JP4957406B2 - Batch process analysis system and batch process analysis method - Google Patents

Description

  The present invention relates to a batch process analysis system and a batch process analysis method for analyzing the status of a batch process based on data acquired from the batch process.

The process manufacturing industry that produces food / beverages, chemicals, clothing, petroleum, raw materials, etc. has industry-specific needs and involves complex processes. The manufacturing process is a batch process with a series of operations, and the following two methods are used for monitoring the batch process.
(1) Data superposition In order to determine whether or not the data change of the current batch process is normal, the current data is displayed as a trend, and superposed with ideal batch data and compared visually.
(2) Automatic comparison This method automatically compares the current batch process and ideal batch data. In this method, determination based on a data value threshold or determination based on a deviation threshold is performed.
JP-A-9-281042

  In order to construct such a system, it is necessary to create an alarm generation logic for detecting an abnormality during system setup. Also, when an alarm occurs, the operator checks the alarm generation logic and checks not only the device where the abnormality occurred, but also the devices that are likely to be related from the alarm generation logic, process knowledge, and past experience, and performs appropriate operations. It is required to execute.

  However, in the method of data superimposition, since a human makes a visual determination by superimposing data, the determination criterion is ambiguous. In addition, when a plurality of batch data are overlaid and compared, there is a risk of overlooking a small difference in visual determination. Furthermore, the number of data that can be overlaid and compared is limited, and a large number of data cannot be compared at one time.

  In addition, in the method based on automatic comparison, even if threshold determination is performed, unless the data value is significantly deviated, no abnormality is detected even if the data change form (trend shape) is different, and accurate determination cannot be performed. In addition, setting an appropriate threshold value is difficult and cumbersome, resulting in a large work burden.

  An object of the present invention is to provide a situation analysis system and a situation analysis method that can easily and quickly analyze the status of a batch process by using the Mahalanobis Taguchi method.

The batch process analysis system of the present invention is a batch process analysis system for analyzing the status of a batch process based on data acquired from the batch process, a data storage means for storing data obtained from the batch process, and the data Reference space creation means for creating a reference space based on the data accumulated in the accumulation means, data acquisition means for obtaining data from a batch process to be analyzed, and the reference space created by the reference space creation means The distance calculation means for calculating the Mahalanobis distance by inputting the data obtained by the data acquisition means, and the display means for graphically displaying the calculation result by the distance calculation means, the reference space creation means Is based on data stored in the data storage means and determined to be normal. The distance calculation means calculates Mahalanobis distance data in batch process units, and the display means calculates the plurality of batch processes of the same type by the distance calculation means. A plurality of Mahalanobis distances are graphically displayed in a comparable form .
According to this batch process analysis system, since the status of the batch process is analyzed by calculating the Mahalanobis distance, the status can be analyzed easily and quickly without relying on visual judgment.

The calculation means includes a contribution degree calculation means for calculating the contribution degree of the data of the batch process to the Mahalanobis distance calculated by the distance calculation means by calculation based on the contribution degree method in the Mahalanobis Taguchi method. You may display the trend graph of the said contribution calculated by the contribution calculation means.

The data acquisition means acquires data from a batch process to be analyzed from the batch start to the batch end based on information from the plant control system, and the distance calculation means is based on information from the plant control system. Then, after the end of the batch, the data obtained between the start of the batch and the end of the batch may be input to the reference space to calculate the Haranobis distance for the batch.

The batch process analysis method of the present invention includes a data accumulation step for accumulating data obtained from a batch process in the batch process analysis method for analyzing the status of the batch process based on data acquired from the batch process, and the data A reference space creation step for creating a reference space based on the data accumulated in the accumulation step, a data acquisition step for obtaining data from a batch process to be analyzed, and the reference space created by the reference space creation step A distance calculation step for calculating the Mahalanobis distance by inputting the data obtained in the data acquisition step, and a display step for graphically displaying a calculation result of the distance calculation step, the reference space creation step In the data storage step, A reference space that is determined to be normal is created based on the data that is determined to be normal stored by the group, and in the distance calculation step, Mahalanobis distance data is calculated in batch process units, and in the display step, the same kind of A plurality of Mahalanobis distances calculated in the distance calculation step for a plurality of batch processes are graphically displayed in a comparable form.
According to this batch process analysis method, since the status of the batch process is analyzed by calculating the Mahalanobis distance, the status can be analyzed easily and quickly without relying on visual judgment.

