JP2023178206A - Abnormality detection device, abnormality detection method and program - Google Patents

Abstract

To provide an abnormality detection method capable of simply checking whether or not there is any abnormality in a monitoring target in a production field.SOLUTION: An abnormality detection method includes: a step (1) of acquiring a standard value for each unit time of a plurality of variables indicating a status of a production line when the production line is properly operating; a step (2) of collecting actual values of the plurality of variables per unit time; a step (3) of calculating, for each unit time, statistics representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables; and a step (4) of detecting, in response to the statics out of a reference range, that abnormality occurs in the production line.SELECTED DRAWING: Figure 1

Description

The present disclosure relates to an anomaly detection device, an anomaly detection method, and a program.

In order to increase production efficiency, it is desirable to detect abnormalities on the production line early. JP 2019-177783 A (Patent Document 1) discloses that the ride comfort of a railway vehicle deteriorates due to an increase in the Mahalanobis distance between characteristic data obtained from an acceleration sensor installed in the railway vehicle and unit data acquired in advance. Discloses technology to detect.

Japanese Patent Application Publication No. 2019-177783 Japanese Patent Application Publication No. 2003-44115

Kishio Tamura, “5th MT system that can determine direction - TS method, T method”, Standardization and Quality Control, 2009, Vol. 62, No. 2 Genichi Taguchi, "Objective function and basic function (6) - Comprehensive prediction using T method -", Quality Engineering, 2005, Vol. 13, No. 3, p.5-10 "Considerations and utilization of quality engineering in the development and design stage - System evaluation and improvement without prototyping and testing", [online], [Retrieved March 22, 2023], Internet <https://foundry.jp/bukai /wp-content/uploads/2012/07/e4806f10b0797ec0932d9317dd92a533.pdf>

Production lines include a wide variety of equipment and provide a large number of variables. Therefore, when the technology disclosed in Patent Document 1 is applied to a production line, it is necessary to compare the reference and actual measured values for each of a large number of variables in order to confirm the presence or absence of an abnormality on the production line. As a result, the work involved in checking whether there is any abnormality on the production line increases.

The present disclosure has been made in view of the above problems, and its purpose is to provide an abnormality detection device, an abnormality detection method, and a program that can easily confirm the presence or absence of an abnormality in a monitoring target at a production site.

According to an example of the present disclosure, an abnormality detection device includes an acquisition section, a collection section, a calculation section, and a detection section. The acquisition unit acquires standard values of a plurality of variables indicating the status of the monitoring target in a unit section when the monitoring target at the production site is operating normally. The collection unit collects actual values of a plurality of variables in a unit interval. The calculation unit calculates a statistic representing a degree of deviation between actual values of the plurality of variables with respect to standard values of the plurality of variables. The detection unit detects that an abnormality has occurred in the monitoring target in response to the statistical amount being outside the reference range.

According to the above disclosure, by checking the results of the detection unit, the user can easily check whether there is an abnormality in the monitoring target at the production site.

In the above disclosure, the anomaly detection device further includes a selection unit that selects an effective variable that has a relatively large influence on the statistic from among the plurality of variables.

According to the above disclosure, the user can identify the effective variable as a candidate cause of the abnormality to be monitored by checking the selection result by the selection unit. As a result, the user can quickly take measures such as adjusting the effective variables, reviewing the maintenance of the monitored object, or reviewing the production plan.

In the above disclosure, the selection unit selects, as the effective variable, a variable that is effective in improving the estimation accuracy of the T-method estimation model that uses a plurality of variables as explanatory variables and a statistic as an objective variable.

According to the above disclosure, effective variables that have a relatively large influence on statistics can be easily selected.

In the above disclosure, the abnormality detection device further includes an analysis unit that performs principal component analysis using actual values and statistics of effective variables for each unit section.

According to the above disclosure, the user can narrow down variables that have a high degree of influence on statistics from among a plurality of effective variables based on the principal component loading amount obtained by principal component analysis. As a result, the user can quickly take measures such as maintaining the production line or reviewing the production plan to adjust the narrowed-down variables.

In the above disclosure, the analysis unit selects two principal components that maximize the variance of the statistics, and calculates the effective variables based on the principal component loadings for the two principal components in each of the effective variables and the statistics. Determine the impact on the statistics.

According to the above disclosure, the user can narrow down effective variables that have a high degree of influence on statistics from among a plurality of valid variables based on the specified degree of influence. As a result, the user can quickly take measures such as adjusting the narrowed-down effective variables and reviewing the maintenance and production plans to be monitored.

In the above disclosure, the acquisition unit uses the virtual model of the monitored target to perform a simulation of the values of the plurality of variables when the monitored target is operated according to the production plan, and standardizes the value of the unit interval obtained by the simulation. Get it as a value.

Alternatively, in the above disclosure, the acquisition unit may acquire values of a plurality of variables in a unit interval when the monitoring target was operating normally in the past as standard values.

In the above disclosure, the statistic is, for example, the standard signal-to-noise ratio. Thereby, the statistic appropriately represents the degree of similarity between the actual value and the standard value.

In the above disclosure, the monitoring target is a production line, and the unit interval is a unit time. According to this disclosure, a user can easily check whether there is an abnormality in the production line by checking the results of the detection unit.

In the above disclosure, the monitored target includes manufacturing equipment. According to this disclosure, a user can easily check whether there is an abnormality in the manufacturing apparatus by checking the results of the detection unit.

In the above disclosure, the manufacturing apparatus repeatedly performs the target operation. The unit interval is a period during which the target operation is executed. According to this disclosure, the plurality of variables indicate the status of the manufacturing apparatus during the period in which the target operation is being performed. Therefore, the user can easily check whether there is an abnormality for each target operation repeatedly executed by the manufacturing apparatus.

According to another example of the present disclosure, an anomaly detection method includes first to fourth steps. The first step is to obtain standard values of a plurality of variables indicating the status of the monitoring target in a unit interval when the monitoring target is operating normally. The second step is a step of collecting actual values of a plurality of variables in a unit interval. The third step is a step of calculating a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables. The fourth step is a step of detecting that an abnormality has occurred in the monitoring target in response to the statistical value being outside the reference range.

According to another example of the present disclosure, a program causes a computer to execute the above abnormality detection method. These disclosures also allow the user to easily check whether there is an abnormality in the monitoring target at the production site.

According to the present disclosure, a user can easily check whether there is an abnormality in a monitoring target at a production site.

1 is a schematic diagram showing the configuration of a system including an abnormality detection device according to an embodiment. It is a diagram showing an example of a production line. FIG. 1 is a schematic diagram showing an example of a hardware configuration of an abnormality detection device according to an embodiment. 1 is a diagram illustrating an example of a functional configuration of an abnormality detection device according to an embodiment. FIG. 3 is a diagram illustrating processing of a simulation execution unit. It is a figure explaining the processing of a calculation part. It is a figure which shows an example of the selection result of an effective variable. It is a figure which shows an example of the analysis result by an analysis part. It is a flowchart which shows an example of the flow of processing of an abnormality detection device. 7 is a diagram illustrating a functional configuration of an abnormality detection device according to Modification 1. FIG. FIG. 2 is a diagram illustrating an example of a manufacturing device to be monitored. FIG. 3 is a diagram showing changes in parameter values during a period in which an injection molding operation is performed. FIG. 3 is a diagram showing standard values and actual values of a plurality of variables indicating the status of the injection molding machine 3. FIG. FIG. 3 is a diagram showing changes in statistics calculated for each injection molding operation. It is a figure which shows the result of the principal component analysis using the actual value of the effective variable for each injection molding period, and a standard SN ratio. FIG. 7 is a diagram illustrating another example of a manufacturing device to be monitored. FIG. 7 is a diagram showing changes in statistics calculated for each joining operation.

