recorded as the program recording medium of the present invention, and this recording medium is, for example, a computer-readable CD-ROM. Therefore, since the abnormality determination program is recorded, the abnormality determination program can be easily installed in the computer.

DESCRIPTION OF SYMBOLS 10 Abnormality determination apparatus 11 Collection part 12 Temporary storage part 13 Classified data storage part 14 Classification part 15 Abnormality determination part 20 Manufacturing apparatus 60 Abnormality determination apparatus 63 Classified data storage part 65 Abnormality determination part

Claims (9)

A collection unit that collects manufacturing data from the manufacturing device in time series;
A temporary storage unit for sequentially storing the manufacturing data collected at the collection unit at each time in a predetermined manufacturing period.
A classified data storage unit that stores the manufacturing data collected by the collection unit in the past and a plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. When,
The manufacturing data of the current manufacturing period stored in the temporary storage unit is converted into a set of manufacturing data stored in the classified data storage unit having a time transition similar to the time transition of the manufacturing data of the current manufacturing period. A classification part for classification;
Compare the time transition of the manufacturing data of the current manufacturing period with the time transition of the manufacturing data stored in the classified data storage unit in the set of manufacturing data in which the manufacturing data of the current manufacturing period is classified, An abnormality determination device comprising: an abnormality determination unit that detects abnormality by obtaining similarity. In the abnormality determination device according to claim 1,
The classification unit and the abnormality determination unit are respectively
When calculating the similarity of the time transition between the manufacturing data,
The manufacturing data determined to be normal belonging to the same classification as the manufacturing data of the current manufacturing period is acquired from the classified data storage unit,
The manufacturing data of the current manufacturing period and the manufacturing data determined to be normal are divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity to the change trend between the time transition of the manufacturing data of the current manufacturing period and the time transition of the manufacturing data determined to be normal,
An abnormality determination device characterized in that a minimum similarity is obtained from the calculated similarities. In the abnormality determination device according to claim 1,
The abnormality determination unit
When calculating the similarity of the time transition between the manufacturing data,
Manufacturing data determined to be abnormal belonging to each classification is obtained from the classified data storage unit,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
An abnormality determination device characterized in that a maximum similarity is obtained from the calculated similarities. In the abnormality determination device according to claim 1,
The abnormality determination unit
When calculating the similarity of the time transition between the manufacturing data,
The manufacturing data determined to be normal belonging to the same classification as the manufacturing data of the current manufacturing period is acquired from the classified data storage unit,
The manufacturing data of the current manufacturing period and the manufacturing data determined to be normal are divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity to the change trend between the time transition of the manufacturing data of the current manufacturing period and the time transition of the manufacturing data determined to be normal,
While obtaining the minimum similarity from the calculated similarities,
Manufacturing data determined to be abnormal belonging to each classification is obtained from the classified data storage unit,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
An abnormality determination device characterized in that a maximum similarity is obtained from the calculated similarities. A collection unit that collects manufacturing data from the manufacturing device in time series;
A temporary storage unit for sequentially storing the manufacturing data collected at the collection unit at each time in a predetermined manufacturing period.
A classified data storage unit that stores the manufacturing data collected by the collection unit in the past and a plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. When,
By comparing the time transition of the manufacturing data of the current manufacturing period stored in the temporary storage unit with the time transition of the manufacturing data of each classification stored in the classified data storage unit, and obtaining similarity With an abnormality determination unit that detects an abnormality,
The abnormality determination unit
When calculating the similarity of the time transition between the manufacturing data,
Manufacturing data determined to be abnormal belonging to each classification is obtained from the classified data storage unit,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
An abnormality determination device characterized in that a maximum similarity is obtained from the calculated similarities. A collection step for collecting manufacturing data from the manufacturing equipment in time series;
Temporary storage step for sequentially storing the manufacturing data collected at the collecting step at each time in a predetermined manufacturing period;
The classified data storage that is the manufacturing data collected in the collecting step and stores the manufacturing data of the plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. Steps,
The manufacturing data of the current manufacturing period stored in the temporary storage step is converted into a set of manufacturing data stored in the classified data storage step having a time transition similar to the time transition of the manufacturing data of the current manufacturing period. A classification step to classify;
Compare the time transition of the manufacturing data of the current manufacturing period with the time transition of the manufacturing data stored in the classified data storage step in the set of manufacturing data in which the manufacturing data of the current manufacturing period is classified, An abnormality determination method comprising: an abnormality determination step of detecting an abnormality by determining similarity. A collection step for collecting manufacturing data from the manufacturing equipment in time series;
Temporary storage step for sequentially storing the manufacturing data collected at the collecting step at each time in a predetermined manufacturing period;
The classified data storage that is the manufacturing data collected in the collecting step and stores the manufacturing data of the plurality of manufacturing periods in the past as a group of manufacturing data having similar time transition trends in the manufacturing period. Steps,
By comparing the time transition of the manufacturing data of the current manufacturing period stored in the temporary storage step with the time transition of the manufacturing data of each classification stored in the classified data storage step, and obtaining similarity With an abnormality determination step for detecting an abnormality,
In the abnormality determination step,
When calculating the similarity of the time transition between the manufacturing data,
Manufacturing data determined to be abnormal belonging to each classification is acquired from the above classified data storage step,
The production data for the current production period is divided at time intervals from the characteristics of time transition,
For each of the divided sections, calculate the degree of similarity with respect to a change trend between the time transition of the manufacturing data in the current manufacturing period and the time transition of the manufacturing data determined to be abnormal.
An abnormality determination method characterized in that a maximum similarity is obtained from the calculated similarities.   The computer according to claim 1, the collection unit according to claim 1, the temporary storage unit, the classified data storage unit, the classification unit, and the abnormality determination unit, or the collection unit according to claim 5, the temporary storage And an abnormality determination program that functions as the classified data storage unit and the abnormality determination unit.   A computer-readable program recording medium on which the abnormality determination program according to claim 8 is recorded.

Images