  According to the batch process analysis system of the present invention, since the status of the batch process is analyzed by calculating the Mahalanobis distance, the status can be analyzed easily and quickly without relying on visual determination.

  According to the batch process analysis method of the present invention, since the status of the batch process is analyzed by calculating the Mahalanobis distance, the status can be analyzed easily and quickly without relying on visual determination.

  Hereinafter, an embodiment of a batch process analysis system according to the present invention will be described with reference to FIGS.

  FIG. 1 is a block diagram illustrating a configuration of a plant control system to which a batch process analysis system according to an embodiment is applied.

  As shown in FIG. 1, the plant control system includes field controllers 2, 2,... Distributed in a plant and field devices 1, 2,. Are provided with a plant control server 3 that integrally controls the monitoring terminal device 4 for monitoring the plant. The field controllers 2, 2,... And the plant control server 3 are connected to each other via a network 7.

  The plant control server 3 is mounted with an application program for collecting data from the field devices 1, 1,... And controlling the field devices 1, 1,. By executing this program, the operator can monitor the state of the plant via the monitoring terminal device 4.

  In addition, a plant information management server 5 and a plant information management terminal device 6 constituting the batch process analysis system of the present embodiment are connected to the plant control system via a network 7.

  As shown in FIG. 1, the plant information management server 5 creates a reference space based on data storage means 51 for storing batch data obtained from a batch process and the batch data stored in the data storage means 51. The reference space creating means 52, the data obtaining means 53 for obtaining batch data from the batch process to be analyzed, and the batch data obtained by the data obtaining means 53 are input to the reference space created by the reference space creating means 52. Thus, a distance calculation unit 54 that calculates the Mahalanobis distance and a contribution calculation unit 55 that calculates the contribution of the batch data of the batch process are configured.

  FIG. 2 is a flowchart showing the operation of the batch process analysis system of this embodiment.

  Steps S1 to S5 in FIG. 2 show an operation procedure for monitoring an abnormality that occurs in the plant control system. This procedure is executed with the start of the batch process based on the control of the plant information management server 5.

  In step S <b> 1 of FIG. 2, batch data is acquired from the plant control server 3 via the data acquisition unit 53. The acquired data includes various process data collected from the field devices 1, 1,.

  Next, in step S2, the distance calculation means 54 inputs the batch data acquired in step S1 into the reference space created by the reference space creation means 52, calculates the Mahalanobis distance, and calculates the calculation result as plant information management. A graph is displayed on the screen of the terminal device 6.

  Next, in step S3, it is determined whether or not the calculated Mahalanobis distance is greater than a predetermined distance. This process is performed by executing the Mahalanobis Taguchi method for calculating the Mahalanobis distance between the normal data group stored in the data storage means 51 within the normal period and the data acquired in step S1. The similarity between the data acquired in step 1 and the normal data group is determined.

  If the determination in step S3 is affirmative, it is determined that the acquired data is abnormal, and the process proceeds to step S4. On the other hand, if the determination in step S3 is negative, it is determined that the acquired data is normal and the process proceeds to step S5. The threshold value of the Mahalanobis distance that is determined to be abnormal may be set by the user.

  FIG. 3 shows a graph display example of the Mahalanobis distance. When the Mahalanobis distance increases and exceeds the determination line, it is determined that there is an abnormality in step S3.

  In step S4, the operator is notified of the plant abnormality by the occurrence of an alarm, and the process returns to step S1.

  On the other hand, in step S5, the batch data (normal data) acquired in step S1 is stored in the data storage means 51, and the process returns to step S1. In this way, only normal data is stored in the data storage means 51.

  Steps S11 to S13 in FIG. 2 show a procedure for creating a reference space. This procedure is executed based on the control of the plant information management server 5.

  In step S11 of FIG. 2, a normal data group within a normal period stored in the data storage means 51 is acquired. Next, in step S12, the reference space creating means 52 creates a reference space based on the normal or ideal data group acquired from the data storage means 51. These data groups may be designated by the user, or normal data groups may be automatically stored in the data storage means 51. Accompanying accumulation of a normal data group as a material for creating the reference space, the abnormality / normality determination accuracy in step S3 is improved. This reference space is used in step S2.

  Next, in step S13, the reference space created in step S12 is stored, and the process returns to step S11.