Embodiments of the present invention will be described in detail with reference to the drawings. Note that the same or corresponding parts in the figures are given the same reference numerals and the description thereof will not be repeated.

§1 Application Example An example of a situation where the present invention is applied will be described with reference to FIG. FIG. 1 is a schematic diagram showing the configuration of a system including an abnormality detection device according to an embodiment. As shown in FIG. 1, the system 1 includes a production line 2, which is an example of a monitoring target at a production site, and an abnormality detection device 100 that detects the occurrence of an abnormality in the production line 2. The abnormality detection device 100 is communicably connected to the production line 2. Note that the monitoring target at the production site is not limited to the production line 2.

The abnormality detection device 100 obtains standard values for each unit time of a plurality of variables indicating the status of the production line 2 when the production line 2 is operating normally (step (1)). The plurality of variables include the number of devices, number of personnel, output, etc. of each process making up the production line 2. The standard values of the plurality of variables represent the standard situation of the production line 2. The unit time is, for example, 5 minutes. Further, the standard values of the plurality of variables may differ for each unit time. For example, the standard value in a unit time immediately after the production line 2 starts operating may be different from the standard value in a unit time several hours after the production line 2 starts operating. Considering such a case, the abnormality detection device 100 acquires a standard value for each unit time.

Next, when the production line 2 starts operating, the abnormality detection device 100 collects actual values of a plurality of variables for each unit time (step (2)). Then, the abnormality detection device 100 calculates a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables for each unit time (step (3)). The abnormality detection device 100 detects that an abnormality has occurred in the production line 2 in response to the statistical value being outside the reference range (step (4)).

In step (3), even if the number of variables indicating the situation of the production line 2 is large, the degree of deviation of the current situation from the standard situation of the production line 2 is expressed by one statistic. For example, a standard SN ratio may be employed as the statistic. However, the statistic is not limited to the standard SN ratio, and may be any parameter that represents the degree of deviation between the standard value and the actual value for a plurality of variables. The standard SN ratio becomes smaller as the degree of deviation between the standard value and the actual value for a plurality of variables increases. Therefore, a decrease in the standard SN ratio means that the actual values of a plurality of variables deviate from the standard values. The threshold value TH is predetermined to be a value when the production line 2 changes from normal to abnormal. In this case, the reference range is a range equal to or greater than the threshold value TH. Therefore, the fact that the standard SN ratio is outside the reference range (that is, less than the threshold TH) means that some kind of abnormality has occurred in the production line 2. Therefore, the user can easily check whether there is any abnormality in the production line 2 by checking the result of step (4) above.

§2 Specific example <Production line example>
FIG. 2 is a diagram showing an example of a production line. The production line 2 shown in FIG. 2 includes a cutting process 21, a parts shipping process 22, an assembly process 23, and a product shipping process 24.

In the cutting process 21, a metal part having a predetermined shape is produced by, for example, cutting a metal piece. The cutting process 21 includes one or more cutting devices 21a, a timer 21b, and counters 21c and 21d.

Each of the one or more cutting devices 21a sends a first signal indicating whether or not it is operating normally, a second signal indicating whether or not it is performing a cutting operation, and a signal indicating that the production of the metal part has been completed. A third signal indicating .

The counter 21c measures the number of cutting devices 21a operating normally in the cutting process 21 (hereinafter referred to as the "number of devices") based on the first signal.

The timer 21b measures the time during which the cutting operation is performed (hereinafter referred to as "cutting time") within a unit time based on the second signal.

The counter 21d measures the number of metal parts manufactured per unit time (hereinafter referred to as "output (cutting process)") based on the third signal.

The parts shipping process 22 transports assembly parts purchased from outside to an assembly process 23. The parts shipping process 22 includes a sensor 22a and an automatic guided vehicle (AGV) 22b. In the parts shipping process 22, after a worker takes out the assembly parts purchased from the outside from the packaging, the automatic transport vehicle 22b transports the assembly parts.

The sensor 22a measures the number of workers (hereinafter referred to as "number of workers (parts shipping process)") every unit time. The sensor 22a is, for example, an image sensor, and measures the number of personnel (parts shipping process) using an image obtained by capturing an image of the parts shipping process 22.

The automatic guided vehicle 22b has a timer 22c and a counter 22d. The timer 22c measures the time during which the assembled parts are being transported (hereinafter referred to as "AGV operating time") per unit time. The counter 22d measures the number of assembled parts that have been transported (hereinafter referred to as "output (components shipping process)") in a unit time.

In the assembly process 23, a product is manufactured by assembling the assembly parts conveyed from the parts shipping process 22 onto the metal parts produced in the cutting process 21. In the assembly step 23, for example, a worker performs assembly work, and the completed product is placed on a belt conveyor.

The assembly process 23 includes a sensor that measures the number of metal parts before assembly (hereinafter referred to as "number of work-in-progress") among the metal parts transported from the cutting process 21 for each unit time. 23a is provided. The sensor 23a is, for example, an image sensor, and measures the number of products in progress using an image obtained by capturing an image of a tray on which metal parts are placed before the parts to be assembled are placed.

Furthermore, the assembly process 23 includes a sensor 23b that measures the number of workers (hereinafter referred to as "number of workers (assembly process)") per unit time, and a sensor 23b that measures the number of workers per unit time. A counter 23c for measuring the number of pieces (hereinafter referred to as "output (assembly process)") is provided. The sensor 23b is, for example, an image sensor, and measures the number of personnel (assembly process) using an image obtained by capturing an image of the assembly process 23. The counter 23c is, for example, a limit switch provided on a belt conveyor on which products are placed. In the product shipping step 24, the product is packed and shipped.

<Hardware configuration of anomaly detection device>
FIG. 3 is a schematic diagram showing an example of the hardware configuration of the abnormality detection device according to the embodiment. As shown in FIG. 3, anomaly detection device 100 typically has a structure that follows a general-purpose computer architecture. Specifically, the abnormality detection device 100 communicates with a processor 101 such as a CPU (Central Processing Unit) or an MPU (Micro-Processing Unit), a memory 102, a storage 103, a display controller 104, and an input interface 105. An interface 106 is included. These units are connected to each other via a bus so that they can communicate data.

Processor 101 implements various processes according to the present embodiment by loading various programs stored in storage 103 into memory 102 and executing them.

The memory 102 is typically a volatile storage device such as DRAM (Dynamic Random Access Memory), and stores programs read from the storage 103.

Storage 103 is typically a nonvolatile magnetic storage device such as a hard disk drive. The storage 103 stores an abnormality detection program 131 and a simulation program 132 that are executed by the processor 101. Various programs installed in the storage 103 are distributed in a state stored in a memory card or the like.

The display controller 104 is connected to the display device 70, and outputs signals for displaying various information to the display device 70 according to internal commands from the processor 101.

Input interface 105 mediates data transmission between processor 101 and input device 75 such as a keyboard, mouse, touch panel, dedicated console, etc. That is, the input interface 105 receives an operation command given by the user operating the input device 75.

The communication interface 106 mediates data transmission between the processor 101 and external equipment (eg, various equipment provided on the production line 2). The communication interface 106 typically includes Ethernet (registered trademark), USB (Universal Serial Bus), and the like. Note that various programs stored in the storage 103 may be downloaded from a distribution server or the like via the communication interface 106.

When using a computer having a structure according to the general-purpose computer architecture as described above, in addition to an application for providing the functions according to this embodiment, an OS for providing the basic functions of the computer is required. (Operating System) may be installed. In this case, the program according to the present embodiment may execute processing by calling necessary modules in a predetermined order and timing from among program modules provided as part of the OS. That is, the program itself according to the present embodiment does not include the above-mentioned modules, and may execute processing in cooperation with the OS.

Alternatively, part or all of the functions provided by executing the abnormality detection program 131 and the simulation program 132 may be implemented as a dedicated hardware circuit.

<Functional configuration of anomaly detection device>
FIG. 4 is a diagram illustrating an example of the functional configuration of the abnormality detection device according to the embodiment. As shown in FIG. 4, the abnormality detection device 100 includes a simulation execution section 11, a collection section 12, a calculation section 13, a detection section 14, a selection section 15, and an analysis section 16. The simulation execution unit 11 is realized by the processor 101 (see FIG. 3) executing the simulation program 132. The collection unit 12 is realized by the communication interface 106 and the processor 101 that executes the abnormality detection program 131. The calculation unit 13, the detection unit 14, the selection unit 15, and the analysis unit 16 are realized by the processor 101 executing the abnormality detection program 131.

The simulation execution unit 11 operates as an acquisition unit that acquires standard values of a plurality of variables indicating the status of the production line 2 in a unit time when the production line 2 is operating normally. Specifically, the simulation execution unit 11 inputs production plan data indicating a production plan into the virtual model of the production line 2, thereby calculating the values of a plurality of variables indicating the status of the production line 2 every unit time (for example, 5 minutes). to simulate the value of . The simulation execution unit 11 may execute the simulation using a known technique (for example, Japanese Patent Laid-Open No. 2003-44115 (Patent Document 2)).

FIG. 5 is a diagram illustrating the processing of the simulation execution unit. The virtual model 2M is created in advance according to the production line 2. In response to input of various data indicating a production plan, the virtual model 2M outputs values for each unit time of a plurality of variables indicating the status of the production line 2 when the production line 2 is operated according to the production plan. As shown in Figure 5, multiple variables indicating the status of production line 2 include "cutting time", "number of devices", "output (cutting process)", "AGV operating time", and "number of personnel ( ``Amount of work completed (parts shipping process),'' ``Number of work in progress,'' ``Number of personnel (assembly process),'' and ``Amount of work (assembly process).''

The virtual model 2M includes a first model 21M corresponding to the cutting process 21, a second model 22M corresponding to the parts shipping process 22, and a third model 23M corresponding to the assembly process 23.

The first model 21M is created according to the specifications of the cutting device 21a. The first model 21M is defined by the following parameters.
- Cutting speed when the cutting device 21a operates normally,
・Cutting amount to make one metal part,
・Preparation time from the completion of cutting a metal part to the start of cutting the next metal part, etc. The first model 21M receives input of production plan data indicating the number of cutting devices 21a to be operated, scheduled operating period, stop period (break time), etc., and calculates the cutting processing time of the cutting process 21 for each unit time, the equipment Output the number and volume values.

The second model 22M is created according to the specifications of the automatic guided vehicle 22b and the specifications of the standard work of the worker in the parts shipping process 22. The second model 22M is defined by the following parameters.
・The time required for a worker to take out one assembled part from its packaging when performing standard work;
- The time required for the worker to set the assembled parts on the automated guided vehicle 22b;
- The time required for the automatic guided vehicle 22b to transport one assembly part to the assembly process 23, etc.
The second model 22M receives input of production plan data indicating the number of workers to be assigned to the parts shipping process 22, scheduled work periods, worker break times, etc., and operates the AGV in the parts shipping process 22 for each unit of time. Output values for time, number of personnel, and output.

The third model 23M is created according to the specifications of the standard work performed by the worker. The third model 23M is defined by, for example, the time required for standard work by a worker to complete one product. The third model 23M receives input of production plan data indicating the number of workers to be placed in the assembly process 23, scheduled work period, break time, etc., output data of the first model 21M, and output data of the second model 22M. Then, the values of the number of work in progress, number of personnel, and output of the assembly process 23 for each unit time are output.

The virtual model 2M is a model based on the premise that equipment and workers perform standard operations in each process. Therefore, the simulation execution unit 11 calculates the values of the plurality of variables output from the virtual model 2M for each unit time (for example, 1 minute, 5 minutes, etc.) when the production line 2 is operating normally. Obtain as a variable value (standard value).

The collection unit 12 shown in FIG. 4 collects actual values of a plurality of variables indicating the current status of the production line 2 for each unit of time (for example, 1 minute, 5 minutes, etc.) from the production line 2. Specifically, the collection unit 12 collects the results for each unit time of "cutting time", "number of devices", and "output (cutting process)" from the timer 21b, counter 21c, and counter 21d of the cutting process 21. Collect each value. The collection unit 12 collects the results for each unit time of "AGV operating time", "number of personnel (parts shipping process)", and "output (parts shipping process)" from the timer 22c, sensor 22a, and counter 22d of the parts shipping process 22. Collect each value. The collection unit 12 collects the actual values for each unit time of "number of work in process", "number of personnel (assembly process)", and "output (assemble process)" from the sensor 23a, sensor 23b, and counter 23c of the assembly process 23. Collect each.

The calculation unit 13 calculates a standard SN ratio as a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables for each unit time. The standard SN ratio can be found in "The concept and utilization of quality engineering at the development and design stage - System evaluation and improvement without prototyping and testing -", [online], [searched on January 4, 2022], Internet <https:/ /foundry.jp/bukai/wp-content/uploads/2012/07/e4806f10b0797ec0932d9317dd92a533.pdf>” (Non-Patent Document 3).

FIG. 6 is a diagram illustrating the processing of the calculation unit. Figure 6 shows multiple variables such as “cutting time”, “number of equipment”, “output (cutting process)”, “AGV operating time”, “number of personnel (parts shipping process)”, and “output (parts shipping process)”. An example is shown in which the following values are used: ``shipping process),'' ``number of work in progress,'' ``number of personnel (assembly process),'' and ``output (assembly process).'' The calculation unit 13 calculates the standard SN ratio of the actual values of the plurality of variables collected by the collection unit 12 with respect to the standard values of the plurality of variables calculated by the simulation execution unit 11 for each unit time. The calculation unit 13 may display a graph 30 showing changes in the calculated standard SN ratio on the display device 70. As a result, the user can grasp the change in the status of the production line 2 by checking the graph 30.