  Thus, in this embodiment, the latest normal data sequentially stored in the data storage unit 51 can be reflected in the reference space by repeating the creation of the reference space. Note that the reference space may be updated in a timely manner without always repeating the creation of the reference space.

In the batch process analysis system according to the present embodiment, not only the abnormality of the ongoing batch process is notified, but also the already completed batch process is analyzed, and the analysis result is displayed on the screen of the plant information management terminal device 6. Can be displayed.
The batch data of the same kind of batch process is ideally always the same data. However, the actual batch data has a deviation from normal or ideal data, and the degree of abnormality of each batch data can be grasped as the Mahalanobis distance calculated for each data.

  In the batch process analysis system of the present embodiment, each abnormality degree is analyzed as a Mahalanobis distance for a plurality of batch data of the same type, and is displayed in a graph on the screen of the plant information management terminal device 6. As shown in FIG. 4A, batch data A, batch data B, and batch data C obtained from a plurality of batch processes of the same type have deviations from normal or ideal data.

  FIG. 4B is a display screen example showing the degree of abnormality of each batch data based on the Mahalanobis distance calculated for each batch data. In the example of FIG. 4B, the degree of abnormality of each batch data is shown by a graph, and the degree of abnormality of batch data C is higher than that of batch data A and batch data B. As described above, the batch process analysis system according to this embodiment can simultaneously analyze a plurality of batch data. In the example of FIG. 4B, the Mahalanobis distance is displayed as the degree of abnormality of each batch data so as not to make the operator aware of the Mahalanobis distance, but even if the Mahalanobis distance of each batch data is directly displayed. Good.

  Furthermore, in the batch process analysis system of the present embodiment, the degree of influence of each batch data on the abnormality is calculated for each time zone, and displayed on the screen of the plant information management terminal device 6 as a graph.

  In the batch process analysis system of the embodiment, when the calculation of the influence degree is requested for the batch data via the plant information management terminal device 6, the contribution degree calculation means 55 (FIG. 1) determines the batch data for each time zone. The degree of contribution (contribution) to the Mahalanobis distance is calculated, and the calculated contribution is displayed as the influence. The contribution method for calculating the contribution is a method for calculating the magnitude of the influence of individual data used for calculating the Mahalanobis distance in the Mahalanobis Taguchi method. By applying the contribution method to the data for each time zone, it is possible to identify the time zone that caused the occurrence of the abnormality.

  FIG. 5 is a diagram illustrating a screen display example of the influence degree by time zone. In this example, the time zone influence degree is displayed for batch data C (FIG. 4) having a high degree of abnormality. Thus, since the influence degree of batch data can be grasped | ascertained for every time slot | zone, an operator can analyze the cause and mechanism of abnormality occurrence easily. Instead of displaying the contribution level as the influence level, the contribution level for each time period may be directly displayed.

  As described above, in the batch process analysis system of the present embodiment, the similarity between specific batch data and normal or ideal batch data can be expressed as a numerical value by calculating the Mahalanobis distance, which is different from visual inspection. Objective and accurate determination is possible. The user does not need to check by overlaying data.

  Further, it is not necessary to set upper and lower limits and the like as in the threshold determination, and it is only necessary to determine batch data as a reference. The Mahalanobis-Taguchi method can be compared through the form of data changes, so it is possible to capture small changes in values that are difficult to set with thresholds. Also, the magnitude of the difference can be expressed by the value of Mahalanobis distance.

  Further, in the data superposition, the number of data that can be compared is limited. However, according to the present invention, it is possible to simultaneously analyze a lot of batch data by calculating the Mahalanobis distance and the contribution degree at a time.

  By the way, an operation support system that supports plant operation by performing process monitoring and operation monitoring of a batch process has a function of monitoring steady operation and unsteady operation, and the batch process analysis method according to the present invention is used as monitoring logic. Can be captured.

  FIG. 6 is a flowchart showing an operation example when the batch process analysis method according to the present invention is incorporated as the monitoring logic of the driving support system.

  In step S31 of FIG. 6, a batch start event occurs in the plant control system, and this is notified to the operation support system, and then the process proceeds to step S32.

  In step S32, batch data acquisition is executed. Next, in step S33, it is determined whether a batch end event has occurred in the plant control system and this has been notified to the operation support system. If this determination is affirmed, the process proceeds to step S34 to end the acquisition of batch data. If the determination is negative, acquisition of the batch data is continued in step S33.