The detection unit 14 shown in FIG. 4 uses the standard SN ratio calculated by the calculation unit 13 to detect that an abnormality has occurred in the production line 2. The standard SN ratio becomes smaller as the degree of deviation between the standard value and the actual value for a plurality of variables increases. Therefore, the detection unit 14 detects that an abnormality has occurred in the production line 2 in response to the standard SN ratio being outside the reference range (that is, less than the threshold TH). For example, in the case of the graph 30 shown in FIG. 6, the detection unit 14 detects that an abnormality has occurred in the production line 2 at timing T when the standard SN ratio becomes less than the threshold value TH. The detection unit 14 may display an error notification on the display device 70 in response to detecting that an abnormality has occurred in the production line 2. Thereby, the user can reliably understand that an abnormality has occurred in the production line 2.

The selection unit 15 shown in FIG. 4 selects an effective variable that has a relatively large influence on the standard SN ratio from among the plurality of variables in response to the detection unit 14 detecting the occurrence of an abnormality. Specifically, the selection unit 15 selects, as effective variables, variables that are effective in improving the estimation accuracy of the T method estimation model that uses a plurality of variables as explanatory variables and the standard SN ratio as an objective variable. Thereby, the user can select an effective variable as a candidate for the cause of an abnormality on the production line 2, and can quickly eliminate the abnormality on the production line 2. Note that the selection unit 15 may select a plurality of effective variables.

The T method is one of the multivariate analysis methods provided by statistician Genichi Taguchi, and is a method for estimating the value of one objective variable from the values of multiple explanatory variables (see Non-Patent Document 1). ).

The selection unit 15 generates a T-method estimation model by performing T-method data processing on a data set showing the actual values of a plurality of variables and the standard SN ratio value for each unit time. In data processing using the T method, a simple regression is performed for each explanatory variable, and a weighted average is performed for each explanatory variable to generate a T method estimation model that calculates an estimated value of the objective variable. The estimated model obtained by data processing using the T method is expressed by the following equation (1).

In equation (1), Xij is the value of the j-th explanatory variable in the i-th unit time after normalization processing. The normalization process is a process of subtracting the average value of the values of variables indicated by the data set. ηj indicates the SN ratio of the j-th explanatory variable. The SN ratio ηj indicates linearity between the value of the j-th explanatory variable and the value of the objective variable, and represents the estimation accuracy of the value of the objective variable. βj represents a proportionality constant of simple regression. Mi is the estimated value of the objective variable at the i-th unit time after the normalization process.

For example, the selection unit 15 uses a two-level orthogonal array to select an effective variable that is effective for improving the estimation accuracy of the target variable from a plurality of variables. The selection unit 15 selects effective variables using the method described in Non-Patent Document 2, for example. The selection of effective variables (also called "item selection") using a two-level orthogonal array is effective in improving the estimation accuracy of the objective variable for each explanatory variable (also called "item"). In addition to selecting a certain explanatory variable, it is possible to exclude explanatory variables that have a negative effect on the estimation accuracy of the target variable (reducing the estimation accuracy). Furthermore, by using a two-level orthogonal array, the influence of interaction between items can be considered. Specifically, according to the method described in Non-Patent Document 2, an orthogonal array is created in which each of a plurality of explanatory variables (items) is given one of two levels: use or not. An overall signal-to-noise ratio is calculated for each row of the table. The selection unit 15 determines whether or not the item has an adverse effect on the estimation effect of the target variable and the estimation of the target variable, based on the total SN ratio when each item is used and when it is not used. The selection unit 15 selects an effective variable from a plurality of variables based on the determination result.

Alternatively, the selection unit 15 may generate a plurality of patterns indicating combinations of two or more variables among the plurality of variables without using a two-level orthogonal array. Then, the selection unit 15 generates a T-method estimation model using the actual values of two or more variables indicated by each pattern as explanatory variables and the standard SN ratio as an objective variable, and generates a total SN ratio of the generated T-method estimation model. Effective variables may be selected based on the magnitude of . For example, the selection unit 15 selects a variable included in a pattern with a maximum total SN ratio as an effective variable.

FIG. 7 is a diagram showing an example of the selection result of effective variables. In the example shown in FIG. 7, "AGC operating time," "number of devices," "number of personnel (parts shipment)," and "processing time" are selected as valid variables. Note that the vertical axis in FIG. 7 indicates the S/N ratio ηj of each variable in the T method estimation model using all of the plurality of variables. As described above, the SN ratio ηj of each variable represents the degree of influence on the standard SN ratio, which is the objective variable.

The analysis unit 16 shown in FIG. 4 performs principal component analysis (PCA) using actual values of effective variables and standard SN ratios for each unit time.

FIG. 8 is a diagram showing an example of an analysis result by the analysis section. FIG. 8 shows the results of principal component analysis for "AGC operating time", "number of devices", "number of personnel (parts shipped)", "processing time", and "standard SN ratio" for each unit time.

The analysis unit 16 selects two principal components that maximize the variance of the standard SN ratio. Specifically, the analysis unit 16 selects the first principal component with the largest contribution rate and the second principal component with the second largest contribution rate.

Alternatively, the analysis unit 16 displays a scatter diagram for each combination of two principal components among the plurality of principal components obtained by principal component analysis on the display device 70, and displays a scatter diagram for each combination of two principal components among the plurality of principal components obtained by the principal component analysis, and Two principal components may be selected. In the scatter diagram, each of the two principal components obtained from the values of "AGC operating time", "number of equipment", "number of personnel (parts shipped)", "processing time" and "standard SN ratio" for each unit time is Points corresponding to principal component scores are plotted. The analysis unit 16 may change the display format of each point in the scatter diagram depending on the value of the standard SN ratio corresponding to the point. For example, the analysis unit 16 displays in black points where the value of the standard SN ratio is within a first range (low range), and points where the value of the standard SN ratio is within a second range (medium range) that is larger than the first range. Points where the value of the standard SN ratio is within the third range (high range) larger than the second range are displayed in white. Thereby, the user only needs to check the scatter diagrams for each combination and select the scatter diagram in which the standard SN ratios are most separated.

As shown in FIG. 8, the analysis unit 16 displays a biplot 40 of the two selected principal components on the display device 70. In the biplot 40, principal component scores and principal component loadings are displayed in an overlapping manner. As described above, the display format of the points corresponding to the principal component scores of each of the two principal components obtained from the data for each unit time differs depending on the value of the standard SN ratio. From this, it is understood that the standard SN ratio is separated by using the two selected principal components.

The principal component loading indicates the correlation between the value of the variable and the principal component score. The larger the principal component loading amount, the stronger the correlation between the principal component and the variable. As shown in FIG. 8, the principal component loading amount for each variable is represented by a vector. Vector v0 indicates the principal component loading amount of the "standard SN ratio". Vector v1 indicates the principal component load amount of "AGC operating time". Vector v2 indicates the principal component loading amount of "number of devices". Vector v3 indicates the principal component loading amount of "number of personnel (shipped parts)". Vector v4 indicates the principal component load amount of "cutting time".

By checking the biplot 40, the user can understand the variables that are affecting the reduction in the standard SN ratio. That is, the user can specify, among the vectors v1 to v4, the variable corresponding to the vector with a large component parallel to the vector v0 as the variable that is affecting the reduction in the standard SN ratio. For example, in the biplot 40 shown in FIG. 8, the user selects the variables "AGV operating time" and "cutting time" corresponding to vectors v1 and v4 which are not orthogonal to the direction of vector v0 and have large magnitudes. can be identified as a variable that influences the reduction in the standard SN ratio.

The variables that are affecting the decrease in the standard SN ratio are considered to be the cause of the abnormality in the production line 2. Therefore, the user can eliminate the abnormality in the production line 2 by adjusting the variables that are affecting the reduction in the standard SN ratio.

The analysis unit 16 identifies and specifies the degree of influence of the effective variable on the standard SN ratio based on the principal component loading amount for the two selected principal components in each of the effective variable and the standard SN ratio. The degree of influence may be displayed on the display device 70. Specifically, for each of the plurality of effective variables, the analysis unit 16 indicates the vector (v1 to v4 in FIG. 8) indicating the principal component loading of the effective variable and the principal component loading of the "standard SN ratio". The absolute value of the inner product with the vector (vector v0 in FIG. 8) is calculated as the degree of influence on the standard SN ratio. Thereby, by checking the degree of influence displayed on the display device 70, the user can easily identify the variable that is influencing the reduction in the standard SN ratio.

<Processing flow of the abnormality detection device>
FIG. 9 is a flowchart showing an example of the process flow of the abnormality detection device. As shown in FIG. 9, the processor 101 of the abnormality detection device 100 first calculates standard values for each unit time of a plurality of variables indicating the status of the production line 2 when the production line 2 is operating normally. Acquire (step S1). In the present embodiment, the processor 101 inputs production plan data into the virtual model 2M, thereby acquiring the values of a plurality of variables for each unit time output from the virtual model 2M as the above-mentioned standard values.

Next, the processor 101 collects actual values of a plurality of variables for each unit time from the production line 2 (step S2). The processor 101 calculates a standard SN ratio representing the degree of deviation between the actual values of the plurality of variables with respect to the standard values of the plurality of variables for each unit time (step S3). The processor 101 displays a graph 30 showing changes in the standard SN ratio on the display device 70 (step S4).

Next, the processor 101 determines whether the standard SN ratio is outside the reference range (that is, less than the threshold TH) (step S5). If the standard SN ratio is within the reference range (that is, greater than or equal to the threshold TH) (NO in step S5), the process returns to step S2.

If the standard SN ratio is outside the reference range (YES in step S5), the processor 101 notifies that an abnormality has occurred in the production line 2 (step S6). For example, the processor 101 displays an error notification screen on the display device 70.

Next, the processor 101 selects effective variables from the plurality of variables that are effective for improving the estimation accuracy of the T method estimation model, in which the actual values of the plurality of variables are used as explanatory variables and the standard SN ratio is used as the objective variable. ). The processor 101 may display a screen showing the selection result on the display device 70. The effective variable is an important item for the estimation accuracy of the standard SN ratio. Therefore, by confirming the selection result, the user can identify the effective variable as a candidate for the cause of the abnormality in the production line 2. As a result, the user can perform maintenance on the production line 2 to adjust effective variables and review the production plan.

Next, the processor 101 performs principal component analysis using the actual values of the effective variables for each unit time and the standard SN ratio (step S8). The processor 101 displays the results of the principal component analysis on the display device 70 (step S9). For example, the processor 101 displays the biplot 40 shown in FIG. 8 on the display device 70. Alternatively, the processor 101 specifies the degree of influence of the effective variable on the standard SN ratio based on the principal component loading amount for each of the valid variable and the standard SN ratio, and displays the specified degree of influence on the display device 70. Good too. This allows the user to narrow down the variables that have a high degree of influence on the standard SN ratio from among the plurality of effective variables. As a result, the user can perform maintenance on the production line 2 to adjust the narrowed down variables. For example, the user performs maintenance on the cutting device 21a in the cutting process 21. Alternatively, the user may review the production plan. For example, the user can adjust the number of workers arranged in the assembly process 23, the number of cutting devices 21a operating in the cutting process 21, and the like.

Next, the processor 101 determines whether the production plan has been reviewed (step S10). Specifically, the processor 101 determines that the production plan has been reviewed in response to input of new production plan data.

If the production plan has been reviewed (YES in step S10), the process returns to step S1. As a result, the standard values for each unit time of a plurality of variables indicating the status of the production line are updated. If the production plan has not been reviewed (NO in step S10), the process returns to step S2.

<Modification 1>
FIG. 10 is a diagram showing a functional configuration of an abnormality detection device according to modification 1. The abnormality detection device 100A shown in FIG. 10 has the same hardware configuration as the abnormality detection device 100 (see FIG. 3). As shown in FIG. 10, the abnormality detection device 100A is different from the abnormality detection device 100 shown in FIG. The data set selection unit 17 is realized by the processor 101 executing the abnormality detection program 131. The performance database 18 is realized by the storage 103.

The performance database 18 includes one or more data sets indicating values of a plurality of variables per unit time when the production line 2 was operating normally in the past.

When the operating conditions of the production line 2 are variable, information indicating the operating conditions of the production line 2 may be added to each data set. The operating conditions include the number of cutting devices 21a operating in the cutting process 21, the number of workers in the parts shipping process 22, the number of workers in the assembly process 23, and the like.

The data set selection unit 17 operates as an acquisition unit that acquires standard values of a plurality of variables indicating the status of the production line 2 in a unit time when the production line 2 is operating normally. Specifically, the data set selection unit 17 selects one or more data sets from the performance database 18 in response to a user's instruction. For example, the user may specify one or more data sets from the performance database 18 when the production line 2 is operating under ideal conditions. Alternatively, if information indicating operating conditions is added to each data set, the user specifies one or more data sets to which information indicating operating conditions that match the production plan is added.

The data set selection unit 17 obtains, as standard values, the values of a plurality of variables for each unit time, which are indicated by the selected one or more data sets. When one data set is selected, the data set selection unit 17 acquires the value indicated by the data set as the standard value. When a plurality of data sets are selected, the data set selection unit 17 selects a representative value (for example, an average value, a median value, etc.) of the values for each unit time indicated by the plurality of data sets for each of the plurality of variables. Obtain as standard value.

According to the first modification, the data set selection unit 17 acquires, as standard values, the values of the plurality of variables per unit time when the production line 2 was operating normally in the past. Thereby, the standard SN ratio calculated by the calculation unit 13 represents the degree of deviation between the situation of the production line 2 when it was operating normally and the current situation. Therefore, the fact that the standard SN ratio is less than the threshold value means that some kind of abnormality has occurred in the production line 2. As a result, the user can easily check whether there is an abnormality in the production line by checking the detection result of the detection unit 14.