  Next, in step S35, the Mahalanobis distance is calculated by inputting the acquired batch data into the already created reference space. The reference space is created based on normal or ideal batch data registered in advance. Next, in step S36, the analysis of the batch data is executed according to the instruction of the operator, and in step S36, the analysis result is output and the process is terminated. The analysis of batch data includes the calculation of contributions for each time period.

  Thus, by applying the batch process analysis method according to the present invention to the operation support system, it is possible to easily and accurately perform process monitoring and operation monitoring of the batch process. As a result of the analysis, when it is determined that the batch process is abnormal, an event notification may be sent to the plant control system and other systems. Further, the user may be notified of the abnormality by mail or alarm.

  As described above, according to the batch process analysis system of the present invention, since the status of the batch process is analyzed by calculating the Mahalanobis distance, the status can be easily and quickly analyzed without relying on visual determination.

  The scope of application of the present invention is not limited to the above embodiment. The present invention can be widely applied to scenes in which analysis based on data from a batch process is performed.

The block diagram which shows the structure of one Embodiment which applied the batch process condition analysis system by this invention to the plant control system. The flowchart which shows operation | movement of the batch process analysis system of this embodiment. The figure which shows the example of a display of the calculation result of Mahalanobis distance. It is a figure which shows the example of a display of an abnormal degree, (a) is a figure which shows several batch data of the same kind, (b) is the abnormal degree of each batch data based on the Mahalanobis distance calculated about each batch data Display screen example showing. The figure which shows the example of a screen display of the influence degree classified by time zone. The flowchart which shows the operation example at the time of taking in the batch process analysis method by this invention as monitoring logic of a driving assistance system.

Explanation of symbols

5 Plant information management server 51 Data storage means 52 Reference space creation means 53 Data acquisition means 54 Distance calculation means 55 Contribution calculation means

Claims (4)

In the batch process analysis system that analyzes the status of the batch process based on the data acquired from the batch process,
Data storage means for storing data obtained from the batch process;
Reference space creation means for creating a reference space based on the data stored in the data storage means;
Data acquisition means for acquiring data from the batch process to be analyzed;
Distance calculation means for calculating Mahalanobis distance by inputting the data obtained by the data acquisition means to the reference space created by the reference space creation means;
Display means for graphically displaying the calculation result by the distance calculation means;
With
The reference space creating means creates a reference space determined to be normal based on data determined to be normal stored in the data storage means,
The distance calculation means calculates Mahalanobis distance data in batch process units,
The batch display analysis system characterized in that the display means graphically displays a plurality of Mahalanobis distances calculated by the distance calculation means for a plurality of batch processes of the same type in a form that can be compared . By means of a calculation based on the contribution method in the Mahalanobis Taguchi method, it is provided with a contribution calculation means for calculating the contribution of batch process data to the Mahalanobis distance calculated by the distance calculation means for each hour,
The batch process analysis system according to claim 1, wherein the display unit displays a trend graph of the contribution calculated by the contribution calculation unit . The data acquisition means recognizes the period from the batch start to the batch end based on information from the plant control system, and acquires data from the batch process to be analyzed during this period,
The distance calculating means recognizes the end of the batch based on information from a plant control system, and inputs data acquired from the start of the batch to the end of the batch into the reference space after the end of the batch. The batch process analysis system according to claim 1, wherein the Halanobis distance is calculated for the batch. In the batch process analysis method for analyzing the status of the batch process based on the data acquired from the batch process,
A data accumulation step for accumulating data obtained from the batch process;
A reference space creation step of creating a reference space based on the data accumulated by the data accumulation step;
A data acquisition step for acquiring data from the batch process to be analyzed;
A distance calculating step of calculating a Mahalanobis distance by inputting the data obtained in the data acquisition step to the reference space created by the reference space creating step;
A display step for graphically displaying a calculation result of the distance calculation step;
With
In the reference space creation step, based on the data determined to be normal accumulated by the data accumulation step, create a reference space determined to be normal,
In the distance calculation step, Mahalanobis distance data is calculated in batch process units,
In the display step, a plurality of Mahalanobis distances calculated in the distance calculation step for a plurality of batch processes of the same type are graphically displayed in a form that can be compared.

Images