The process flow of the abnormality detection device 100A in Modification 1 is similar to the flowchart shown in FIG. 9 . Note that if the information indicating the operating conditions is not added to the data set, step S10 is omitted, and the process returns to step S1 after step S9.

<Modification 2>
In the embodiment shown in FIG. 1, a production line 2 is monitored at the production site. However, the object to be monitored is not limited to the production line 2. For example, the monitoring target may be a manufacturing device. Further, the abnormality detection device 100 according to the embodiment shown in FIG. 1 calculates a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables for each unit time. However, the abnormality detection device 100 may calculate the statistics for each unit section. The unit interval is, for example, a period during which the target operation is executed by the manufacturing apparatus.

(First example of manufacturing equipment)
FIG. 11 is a diagram illustrating an example of a manufacturing apparatus to be monitored. FIG. 11 shows an injection molding machine 3 as a manufacturing device to be monitored. The injection molding machine 3 includes a mold 31, a cylinder 32, a screw 33, a hopper 34, a screw drive device 35, and a sensor group 36.

The cylinder 32 is a cylindrical member and has an internal space into which resin is supplied. Hopper 34 supplies resin to the interior space of cylinder 32. A screw 33 is inserted into the internal space of the cylinder 32. A screw drive device 35 is connected to the base end of the screw 33. The screw 33 is rotated under the control of a screw drive device 35 and is movable in the longitudinal direction of the cylinder 32 . The sensor group 36 measures the values of various parameters indicating the status of the injection molding machine 3.

The injection molding machine 3 repeatedly performs injection molding operations. The injection molding operation is an operation in which the resin contained in the internal space of the cylinder 32 is injected into the mold 31 by moving the screw 33 from the upper position to the lower position in FIG.

FIG. 12 is a diagram showing changes in parameter values during a period in which an injection molding operation is performed. Hereinafter, the period during which the injection molding operation is performed will be referred to as an "injection molding period." In the graph shown in FIG. 12, the horizontal axis shows the elapsed time from the start of the injection molding operation, and the vertical axis shows the values of various parameters measured by the sensor group 36. As shown in FIG. 12, the sensor group 36 measures screw position, injection pressure, and injection speed as parameters indicating the status of the injection molding machine 3. The screw position is expressed, for example, by the distance D between the tip of the cylinder 32 and the tip of the screw 33 (see FIG. 11). In FIG. 12, a line 37 shows the evolution of the screw position during the injection molding period. Line 38 shows the evolution of the injection pressure. Line 39 shows the evolution of the injection speed.

When the monitoring target is the injection molding machine 3, the abnormality detection device according to the second modification uses a plurality of variables indicating the status of the injection molding machine 3 during the injection molding period when the injection molding machine 3 is operating normally. Get the standard value of. For example, the simulation execution unit 11 shown in FIG. 4 uses a virtual model of the injection molding machine 3 to execute a simulation of the values of a plurality of variables when the injection molding machine 3 is operated according to the production plan, and obtains results from the simulation. Obtain the values of multiple variables in the injection molding operation as standard values. Alternatively, the data set selection unit 17 shown in FIG. 10 acquires, as standard values, the values of a plurality of variables during an injection molding period when the injection molding machine 3 was operating normally in the past. The plurality of variables includes, for example, the screw position, injection pressure, and injection speed at each of the elapsed times t1, t2, . . . , tn from the start of the injection molding period. In this case, the plurality of variables includes 3×n variables.

When the monitoring target is the injection molding machine 3, the collection unit 12 acquires actual values of a plurality of variables indicating the status of the injection molding machine 3 for each injection molding period. The collection unit 12 may acquire actual values of a plurality of variables from the sensor group 36 or may acquire actual values of a plurality of variables from a control device (not shown) that controls the injection molding machine 3 (for example, a PLC). good.

FIG. 13 is a diagram showing standard values and actual values of a plurality of variables indicating the status of the injection molding machine 3. The line 37a represents the standard values of the screw positions (n variables) at the elapsed times t1, t2, . . . , tn from the start of the injection molding period. The line 38a represents the standard value of the injection pressure (n variables) for the elapsed times t1, t2, . . . , tn. The line 39a represents the standard value of the injection speed (n variables) for the elapsed times t1, t2, . . . , tn. A line 37b represents actual values of screw positions (n variables) at elapsed times t1, t2, . . . , tn from the start of the injection molding period. The line 38b represents actual values of the injection pressure (n variables) over the elapsed times t1, t2, . . . , tn. A line 39b represents actual values of the injection speed (n variables) at elapsed times t1, t2, . . . , tn.

The calculation unit 13, the detection unit 14, the selection unit 15, and the analysis unit 16 perform the processing described in the above embodiment using the standard values and actual values of the 3×n variables shown in FIG. For example, as shown in FIG. 13, the calculation unit 13 calculates a standard SN ratio as a statistic representing the degree of deviation of the actual values of 3×n variables from the standard values of 3×n variables.

FIG. 14 is a diagram showing changes in statistics calculated for each injection molding operation. FIG. 14 shows a graph in which the horizontal axis represents the number of injection molding operations and the vertical axis represents the statistical amount (standard SN ratio) calculated by the calculation unit 13. The seven digit number corresponding to each point in the graph represents the number of the product produced by the injection molding operation. Many defective products were found in products manufactured by injection molding operations in which the standard S/N ratio was outside the standard range (that is, below the threshold TH). From this, the user can check whether a defective product has occurred based on the error notification from the detection unit 14.

FIG. 15 is a diagram showing the results of principal component analysis using actual values of effective variables and standard SN ratios for each injection molding period. FIG. 15 shows a biplot of two principal components PC1 and PC2 selected by the analysis unit 16. In the biplot shown in FIG. 15, points corresponding to the principal component scores of each of the two principal components obtained from the actual values of the effective variables for each injection molding operation are plotted. Points corresponding to injection molding operations in which the standard SN ratio is within the reference range (that is, greater than or equal to the threshold value TH) are concentrated within the frame line 50a. On the other hand, points corresponding to injection molding operations where the standard SN ratio is outside the reference range (that is, less than the threshold TH) are concentrated within the frame line 50b. Therefore, by checking the biplot shown in FIG. 15, the user can easily identify the injection molding operation in which the abnormality has occurred.

Furthermore, by checking the biplot shown in FIG. 15, the user can consider the mechanism by which the abnormality occurs. In the biplot shown in FIG. 15, the variables "Speed_time_315 (injection speed at elapsed time t315)", "Pressure_time_315 (injection pressure at elapsed time t315)", "Speed_time_316 (injection speed at elapsed time t316)", "Speed_time_317 (elapsed time Vectors v10, v11, v12, v13, and v14 respectively corresponding to "injection speed at t317" and "Cylinder_time_318 (screw position at elapsed time t318)" are oriented in different directions from vectors corresponding to other variables. Therefore, the user can understand that the injection speed and screw position around the elapsed time t315 to t318 greatly contribute to the abnormality of the injection molding machine 3.

(Second example of manufacturing equipment)
FIG. 16 is a diagram showing another example of a manufacturing device to be monitored. In FIG. 16, a mounting machine 4 is shown as a manufacturing apparatus to be monitored. The mounting machine 4 includes a chuck 44, a drive device 45, and a sensor group 46.

The chuck 44 holds the chip 42 to which the bumps 43 are attached. The drive device 45 controls the generation of ultrasonic waves, moves the chuck 44 up and down, and controls the force with which the chuck 44 holds the chip 42 . The sensor group 46 measures the values of various parameters indicating the status of the mounting machine 4.

As shown in FIG. 16, the drive device 45 lowers the chuck 44 holding the chip 42. When the bumps 43 come into contact with the substrate 41, the mounting machine 4 performs a bonding operation to bond the bumps 43 and the substrate 41 together. Specifically, the drive device 45 generates ultrasonic waves to promote bonding between the bumps 43 and the substrate 41, and further lowers the chuck 44 while applying a load. When the bump 43 is deformed to a predetermined amount, the drive device 45 stops generating ultrasonic waves and reduces the holding force of the chip 42 by the chuck 44. This completes the joining operation. Thereafter, the drive device 45 raises the chuck 44. The mounting machine 4 performs the above bonding operation every time the chuck 44 holds a new chip 42.

The sensor group 46 measures the current and voltage applied to the drive device 45 and the amount of deformation of the bump 43 as parameters indicating the status of the mounting machine 4. The amount of deformation of the bump 43 is expressed by the distance between the chip 42 and the substrate 41.

When the monitoring target is the mounting machine 4, the abnormality detection device according to modification 2 detects the period during which the mounting machine 4 is operating normally and the bonding operation is performed (hereinafter referred to as the "bonding operation period"). ) to obtain the standard values of a plurality of variables indicating the status of the mounting machine 4. For example, the simulation execution unit 11 shown in FIG. 4 uses a virtual model of the mounting machine 4 to simulate the values of a plurality of variables when the mounting machine 4 is operated according to the production plan, and Obtain the values of multiple variables in an operation as standard values. Alternatively, the data set selection unit 17 shown in FIG. 10 acquires, as standard values, the values of a plurality of variables during a bonding operation period when the mounting machine 4 was operating normally in the past. The plurality of variables include, for example, the voltage, current, and amount of deformation at each of the elapsed times t1, t2, . . . , tm from the start of the bonding operation period. In this case, the plurality of variables includes 3×m variables.

When the monitoring target is the mounting machine 4, the collection unit 12 acquires actual values of a plurality of variables indicating the status of the mounting machine 4 for each bonding operation period. The collection unit 12 may acquire actual values of a plurality of variables from the sensor group 46, or may acquire actual values of a plurality of variables from a control device (for example, PLC), not shown, that controls the mounting machine 4. .

The calculation unit 13, the detection unit 14, the selection unit 15, and the analysis unit 16 perform the processing described in the above embodiment using the standard values and actual values of 3×m variables. For example, the calculation unit 13 calculates the standard SN ratio as a statistic representing the degree of deviation between the actual values of 3×m variables and the standard values of 3×m variables.

FIG. 17 is a diagram showing changes in statistics calculated for each joining operation. FIG. 17 shows a graph in which the horizontal axis is the joining operation number (mounting number) and the vertical axis is the statistic (standard SN ratio) calculated by the calculation unit 13. Many defective products were found in products manufactured by bonding operations in which the standard SN ratio was outside the reference range (that is, below the threshold TH). From this, the user can check whether a defective product has occurred based on the error notification from the detection unit 14.

(Another example of unit interval)
In the above example, the unit section is the period during which the operation of the manufacturing apparatus (injection molding operation or joining operation) is performed. However, the unit section is not limited to this. For example, the unit section may be each of a plurality of sections included in the range that the physical quantity of the manufacturing device can take.

For example, when the manufacturing device is the injection molding machine 3 shown in FIG. 11, the screw position is from the position of the screw 33 shown in the upper part of FIG. 11 (represented by distance D1) to the position of the screw 33 shown in the lower part of FIG. (represented by distance D2). In this case, each of a plurality of sections included in the range from distance D1 to distance D2 is set as a unit section. For example, a section where the distance D is D1 to D3 and a section where the distance D is D3 to D2 are set as unit sections.

Alternatively, a section in which a physical quantity obtained from a tensile tester or a compression tester installed at a production site falls within a specific range may be set as a unit section. For example, a section in which the amount of strain seen in a stress-strain diagram falls within a predetermined range is set as a unit section. Alternatively, in the relationship between the cooling rate and hardness of metal heated at a production site, a section in which the cooling rate falls within a predetermined range may be set as a unit section.

§3 Supplementary notes As described above, this embodiment includes the following disclosures.

(Configuration 1)
an acquisition unit (101, 11, 17) that acquires standard values for each unit time of a plurality of variables indicating the status of the production line (2) when the production line (2) is operating normally;
a collection unit (101, 12) that collects actual values of the plurality of variables for each unit time;
a calculation unit (101, 13) that calculates, for each unit time, a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
An abnormality detection device (100, 100A) comprising: a detection unit (101, 14) that detects that an abnormality has occurred in the production line in response to the statistical amount being outside a reference range.

(Configuration 2)
An acquisition unit that acquires standard values of a plurality of variables indicating the status of the monitoring target (2, 3, 4) in a unit interval when the monitoring target (2, 3, 4) at the production site is operating normally. (101, 11, 17) and
a collection unit (101, 12) that collects actual values of the plurality of variables in the unit section;
a calculation unit (101, 13) that calculates a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
An abnormality detection device (100, 100A) comprising: a detection unit (101, 14) that detects that an abnormality has occurred in the monitoring target in response to the statistical amount being outside a reference range.

(Configuration 3)
The abnormality detection device according to configuration 1 or 2, further comprising a selection unit (101, 15) that selects an effective variable that has a relatively large influence on the statistic from among the plurality of variables.

(Configuration 4)
Configuration 3, wherein the selection unit (101, 15) selects, as the effective variable, a variable that is effective in improving the estimation accuracy of the T method estimation model in which the plurality of variables are used as explanatory variables and the statistic is used as the objective variable. Anomaly detection device (100, 100A) described in .

(Configuration 5)
The abnormality detection device according to configuration 3 or 4 (100, 100A ).

(Configuration 6)
The abnormality detection device according to configuration 3 or 4 (100, 100A ).

(Configuration 7)
The analysis section (101, 16)
Select two principal components that maximize the variance of the statistics,
The abnormality according to configuration 5 or 6, wherein the degree of influence of the effective variable on the standard SN ratio is specified based on the principal component loading amount for the two principal components in each of the effective variable and the statistic. Detection device (100, 100A).

(Configuration 8)
The acquisition unit (101, 11)
Using a virtual model (2M) of the production line (2), performing a simulation of the values of the plurality of variables when the production line (2) is operated according to the production plan,
The abnormality detection device (100) according to any one of configurations 1 and 3 to 7, wherein the value for each unit time obtained by the simulation is acquired as the standard value.

(Configuration 9)
The acquisition unit (101, 11)
A virtual model (2M) of the monitoring target (2, 3, 4) is used to simulate the values of the plurality of variables when the monitoring target (2, 3, 4) is operated according to the production plan. ,
The abnormality detection device (100) according to any one of configurations 2 to 7, wherein the value of the unit interval obtained by the simulation is acquired as the standard value.

(Configuration 10)
Configurations 1 and 3, wherein the acquisition unit (101, 17) acquires, as the standard value, the values of the plurality of variables for each unit time when the production line (2) was operating normally in the past. 8. The abnormality detection device (100A) according to any one of 7 to 8.

(Configuration 11)
The acquisition unit (101, 17) is configured to acquire, as the standard value, the values of the plurality of variables in the unit interval when the monitoring target (2, 3, 4) was operating normally in the past. 8. The abnormality detection device (100A) according to any one of 2 to 7.

(Configuration 12)
The abnormality detection device (100, 100A) according to any one of configurations 1 to 11, wherein the statistic is a standard SN ratio.

(Configuration 13)
The monitoring target is a production line (2),
The abnormality detection device (100, 100A) according to any one of configurations 1 and 3 to 12, wherein the unit interval is a unit time.

(Configuration 14)
The monitoring target is an abnormality detection device (100, 100A) according to any one of configurations 2 to 12, including a manufacturing device (3, 4).

(Configuration 15)
The manufacturing device (3, 4) repeatedly executes the target operation,
The abnormality detection device (100, 100A) according to configuration 14, wherein the unit section is a period during which the target operation is executed.

(Configuration 16)
a step (S1) of obtaining standard values for each unit time of a plurality of variables indicating the status of the production line (2) when the production line (2) is operating normally;
a step (S2) of collecting actual values of the plurality of variables for each unit time;
a step (S3) of calculating, for each unit time, a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
An abnormality detection method comprising the step of detecting that an abnormality has occurred in the production line in response to the statistical value being outside a reference range (S6).

(Configuration 17)
A program (131, 132) that causes a computer (101) to execute an abnormality detection method,
The abnormality detection method is
a step (S1) of obtaining standard values for each unit time of a plurality of variables indicating the status of the production line (2) when the production line (2) is operating normally;
a step (S2) of collecting actual values of the plurality of variables for each unit time;
a step (S3) of calculating, for each unit time, a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
A program (131, 132) comprising a step (S6) of detecting that an abnormality has occurred in the production line in response to the statistical value being outside a reference range.

(Configuration 18)
Obtaining standard values of a plurality of variables indicating the status of the monitoring target (2, 3, 4) in a unit interval when the monitoring target (2, 3, 4) at the production site is operating normally ( S1) and
a step (S2) of collecting actual values of the plurality of variables in the unit interval;
calculating a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables (S3);
An abnormality detection method comprising: detecting that an abnormality has occurred in the monitoring target in response to the statistical value being outside a reference range (S6).

(Configuration 19)
A program that causes a computer to execute an anomaly detection method,
The abnormality detection method is
Obtaining standard values of a plurality of variables indicating the status of the monitoring target (2, 3, 4) in a unit interval when the monitoring target (2, 3, 4) at the production site is operating normally ( S1) and
a step (S2) of collecting actual values of the plurality of variables in the unit interval;
calculating a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables (S3);
A program (131, 132) comprising a step (S6) of detecting that an abnormality has occurred in the monitoring target in response to the statistical amount being outside a reference range.

Although the embodiments of the present invention have been described, the embodiments disclosed this time should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is indicated by the claims, and it is intended that all changes within the meaning and range equivalent to the claims are included.

1 system, 2 production line, 2M virtual model, 3 injection molding machine, 4 mounting machine, 11 simulation execution section, 12 collection section, 13 calculation section, 14 detection section, 15 selection section, 16 analysis section, 17 dataset selection section , 18 Performance database, 21 Cutting process, 21M 1st model, 21a Cutting device, 21b, 22c Timer, 21c, 21d, 22d, 23c Counter, 22 Parts shipping process, 22M 2nd model, 22a, 23a, 23b Sensor, 22b automatic guided vehicle, 23 assembly process, 23M third model, 24 product shipping process, 30 graph, 31 mold, 32 cylinder, 33 screw, 34 hopper, 35 screw drive device, 36, 46 sensor group, 40 biplot , 41 substrate, 42 chip, 43 bump, 44 chuck, 45 drive device, 70 display device, 75 input device, 100,100A abnormality detection device, 101 processor, 102 memory, 103 storage, 104 display controller, 105 input interface, 106 Communication interface, 131 Anomaly detection program, 132 Simulation program.

Claims (13)

an acquisition unit that acquires standard values of a plurality of variables indicating the status of the monitoring target in a unit interval when the monitoring target is operating normally at the production site;
a collection unit that collects actual values of the plurality of variables in the unit interval;
a calculation unit that calculates a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
An abnormality detection device comprising: a detection unit that detects that an abnormality has occurred in the monitoring target in response to the statistical amount being outside a reference range. The abnormality detection device according to claim 1, further comprising a selection unit that selects an effective variable having a relatively large influence on the statistic from among the plurality of variables. The abnormality according to claim 2, wherein the selection unit selects, as the effective variable, a variable that is effective for improving estimation accuracy of a T-method estimation model in which the plurality of variables are explanatory variables and the statistic is an objective variable. Detection device. The abnormality detection device according to claim 2, further comprising an analysis unit that performs principal component analysis using the actual value of the effective variable for each unit section and the statistical amount. The analysis department is
Select two principal components that maximize the variance of the statistics,
The anomaly detection device according to claim 4, wherein the degree of influence of the effective variable on the statistic is specified based on principal component loads for the two principal components in each of the effective variable and the statistic. . The acquisition unit includes:
Using the virtual model of the monitoring target, performing a simulation of the values of the plurality of variables when the monitoring target is operated according to a production plan,
The abnormality detection device according to claim 1, wherein the value of the unit interval obtained by the simulation is acquired as the standard value. The abnormality detection device according to claim 1, wherein the acquisition unit acquires, as the standard value, the values of the plurality of variables in the unit section when the monitoring target was operating normally in the past. The abnormality detection device according to claim 1, wherein the statistic is a standard SN ratio. The monitoring target is a production line,
The abnormality detection device according to any one of claims 1 to 8, wherein the unit interval is a unit time. The abnormality detection device according to claim 1 , wherein the monitoring target includes a manufacturing device. The manufacturing device repeatedly executes the target operation,
The abnormality detection device according to claim 10, wherein the unit interval is a period during which the target operation is executed. obtaining standard values of a plurality of variables indicating the status of the monitoring target in a unit interval when the monitoring target at the production site is operating normally;
collecting actual values of the plurality of variables in the unit interval;
calculating a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
An abnormality detection method comprising the step of detecting that an abnormality has occurred in the monitoring target in response to the statistical value being outside a reference range. A program that causes a computer to execute an anomaly detection method,
The abnormality detection method is
obtaining standard values of a plurality of variables indicating the status of the monitoring target in a unit interval when the monitoring target at the production site is operating normally;
collecting actual values of the plurality of variables in the unit interval;
calculating a statistic representing the degree of deviation of the actual values of the plurality of variables from the standard values of the plurality of variables;
A program comprising the step of detecting that an abnormality has occurred in the monitoring target in response to the statistical value being outside a reference range.

Images